AI powered blockchain framework for predictive temperature control in smart homes using wireless sensor networks and time shifted analysis

Abstract

In the context of smart homes, efficiently managing temperature control while optimizing energy consumption and ensuring data security remains a significant challenge. Traditional thermostat-based systems lack predictive capabilities, and energy consumption often spikes during peak hours, leading to inefficiency. Additionally, the security of sensitive data in smart home environments is a growing concern. This paper presents a novel AI-powered blockchain framework for predictive temperature control in smart homes, leveraging wireless sensor networks (WSNs) and time-shifted analysis. The framework integrates machine learning (ML) algorithms for predictive temperature management, blockchain technology for secure data handling, and edge computing for real-time data processing, resulting in a highly efficient and secure system. Key innovations include the dynamic detection of heating and cooling events, predictive scheduling based on historical data, and blockchain-based decentralized energy trading. Performance evaluation demonstrates that the system accurately detects radiator heat-on events with a 28.5% success rate, while radiator cooling event detection achieves 37.3% accuracy. Scheduled heat-on events were triggered with 68.4% reliability, and the system’s machine learning component successfully reduced energy consumption by 15.8% compared to traditional thermostat controls, by adjusting heating based on predictive analysis. Additionally, the time-shifted data processing reduces peak-time computational load by 22%, contributing to overall energy efficiency and system scalability. The integration of blockchain ensures tamper-proof data security, eliminating unauthorized data access, and improving trust in smart home environments. These results illustrate the potential of combining AI, blockchain, and WSNs to create a robust, energy-efficient, and secure smart home temperature control system, offering significant improvements over traditional solutions.

Introduction

As the adoption of smart homes continues to expand globally, the need for efficient, secure, and autonomous temperature control systems has become increasingly important¹. Traditional thermostat-based systems, while effective in basic temperature regulation, lack the sophistication necessary to optimize energy usage and improve user comfort in dynamic environments². These systems often fail to account for real-time data, external factors, or user behavior patterns, leading to inefficient heating and cooling cycles that contribute to energy waste and increased costs³. Recent advancements in WSNs and Internet of Things (IoT) technologies have introduced new opportunities for more precise temperature control⁴. WSNs enable the continuous monitoring of environmental parameters, such as room temperature and radiator activity, within a smart home⁵. However, managing and processing the large volume of data generated by these networks, while maintaining data security and minimizing energy consumption, remains a significant challenge⁶.

Blockchain technology has emerged as a promising solution to address the issues of data security and privacy in IoT systems⁷. Its decentralized architecture ensures tamper-proof data storage, while smart contracts enable automated decision-making processes, such as adjusting temperature settings based on predefined conditions⁸. Despite these advantages, blockchain’s high computational demands can lead to latency and inefficiency when applied to real-time data processing in smart homes⁹. To further enhance the capabilities of smart home temperature control, ML techniques have proven effective in predictive analytics, enabling systems to anticipate user needs based on historical data¹⁰. By integrating ML models with blockchain and WSNs, it is possible to develop a more responsive and energy-efficient system that adjusts temperature proactively, reducing the need for manual intervention¹¹.

Furthermore, the integration of edge computing offers a viable solution to the problem of latency by processing data locally at the home level¹². This reduces the reliance on centralized cloud systems, enhancing real-time response and improving overall system performance¹³. Additionally, time-shifted data analysis can be employed to reduce computational loads during peak times, ensuring that energy consumption is optimized without compromising the accuracy or responsiveness of the system¹⁴. The integration of smart technologies in residential environments has been a rapidly growing field of research over the past decade, driven by the increasing demand for energy-efficient and automated systems in modern homes¹⁵. WSNs, IoT devices, and smart thermostats are among the key enablers of intelligent home systems, allowing for real-time monitoring and control of environmental conditions such as temperature, lighting, and security¹⁶. The deployment of these systems at scale has brought about new challenges related to data security, computational efficiency, and energy management, which have been the focus of recent research¹⁷. WSNs have been extensively studied for their ability to monitor and control environmental parameters in real time¹⁸. In the context of smart homes, WSNs enable continuous tracking of room temperature, humidity, and occupancy, providing the data needed for adaptive heating and cooling systems¹⁹. Several studies have demonstrated the effectiveness of WSNs in improving energy efficiency in smart homes. For example, research²⁰ showed that sensor-driven temperature control systems could reduce heating energy consumption by up to 20% when compared to traditional thermostat systems.

Literature survey

Blockchain technology has emerged as a robust solution²¹ for addressing security and privacy concerns in IoT systems²². By decentralizing data storage and securing transactions with cryptographic algorithms, blockchain ensures the immutability of data records and protects against unauthorized access^23,24 is among the first to explore the use of blockchain for securing IoT data in smart homes, proposing a lightweight framework that protects user data from external attacks. The use of smart contracts in blockchain further enhances automation by enabling devices to execute predefined actions autonomously, based on specified triggers²⁵. The computational intensity of blockchain, particularly its reliance on consensus mechanisms, can introduce delays that reduce system efficiency for real-time applications²⁶.The integration of blockchain technology and machine learning techniques to enhance the security management of 6G wireless networks is explored in²⁷. The application of blockchain technology to enhance various aspects of smart city infrastructure is examined in²⁸. It highlights how blockchain can improve security, transparency, and efficiency in urban systems such as energy management, transportation, and governance, enabling better data handling and fostering trust among citizens and stakeholders. The article²⁹ explores the integration of blockchain technology to enhance the security and privacy of smart devices in the IoT environment.

The application of ML for predictive temperature control has gained considerable attention in recent years due to its potential for improving system responsiveness and energy efficiency³⁰. ML algorithms can analyze historical temperature data, occupancy patterns, and even external factors such as weather to forecast heating or cooling needs, allowing systems to adjust preemptively³¹. Studies such as³² demonstrated that machine learning-based systems could reduce energy consumption by up to 18% compared to traditional reactive control systems, by predicting when heating or cooling is required based on user behavior. These predictive systems, however, require robust data³³ handling mechanisms to ensure that real-time and historical data are processed securely and efficiently. The integration of edge computing into smart home ecosystems has been proposed as a solution to the latency and bandwidth issues associated with centralized cloud processing³⁴. Edge computing allows data to be processed locally, closer to the source of data generation, which reduces the delays inherent in cloud-based systems³⁵. Research³⁶ highlighted the potential of edge computing in reducing latency and improving real-time decision-making in IoT systems, particularly in scenarios requiring immediate responses, such as smart temperature control. A transformative advancement in³⁷, showcasing significant improvements across several key dimensions of industrial operations. Overall, the synergy between AI and blockchain technologies³⁸ leads to a notable increase in productivity, operational reliability, and data security, setting a new standard for industrial excellence.

The integration of explainable AI with blockchain technology in³⁹ significantly enhances financial decision-making by addressing key issues of transparency and trust. The blockchain-modeled edge-computing-based smart home in⁹ demonstrates notable improvements in efficiency and security for smart home environments⁴⁰. The IoT-based smart home automation system utilizing blockchain and deep learning models in⁴¹ showcases impressive advancements in home automation, security, and efficiency. The differential privacy model integrated into a blockchain-based smart home architecture in²⁴ offers substantial improvements in user data privacy and system security. The BEDS (Blockchain Energy-Efficient IoE Sensors Data Scheduling) system significantly enhances the management and efficiency of data within smart home and vehicle applications in⁴². The collaborative approach of securing smart grid data using blockchain and WSNs in⁴³ demonstrates notable advancements in data integrity and system reliability. The BS-SCRM (Blockchain and Swarm Intelligence-Based Secure Wireless Sensor Networks) approach in⁴⁴ introduces a novel method for enhancing the security and efficiency of WSNs. The article¹⁴ presents an advanced approach to optimizing microgrid operations by balancing energy distribution and capacity scheduling. The article⁴⁵ explores an innovative approach to optimizing the performance of MHz wireless power transfer systems through time-shifted control techniques. A novel smart home system in⁴⁶ introduces that leverages a sophisticated algorithm for monitoring and managing the link status of WSNs. The link status awareness algorithm plays a crucial role in maintaining reliable communication between sensors and control systems by continuously assessing the quality and stability⁴⁷ of network connections. A sophisticated approach in⁴⁸ presents to optimizing renewable energy⁴⁹ use in smart homes through advanced forecasting and scheduling techniques. The effectiveness of combining multiple machine learning algorithms in⁵⁰ highlights to enhance the accuracy of energy consumption forecasts in smart homes.

The study⁵¹ investigates how trust influences the adoption and usage intentions of AI-powered smart home devices among younger generations. The study reveals that trust plays a critical moderating role in shaping users’ willingness to integrate these devices into their homes. The paper⁵² provides an overview of AI-driven energy management techniques, highlighting their applicability in optimizing smart home systems, particularly in temperature control and energy efficiency. The work⁵³ examines the role of blockchain technology in securing data exchanges in smart homes, focusing on the challenges and solutions for integrating blockchain in home automation systems. The paper⁵⁴ discusses the role of WSNs in smart home systems, particularly in real-time data collection and monitoring, which are essential for the proposed predictive temperature control framework. The study⁵⁵ presents various predictive control strategies, including machine learning approaches, to improve energy efficiency in smart homes, directly relevant to the predictive scheduling aspect of the proposed system. The research⁵⁶ explores the integration of blockchain for decentralized energy trading, which aligns with the proposed system’s feature of blockchain-based energy trading for smart homes.

The article⁵⁷ provides an extensive review of various data aggregation techniques used to optimize the performance of WSNs, focusing on reducing energy consumption, improving data accuracy, and enhancing network lifetime⁵⁸ explores the integration of cognitive agents in the IoT to enable context-aware data perception, enhancing the ability to adapt and respond intelligently to dynamic environmental conditions⁵⁹ presents a method for improving the accuracy of node localization in wireless sensor networks by utilizing mobile sinks and agent-based algorithms, enhancing the overall performance and scalability of the system⁶⁰ proposes a hybrid architecture combining centralized and peer-to-peer models to improve resource discovery and secure communication within the Internet of Things (IoT), offering enhanced scalability and reliability in IoT networks. The results of the comparative analysis of the articles reviewed are detailed in Table 1, which provides a comprehensive overview of the advancements and performance outcomes across various smart home and energy management technologies.

Table 1 Comparing this Article and related works.

The key contributions of the proposed approach are threefold: (1) the integration of AI and Blockchain for predictive temperature management and secure data handling, (2) the development of a novel framework that combines predictive scheduling and dynamic event detection, and (3) the evaluation of the system’s performance in terms of energy efficiency, security, and scalability.

Despite the significant advancements in smart home temperature control systems, several critical research gaps remain in the integration of emerging technologies such as blockchain, ML, WSNs, and edge computing. Current approaches either focus on isolated aspects of temperature management (e.g., predictive analytics or security) or lack the computational efficiency needed for real-time applications. These gaps are becoming more pronounced as smart home environments grow increasingly complex and demand secure, scalable, and energy-efficient solutions.

Research Gaps:

• 1. Limited Integration of Technologies: While blockchain has shown promise for securing IoT systems, few studies have explored its integration with predictive ML models for smart home temperature control. This gap limits the ability to create systems that are both secure and adaptive to user behavior and environmental conditions.

• 2. Lack of Predictive Control with Data Security: Traditional smart home systems often rely on historical data for reactive temperature control, lacking the predictive capabilities of machine learning. At the same time, ensuring data integrity and privacy in IoT environments remains a challenge, especially when dealing with sensitive home environment data.

• 3. Latency and Computational Bottlenecks: Centralized cloud-based systems used in many smart home applications face significant latency issues and computational bottlenecks, especially during peak data loads. These limitations hinder real-time control and scalability, which are critical for large-scale smart home deployments.

• 4. Energy Management and Optimization: Existing solutions typically focus on reducing energy consumption but do not fully exploit the potential of dynamic energy pricing or decentralized energy trading in smart homes, which could enhance both energy efficiency and user cost savings.

Contributions:

This paper addresses these research gaps by proposing a comprehensive, AI-powered blockchain framework for smart home temperature control that integrates WSNs, ML-based predictive analytics, and edge computing for time-shifted data processing. The key contributions are:

Integration of blockchain with predictive ML

A novel approach that combines blockchain technology with ML for predictive temperature control. Blockchain ensures secure data handling, while ML optimizes heating and cooling based on real-time and historical data.

Time-Shifted data processing with edge computing

The framework introduces edge computing to reduce latency and improve real-time responsiveness. By processing data locally and utilizing time-shifted analysis, the system decreases peak-time computational loads, enhancing overall performance.

Event detection and predictive scheduling

The system employs advanced WSNs to accurately detect radiator events (heat-on, cooling, etc.) and uses predictive models to schedule heating in a way that minimizes energy consumption, based on real-world data analysis.

Decentralized energy trading and dynamic pricing

The framework incorporates blockchain-enabled peer-to-peer energy trading and dynamic pricing models, allowing smart home users to trade surplus energy in a secure, decentralized marketplace. This feature optimizes energy usage while reducing costs.

Energy efficiency and scalability

By integrating ML for predictive control and using blockchain for secure decentralized management, the framework significantly improves energy efficiency, with reductions in heating energy, while ensuring scalability for broader smart home deployments.

For the proposed work in Predictive Temperature Control and Energy Consumption Management using Machine Learning, various AI and ML models can be employed to predict the temperature and energy consumption while optimizing HVAC system settings. Below is an overview of the AI and ML models used, along with recent papers related to this field:

AI and ML Models Used⁶¹:

• 1. Artificial Neural Networks (ANN): ANNs can be employed for time-series prediction of temperature and energy consumption. They learn complex nonlinear relationships from historical data, making them suitable for predicting energy demand and HVAC system control⁶², explores the use of deep neural networks for energy consumption prediction in smart homes, achieving high accuracy by integrating weather and occupancy data⁶³.

• 2. Support Vector Machines (SVM): SVMs can be used for regression tasks to predict continuous values, such as temperature or energy consumption⁶⁴. They work well with high-dimensional data, making them suitable for smart home data⁶⁵ investigates SVM-based models for predicting smart home energy usage, outperforming traditional regression models.

• 3. Random Forest (RF): Random Forest is an ensemble method that can be used for predicting temperature and energy consumption. It performs well with a large number of input features and is robust to overfitting⁶⁶⁶⁷ uses Random Forest for energy consumption forecasting and demonstrates energy-saving improvements is used .

• 4. Recurrent Neural Networks (RNN) / Long Short-Term Memory (LSTM): These models are particularly effective for time-series data and can be used to predict temperature and energy consumption based on historical time-dependent data⁶⁸⁶⁹ explores the use of LSTM models for predicting energy consumption in smart homes, providing superior results in terms of accuracy over traditional methods.

• 5. Decision Trees (DT): Decision Trees can help model the decision-making process for temperature control based on input data (e.g., occupancy, time of day, weather). They are interpretable and easy to implement⁷⁰⁷¹ applies decision trees to predict energy demand in residential buildings, incorporating feature selection to improve model accuracy.

The second part of the article focuses on Problem Formulation, addressing key components such as Predictive Temperature Control. The third section follows with the Results and Discussion, providing a detailed simulation analysis and comparison of the proposed method, while the fourth section offers the Conclusion for a final summary.

The proposed work

The rapid advancement of smart home technologies has highlighted the need for systems that not only provide comfort but also optimize energy usage and ensure data security. Traditional thermostat systems are limited in their ability to predict and adapt to varying environmental conditions and user behaviors, often resulting in inefficient heating schedules and increased energy consumption. Additionally, the centralized handling of sensitive data in these systems raises concerns about privacy and vulnerability to unauthorized access. To address these challenges, there is a pressing need for an integrated framework that combines predictive analytics, secure data management, and efficient processing capabilities. Specifically, the problem centers on developing a system that can accurately detect heating and cooling events, reliably trigger scheduled heat-on events, reduce energy consumption through predictive adjustments, and maintain data integrity and security within the smart home environment. This necessitates leveraging advanced technologies such as AI-powered machine learning algorithms, blockchain for secure and transparent data handling, wireless sensor networks for real-time environmental monitoring, and time-shifted analysis to optimize computational efficiency.

In this section, we define the mathematical framework and relationships necessary to develop an AI-powered blockchain framework for predictive temperature control in smart homes. The system aims to optimize temperature regulation while ensuring secure data handling and improving energy efficiency. The key components of the system include ML algorithms for prediction, blockchain for secure data management, WSNs for real-time monitoring, and time-shifted analysis for optimized computational efficiency.

Figure 1 illustrates the flowchart for solving the problem, outlining the step-by-step process and the sequence of operations involved. It provides a clear visual representation of the method used to address the issue.

Fig. 1

Flowchart depicting the problem-solving process.

The AI-Powered Blockchain Framework for Predictive Temperature Control in Smart Homes can be modeled as a holistic system integrating predictive control, energy optimization, and blockchain technology. The architecture can be represented as follows:

The system takes the following inputs:

$$\:\text{\:Inputs\:}=\left\{{T}_{\text{outdoor\:}}\left(t\right),{T}_{\text{current\:}}\left(t\right),{H}_{\text{historical\:}}\left(t\right),{U}_{\text{user\:}}\right\}$$

(1)

where $\:{U}_{\text{user\:}}$ represents user-defined preferences, such as the desired temperature range. The predictive model forecasts future temperature:

$$\:{T}_{\text{pred\:}}\left(t\right)=g\left({T}_{\text{current\:}}\left(t\right),{T}_{\text{outdoor\:}}\left(t\right),{H}_{\text{historical\:}}\left(t\right)\right)$$

(2)

with the control law ensuring:

$$\:{T}_{\text{desired\:}}-\delta\:\le\:{T}_{\text{actual\:}}\left(t\right)\le\:{T}_{\text{desired\:}}+\delta\:$$

(3)

The optimization function minimizes energy consumption:

$$\:\text{m}\text{i}\text{n}{E}_{\text{total\:}}={\int\:}_{{t}_{1}}^{{t}_{2}}\:{P}_{\text{heating\:}/\text{\:cooling\:}}\left(t\right)dt$$

(4)

subject to constraints on temperature stability. Sensor data and control signals are securely stored in a blockchain:

$$\:H\left({\text{T}\text{x}}_{i}\right)=\text{H}\text{a}\text{s}\text{h}\left({T}_{\text{sensor\:}},{P}_{\text{energy\:}},t\right)$$

(5)

ensuring immutability and transparency in the data. The system detects heating and cooling events dynamically:

$$\:{E}_{\text{event\:}}\left(t\right)=\frac{d{T}_{\text{actual\:}}\left(t\right)}{dt}$$

(6)

and uses historical patterns for scheduling:

$$\:S\left(t\right)=\text{a}\text{r}\text{g}\text{m}\text{i}\text{n}{D}_{\text{shifted\:}}\left(t\right)$$

(7)

The outputs are:

$$\:\text{\:Outputs\:}=\left\{{T}_{\text{actual\:}}\left(t\right),{P}_{\text{Optimized\:}}\text{},{E}_{\text{events\:}},\text{\:Blockchain\:State\:}\right\}$$

(8)

where $\:{P}_{\text{Optimized\:}}$​ represents optimized power usage and $\:{E}_{\text{events\:}}$​ identifies critical heating/cooling events.

This integrated system provides a robust, secure, and energy-efficient solution for temperature control in smart homes, demonstrating the novelty and practical relevance of the proposed framework.

Predictive temperature control

Let $\:T\left(t\right)$ represent the indoor temperature at time $\:t$, measured by the wireless sensors. The goal is to predict the future temperature $\:T(t+{\Delta\:}t)$ based on historical temperature data, user preferences, and environmental conditions.

To model the temperature dynamics, we can use a simple heat transfer equation:

$$\:\frac{dT}{dt}=\frac{1}{C}\left(P\left(t\right)-U\left(T\left(t\right)-{T}_{ext}\left(t\right)\right)\right)$$

(9)

$\:C$ is the thermal capacity of the room,$\:\:P\left(t\right)$ is the power supplied by the heating system at time $\:{t}_{r}$,$\:\:U$ is the overall heat transfer coefficient of the building and$\:\:{T}_{ext}\left(t\right)$ is the external temperature.

The ML model uses historical temperature data $\:\left\{T\left({t}_{1}\right),T\left({t}_{2}\right),\dots\:,T\left({t}_{n}\right)\right\}$ and corresponding energy consumption $\:P\left(t\right)$ to predict future temperature changes. The prediction function can be expressed as:

$$\:T(t+{\Delta\:}t)=f\left(T\left(t\right),P\left(t\right),{T}_{ext}\left(t\right)\right)$$

(10)

where $\:f$ is the learned function based on past data using a machine learning algorithm, such as Recurrent Neural Networks (RNNs) or Long Short-Term Memory (LSTM) models.

Energy consumption optimization

The objective is to minimize the energy consumption $\:E$, while maintaining user comfort, represented by a set temperature range $\:\left[{T}_{\text{min\:}},{T}_{\text{max\:}}\right]$.

The energy consumed by the heating system over a period $\:T$ is:

$$\:E={\int\:}_{0}^{T}P\left(t\right)dt$$

(11)

where $\:P\left(t\right)$ is the power required to maintain the temperature within the desired range. Using predictive temperature control, the heating system can adjust the power supply based on the forecasted temperature $\:T(t+\varDelta\:t)$, thereby reducing unnecessary energy use.

The system seeks to minimize $\:E$ under the constraint:

$$\:{T}_{min}\le\:T\left(t\right)\le\:{T}_{max}$$

(12)

Time-Shifted analysis

To reduce peak-time computational load, the system leverages time-shifted analysis, where non-urgent computations, such as data processing or historical analysis, are performed during off-peak times.

Let $\:{C}_{\text{p}\text{e}\text{a}\text{k}}$ ​be the computational load during peak hours, and $\:{C}_{\text{o}\text{f}\text{f}}$ ​ be the load during off-peak hours. Time-shifted analysis aims to minimize $\:{C}_{\text{p}\text{e}\text{a}\text{k}}$​ by shifting part of the workload to off-peak times. The relationship can be modeled as:

$$\:{C}_{\text{t}\text{o}\text{t}\text{a}\text{l}}={C}_{\text{p}\text{e}\text{a}\text{k}}+{C}_{\text{o}\text{f}\text{f}}$$

(13)

with the goal to reduce $\:{C}_{\text{p}\text{e}\text{a}\text{k}}$​, where:

$$\:{C}_{\text{p}\text{e}\text{a}\text{k}}={C}_{\text{t}\text{o}\text{t}\text{a}\text{l}}-\varDelta\:{C}_{\text{s}\text{h}\text{i}\text{f}\text{t}}$$

(14)

and $\:\varDelta\:{C}_{\text{s}\text{h}\text{i}\text{f}\text{t}}$ represents the load shifted to off-peak times. This reduces overall peak-time load by a percentage $\:\varDelta\:{C}_{\text{s}\text{h}\text{i}\text{f}\text{t}}/{C}_{\text{t}\text{o}\text{t}\text{a}\text{l}}$​.

Blockchain for secure data handling

For data security, blockchain technology is integrated to ensure tamper-proof and transparent data management. Each temperature reading and energy consumption record is stored as a block in the blockchain. Let $\:D\left(t\right)$ represent the data at time $\:t$ (e.g., temperature readings, energy consumption). The blockchain ensures that $\:D\left(t\right)$ cannot be altered once recorded.

The data is secured using a cryptographic hash function $\:H$, where:

$$\:H\left(D\left(t\right)\right)=hash\left(D\left(t\right)\right)$$

(15)

Each new block includes the hash of the previous block $\:H\left(D\right(t-1\left)\right)$, ensuring data immutability:

$$\:B\left(t\right)=\left\{\right(D\left(t\right),\:H\left(D\right(t-1\left)\right)\}\:$$

(16)

This chain of blocks guarantees that any attempt to tamper with historical data will be easily detected, as it would alter the hash values in subsequent blocks.

System optimization and customer satisfaction

The final objective is to optimize the system for both energy efficiency and user satisfaction. Let SSS represent customer satisfaction, which depends on maintaining the desired temperature range and minimizing energy costs. The overall optimization problem can be formulated as a multi-objective problem:

$$\:\text{\:minimize\:}\:E,\:\text{m}\text{a}\text{x}\text{i}\text{m}\text{i}\text{z}\text{e}\:S$$

(17)

subject to the constraints $\:{T}_{min}\le\:T\left(t\right)\le\:{T}_{max}$​ and secure data handling through blockchain. The system seeks to balance energy efficiency with user comfort and data security.

Dynamic event detection for heating and cooling

The system must dynamically detect heating events (e.g., radiator turning on) and cooling events (e.g., radiator turning off). These events can be modeled as binary occurrences based on the rate of temperature change over time.

Let $\:\varDelta\:T=T(t+1)-T\left(t\right)$ represent the change in temperature between time intervals. Define $\:H\left(t\right)$ as a heating event and $\:C\left(t\right)$ as a cooling event, which are triggered when certain thresholds are crossed:

• $\:H\left(t\right)=1$ if $\:{\Delta\:}T>{\Delta\:}{T}_{\text{heat-on,\:}}$ where $\:{\Delta\:}{T}_{\text{heat-on\:}}$ is the minimum temperature change to trigger a heating event.

• $\:C\left(t\right)=1$ if $\:{\Delta\:}T<{\Delta\:}{T}_{\text{cool-off\:}}$, where $\:{\Delta\:}{T}_{\text{cool-off\:}}$ is the maximum temperature drop to trigger a cooling event.

The framework can further be refined by employing machine learning models that dynamically learn from data to refine these thresholds and account for varying environmental conditions:

$$\:H\left(t\right)=1\text{\:if\:}{\Delta\:}T>{f}_{\theta\:}\left(T\left(t\right),{T}_{ext}\left(t\right),P\left(t\right)\right)$$

(18)

where $\:{f}_{\theta\:}$ is a learned function that dynamically adjusts thresholds based on environmental and system variables.

Predictive scheduling using historical data

To optimize energy usage, the system uses predictive scheduling based on historical data. The goal is to anticipate future temperature changes and schedule heating or cooling events accordingly.

Define $\:\mathcal{H}\left(t\right)=\left\{H\right(t-\tau\:),\dots\:,H(t\left)\right\}$ as the history of heating events over the time period $\:\tau\:$, and similarly, $\:\mathcal{C}\left(t\right)=\left\{C\right(t-\tau\:),\dots\:,C(t\left)\right\}$ as the history of cooling events. The system uses these historical patterns to predict future events:

$$\:P\left(H\right(t+1)=1)={f}_{\mathcal{H},\mathcal{C},\theta\:}\left(T\left(t\right),{T}_{ext}\left(t\right),P\left(t\right)\right)$$

(19)

where $\:P\left(H\right(t+1)=1)$ is the probability of a heating event occurring at time $\:t+1$, predicted using a machine learning algorithm trained on historical data $\:\mathcal{H}$ and $\:\mathcal{C}$.

The system then schedules heating and cooling events based on the predicted probabilities to minimize energy consumption while maintaining user comfort. If $\:P\left(H\right(t+1)=1)>{P}_{\text{threshold,\:}}$ the system preemptively triggers a heating event, reducing energy spikes.

Blockchain-Based decentralized energy trading

One of the key innovations is the integration of blockchain for decentralized energy trading. Users within a smart home network can buy or sell surplus energy generated from renewable sources (e.g., solar panels) using blockchain smart contracts.

Let $\:{E}_{prod}\left(t\right)$ represent the energy produced by a renewable energy source at time $\:t$, and $\:{E}_{cons}\left(t\right)$​ represent the energy consumed by the smart home at time $\:t$. The surplus energy is given by:

$$\:{E}_{surplus}\left(t\right)={E}_{prod}\left(t\right)-{E}_{cons}\left(t\right)$$

(20)

where $\:{E}_{surplus}\left(t\right)>0$ represents excess energy that can be sold, and $\:{E}_{surplus}\left(t\right)<0$ represents a deficit that can be compensated by purchasing energy.

Using blockchain smart contracts, energy transactions are automated between smart homes. Let $\:{p}_{buy}\left(t\right)$ and $\:{p}_{sell}\left(t\right)$ represent the buying and selling prices at time $\:t$. The smart contract automates the energy trade when:

$$\begin{gathered} {\text{if}}~{E_{{\text{surplus~}}}}\left( t \right)>0,{\text{~sell energy at price}},\;{p_{{\text{sell~}}}}\left( t \right), \hfill \\ {\text{if}}~{E_{{\text{surplus~}}}}\left( t \right)<0,{\text{~buy energy at price}}\;{p_{{\text{buy~}}}}\left( t \right). \hfill \\ \end{gathered}$$

(21)

Each energy trade is recorded on the blockchain, ensuring transparency and immutability. The smart contract logic can be formalized as:

$$\:\text{\:Energy\:trade\:contract\::\:}\left\{\begin{array}{ll}\text{\:If\:}{E}_{\text{surplus\:}}\left(t\right)>0,&\:\text{\:sell\:energy\:}\\\:\text{\:If\:}{E}_{\text{surplus\:}}\left(t\right)<0,&\:\text{\:buy\:energy\:}\end{array}\right.$$

(22)

The blockchain ledger ensures that energy trades are secured and logged without requiring a central authority, maintaining trust among users in the decentralized energy market.

Wireless sensor network optimization

WSNs play a vital role in real-time monitoring of temperature and environmental conditions in smart homes. However, optimizing the energy efficiency and reliability of the WSN itself is essential.

Let $\:N\left(t\right)$ represent the number of active sensors at time $\:t$, and $\:{P}_{s}\left(t\right)$ represent the power consumption of the WSN at time $\:t$. The goal is to minimize the power consumption of the WSN while maintaining sufficient sensor coverage.

The optimization problem can be formulated as:

$$\:\begin{array}{c}\text{m}\text{i}\text{n}\text{i}\text{m}\text{i}\text{z}\text{e}{P}_{s}\left(t\right)=\sum\:_{i=1}\:\:{P}_{\text{sensor\:}}\left(i\right)\\\:\text{\:subject\:to\:sensor\:coverage\:constraints:\:}{C}_{\text{m}\text{i}\text{n}}\le\:C\left(t\right)\end{array}$$

(23)

where $\:{P}_{\text{sensor\:}}\left(i\right)$ is the power consumption of the $\:i$-th sensor, and $\:C\left(t\right)$ represents the coverage of the WSN, which must exceed a minimum threshold $\:{C}_{\text{m}\text{i}\text{n}}$ for reliable temperature monitoring.

To reduce power consumption, time-shifted data analysis and adaptive sensing can be employed, where only a subset of sensors is active during certain periods, depending on predicted events.

The system can dynamically deactivate sensors when they are not required, using the predicted temperature changes $\:T(t+{\Delta\:}t)$ from the machine learning model. This reduces sensor power consumption:

$$\:N\left(t\right)={f}_{N}\left(T\right(t),P(t),C(t\left)\right)$$

(24)

The interaction between the edge server and IoMT devices involves a collaborative exchange of data and computational tasks, which ensures efficient operation in the system. Each IoMT device independently collects and processes local data, generating model parameters based on its specific environment and tasks. These parameters are periodically transmitted to the edge server for aggregation.

The edge server plays a pivotal role in this framework by acting as a central coordinator. It aggregates the model parameters received from multiple devices using advanced techniques, such as weighted averaging or federated optimization, depending on the importance and quality of the data from each device. This aggregation process ensures that the global model is continually updated while preserving the privacy of individual devices since raw data is never directly shared.

To manage real-time updates, the edge server employs a systematic communication protocol that prioritizes low-latency and secure data transfer. The server can handle asynchronous updates, allowing devices with varying computational and network capabilities to participate effectively. Additionally, the edge server uses error-checking and version.

Optimization of objective functions

To formulate the overall optimization problem, we aim to minimize energy consumption, $\:E$, and computational load, $\:{C}_{total}$, while maximizing user satisfaction, $\:S$, and ensuring secure data handling. This leads to a multi-objective optimization problem:

$$\:\text{m}\text{i}\text{n}\text{i}\text{m}\text{i}\text{z}\text{e}\,\mathcal{L}={\alpha\:}_{1}E+{\alpha\:}_{2}{C}_{\text{total\:}}-{\alpha\:}_{3}S$$

(25)

where $\:{\alpha\:}_{1},{\alpha\:}_{2}$, and $\:{\alpha\:}_{3}$ are weights representing the relative importance of energy consumption, computational efficiency, and user satisfaction.

Subject to the constraints:

$$\begin{aligned}&{T}_{\text{m}\text{i}\text{n}}\le\:T\left(t\right)\le\:{T}_{\text{m}\text{a}\text{x}}\\ &{C}_{\text{m}\text{i}\text{n}}\le\:C\left(t\right)\\ &\text{Blockchain}\text{\:security\:and\:immutability.\:}\end{aligned}$$

(26)

Blockchain-based data logs enhance system security by providing an immutable and tamper-proof ledger for recording all system transactions and events. Each data log is cryptographically secured, ensuring that once a block is added to the chain, it cannot be altered without the consensus of the network. This feature prevents unauthorized access and data manipulation. Additionally, the decentralized nature of blockchain eliminates single points of failure, making the system resilient against cyberattacks. By incorporating these secure data logs, the proposed framework ensures the integrity and confidentiality of temperature control data in smart homes, thereby fostering user trust and system reliability.

Advanced energy consumption optimization with constraints

The previous formulation provided a basic energy optimization model. We can enhance this by incorporating dynamic energy pricing and more granular control over energy usage based on real-time conditions.

Let $\:P\left(t\right)$ represent the dynamic price of energy at time $\:t$, which varies based on demand and supply in the energy market. The cost of energy consumption, $\:{C}_{E}$​, over a period $\:T$ can be expressed as:

$$\:{C}_{E}={\int\:}_{0}^{T}\:P\left(t\right)\cdot\:{P}_{\text{cons\:}}\left(t\right)dt$$

(27)

where $\:{P}_{\text{cons\:}}\left(t\right)$ is the power consumption at time $\:t$.

The objective is to minimize the energy cost $\:{C}_{E}$ while maintaining comfort, subject to dynamic pricing. Therefore, the optimization function becomes:

$$\:\text{m}\text{i}\text{n}\text{i}\text{m}\text{i}\text{z}\text{e}{C}_{E}={\int\:}_{0}^{T}\:P\left(t\right)\cdot\:{P}_{\text{cons\:}}\left(t\right)dt$$

(28)

subject to the constraints:

$$\begin{aligned}&{T}_{\text{m}\text{i}\text{n}}\le\:T\left(t\right)\le\:{T}_{\text{m}\text{a}\text{x}}\\ & {E}_{\text{cons\:}}\le\:{E}_{\text{m}\text{a}\text{x}}\end{aligned}$$

(29)

The dynamic pricing $\:P\left(t\right)$ can be modeled as a function of market conditions and predicted demand:

$$\:P\left(t\right)=\:{f}_{\text{d\:}}\left(D(t\right),S\left(t\right))$$

(30)

where $\:D\left(t\right)$ is the predicted demand and $\:S\left(t\right)$ is the available energy supply. Incorporating dynamic pricing incentivizes the system to reduce energy usage during peak periods and shift demand to off-peak times, which leads to cost savings.

Blockchain-Based consensus for secure data handling

Blockchain consensus mechanisms ensure the integrity of data within the system. Given the decentralized nature of smart homes, where each home or node is an independent agent, a consensus algorithm like Proof of Stake (PoS) or Delegated Proof of Stake (DPoS) is appropriate to validate transactions without high energy costs.

Let $\:D\left(t\right)$ represent a data block (e.g., sensor readings, energy trades), and let $\:V\left(t\right)$ represent the set of validators in the network at time $\:t$. Each validator $\:{\nu\:}_{i}\in\:V\left(t\right)\:$proposes a block $\:{B}_{i}\left(t\right)$, where the block contains a cryptographic hash of the previous block and the new data to be added.

The consensus algorithm requires that a majority of validators approve the new block. The total number of validators that approve a block $\:{B}_{i}\left(t\right)$ is denoted as $\:{A}_{i}\left(t\right)$. The consensus is reached when:

$$\:{A}_{i}\left(t\right)\ge\:\frac{\left|V\left(t\right)\right|}{2}$$

(31)

where $\:\left|V\left(t\right)\right|$ is the total number of validators. If the block is approved by the majority, it is added to the blockchain, ensuring data integrity.

Algorithm: Blockchain-Based consensus for smart home temperature control

Step 1: Initialization.

• Define the network nodes N={n₁,n₂,…,n_k}.

• Each node n_i​ maintains a local blockchain ledger L_i​.

• Initialize consensus threshold T (e.g., 51%).

Step 2: Data Collection.

• Each node collects temperature data D_i​ from its associated WSN.

• The data includes room temperature T_r​, radiator temperature T_h​, and time-stamped events.

Step 3: Block Proposal.

Node n_i​ prepares a candidate block B_i​ with:

• Sensor readings.

• Predicted temperature control actions.

• Previous block hash H_prev​.

Step 4: Validation.

• Broadcast B_i​ to all nodes in the network.

• Each node validates B_i​ by:

• Verifying sensor data integrity.

• Checking consistency with predicted control actions.

Step 5: Consensus Mechanism.

• Nodes perform a voting process.

• Accept or reject Bi​ based on validation.

• Count the votes V(B_i).

• If V(B_i) ≥ T, B_i is accepted and added to the blockchain.

Step 6: Blockchain Update.

• Update local ledger L_i​ to include the new block.

Step 7: Execution of Control Actions.

• Apply the predictive temperature control actions specified in B_i​.

Step 8: Repeat.

• Continue the process for the next data interval.

This algorithm ensures secure and decentralized management of temperature control in smart homes while leveraging blockchain for data integrity and trust.

Decentralized energy trading incentives

To encourage energy trading between smart homes, an incentive mechanism can be introduced based on a reward structure for participants who trade energy efficiently. Each user $\:{u}_{i}$ has a surplus $\:{E}_{\text{surplus\:}}\left(t\right)$ or deficit $\:{E}_{\text{deficit\:}}\left(t\right)$ of energy at time $\:t$, as discussed earlier. The system assigns rewards $\:{R}_{i}\left(t\right)$ to users based on their contribution to the energy market.

Let $\:N\left(t\right)$ be the total number of users in the network. The reward for user $\:{u}_{i}$ at time $\:t$ is proportional to the energy they contribute to the system and inversely proportional to the overall demand:

$$\:{R}_{i}\left(t\right)=\frac{{E}_{\text{surplus\:}}\left(t\right)}{\sum\:_{j=1}^{N\left(t\right)}\:\:{E}_{\text{deficit\:}}\left(t\right)}\cdot\:{R}_{\text{total\:}}\left(t\right)$$

(32)

where $\:{R}_{\text{total\:}}\left(t\right)$ is the total reward available at time $\:t$, which is determined by the blockchain network. This reward system encourages users to contribute surplus energy to the grid, promoting decentralized energy management.

Time-Shifted load balancing with Priority-Based control

In the proposed system, time-shifted load balancing optimizes computational resources by deferring non-critical computations to off-peak times. This is particularly useful for resource-constrained smart home devices. The framework uses priority-based control to assign priority levels to tasks.

Let $\:\mathcal{T}\left(t\right)=\left\{{T}_{1}\left(t\right),{T}_{2}\left(t\right),\dots\:,{T}_{n}\left(t\right)\right\}$ represent the set of tasks at time $\:t$, and let $\:{P}_{i}\left(t\right)$ be the priority of task $\:{T}_{i}\left(t\right)$. Tasks with lower priority are deferred to off-peak times, thereby reducing peaktime computational load.

The optimization objective is to minimize the peak computational load $\:{C}_{peak}$ by shifting lowerpriority tasks. The load is reduced according to the equation:

$$\:{C}_{\text{peak\:}}=\sum\:_{i=1}^{{N}_{\text{high\:}}\left(t\right)}\:C\left({T}_{i}\left(t\right)\right)+\sum\:_{i={N}_{\text{high\:}}\left(t\right)+1}^{N\left(t\right)}\:C\left({T}_{i}\left(t\right)\right)\cdot\:\delta\:\left({T}_{i}\left(t\right)\right)$$

(33)

where $\:{N}_{\text{high\:}}\left(t\right)$ is the number of high-priority tasks, $\:C\left({T}_{i}\left(t\right)\right)$ is the computational cost of task $\:{T}_{i}\left(t\right)$, and $\:\delta\:\left({T}_{i}\left(t\right)\right)$ is the binary variable indicating whether task $\:{T}_{i}\left(t\right)$ has been deferred to offpeak times.

In a decentralized smart home network, multi-agent collaboration allows multiple homes to collaborate in managing energy and computational load. Each smart home is treated as an agent $\:{a}_{i}$​, and the collaboration aims to minimize total system energy consumption while maintaining comfort across all agents.

Let $\:{E}_{i}\left(t\right)$ represent the energy consumption of agent $\:{a}_{i}$ at time $\:t$. The total energy consumption $\:{E}_{\text{total\:}}\left(t\right)$ of the network is the sum of energy consumption across all agents:

$$\:{E}_{total}\left(t\right)=\sum\:_{i=1}^{N}\:{E}_{i}\left(t\right)$$

(34)

Each agent shares its load with others, reducing peak demand. The collaborative optimization function is:

$$\:\text{m}\text{i}\text{n}\text{i}\text{m}\text{i}\text{z}\text{e}{E}_{\text{total\:}}\left(t\right)=\sum\:_{i=1}^{N}\:{E}_{i}\left(t\right)\cdot\:{w}_{i}\left(t\right)$$

(35)

where $\:{w}_{i}\left(t\right)$ is the weight assigned to each agent based on their energy-sharing contribution.

The collaboration ensures that energy is distributed efficiently, and peak demand is reduced by sharing surplus energy between homes in the network.

The proposed system can implement an energy-aware control algorithm to manage heating and cooling based on real-time predictions and sensor data. The control algorithm calculates the optimal heating or cooling schedule by predicting the energy consumption required to maintain the desired temperature.

Define the control signal $\:u\left(t\right)$ that represents the power adjustment made by the system at time $\:t$. The control algorithm minimizes energy consumption while maintaining comfort within a defined range $\:\left[{T}_{\text{m}\text{i}\text{n}},{T}_{\text{m}\text{a}\text{x}}\right]$:

$$\:\text{m}\text{i}\text{n}\text{i}\text{m}\text{i}\text{z}\text{e}J={\int\:}_{0}^{T}\:\left(P\left(t\right)+\lambda\:{\left(T\left(t\right)-{T}_{set}\right)}^{2}\right)dt$$

(36)

where $\:\lambda\:$ is a penalty term for deviations from the setpoint temperature $\:{T}_{\text{set\:}}$.

The control signal is updated based on the following relationship:

$$\:u(t+1)=u\left(t\right)-\eta\:\nabla\:J$$

(37)

where $\:\eta\:$ is the learning rate and $\:\nabla\:J$ is the gradient of the objective function with respect to the control signal.

This adaptive control algorithm ensures that the system learns over time, adjusting energy usage to maintain the desired temperature while minimizing cost.

Algorithm: predictive temperature control and energy consumption using machine learning

Data collection

Collect historical temperature data, energy consumption data, weather conditions, and occupancy patterns from IoT sensors (e.g., temperature sensors, occupancy sensors, weather forecasts). Collect the system parameters such as heating and cooling system efficiency, energy consumption rates, and other relevant data.

Data preprocessing

Clean the data by handling missing values, removing outliers, and normalizing/standardizing the data. Create additional features based on historical data (e.g., moving averages of temperature, occupancy trends, etc.).

Model training

Select an appropriate ML model (e.g., Decision Trees, Random Forest, Support Vector Machines, Neural Networks, etc.). Split the data into training and testing sets. Train the ML model using the historical temperature and energy consumption data, incorporating the weather and occupancy data as input features. Optimize model hyperparameters for better performance.

Temperature prediction

Use the trained ML model to predict future temperature based on current temperature, weather forecast, and occupancy patterns. Predict the temperature setpoints for future hours or days based on this analysis.

Energy consumption prediction

Predict energy consumption for the heating and cooling system using the model based on the predicted temperature and occupancy data. Optimize the energy consumption prediction by adjusting the system’s heating/cooling demand to match predicted temperature deviations.

Temperature control decision

Compare the predicted temperature to the desired temperature setpoint. If the predicted temperature is above or below the target, trigger the HVAC (Heating, Ventilation, and Air Conditioning) system to adjust. Adjust the heating/cooling system settings to bring the temperature closer to the desired setpoint while minimizing energy consumption.

Optimization of energy consumption

Apply energy-efficient strategies such as predictive scheduling (heating/cooling during off-peak times), adjusting setpoints based on predicted trends, or controlling HVAC systems based on occupancy data. Use reinforcement learning techniques, if applicable, to adapt and optimize the temperature control and energy usage over time.

Real-time adaptation

Continuously monitor and update predictions using real-time data from IoT sensors, modifying the heating/cooling strategy as needed. Re-train the model periodically with new data to ensure that the system stays accurate.

Output

The system generates optimal heating/cooling schedules and real-time control adjustments. Display energy consumption predictions and provide recommendations for further optimization.

Evaluation

Evaluate the model’s accuracy by comparing predicted energy consumption and temperature control results against actual outcomes. Use performance metrics like MAE, RMSE, or energy savings percentage to assess the performance of the predictive system.

The algorithm leverages machine learning models such as decision trees, neural networks, or other time-series models, combined with optimization strategies like predictive scheduling and reinforcement learning, to efficiently manage energy in smart homes. By integrating IoT with real-time sensor data and weather forecasting, it predicts temperature and energy consumption trends. This enables optimal HVAC scheduling, minimizing energy usage while maintaining comfort, ultimately achieving intelligent temperature control and energy optimization based on both real-time and historical data.

Simulation model

To begin the Results and Discussion section, the analysis of the simulation results focuses on evaluating the effectiveness and performance of the proposed AI-powered blockchain framework for predictive temperature control in smart homes. The results are presented in a structured manner, examining key parameters such as accuracy, efficiency, and energy savings, as well as the security benefits provided by the blockchain integration.

The framework’s ability to predict heating and cooling events based on historical data is analyzed, with particular attention given to the system’s detection rates for radiator heat-on, cooling events, and scheduled heating events. These results are benchmarked against traditional thermostat control methods to highlight the improvements achieved through predictive machine learning algorithms. Additionally, the framework’s energy consumption reduction is quantified, demonstrating the impact of predictive scheduling on overall energy efficiency.

Blockchain’s role in securing wireless sensor network data and enabling decentralized energy trading is also discussed. The system’s tamper-proof nature is examined, along with its ability to prevent unauthorized access to sensitive data, thereby improving trust and transparency in smart home environments.

The discussion will also touch upon the scalability of the system, particularly the time-shifted data processing method, which significantly reduces peak computational loads. This reduction in processing demand ensures that the system remains efficient, even in large-scale smart home deployments. The results will be compared to existing solutions in the literature, highlighting the innovation and contributions of the proposed framework.

Each of these aspects is discussed in detail to provide a comprehensive understanding of how the AI-powered blockchain framework enhances both operational efficiency and security in smart homes.

Simulation procedure

Table 2 outlines the steps of the proposed algorithm, providing a clear roadmap for the process from input data collection to output generation. Each step corresponds to a specific phase in the methodology, with the relevant equations listed alongside. Starting with the collection of input data, the algorithm progresses through system modeling, parameter estimation, adaptive control, optimization, cybersecurity integration, real-time monitoring, and performance evaluation. For each of these stages, mathematical equations are used to define and adjust system parameters, ensuring that the model is optimized and resilient against potential threats. The final output is the result of optimized control signals or updated system parameters, ensuring efficient and stable performance. This systematic approach ensures that all crucial factors such as parameter estimation, optimization, security, and real-time monitoring are integrated effectively.

Table 2 Overview of the proposed algorithm with corresponding equations for each step.

• 1. Experimental Setup: The experimental setup used in the study involved both simulation and real-time testing. The system was tested in a controlled environment where the proposed AI-powered blockchain framework for predictive temperature control was implemented. Real-time data from WSNs deployed in a simulated smart home environment were collected to evaluate the performance of the system. The simulation models included a combination of environmental variables such as temperature, humidity, and energy consumption patterns. The real-time scenarios were designed to mimic typical smart home temperature control situations, with heating and cooling events, predictive scheduling, and energy trading based on historical data.

The experimental setup can be described as follows:

• Environment: A simulated smart home with wireless sensors monitoring room temperature, humidity, and radiator status.

• Data Collection: Real-time data logging via WSNs for temperature, humidity, and event detection (heating, cooling).

• Technology Stack: AI algorithms for predictive temperature control, blockchain for secure data handling, and edge computing for real-time data processing.

• Performance Metrics: Success rates for event detection (heat-on, cooling, scheduled heat-on), energy consumption savings, and time-shifted load balancing.

• 2. Real-Time Scenarios: The system works with real-time scenarios. The integration of real-time data processing with edge computing ensures that the temperature control adjustments occur dynamically based on the collected data. This means that the system can make decisions and update control actions (such as turning the radiator on/off or adjusting the temperature) in real time. Furthermore, the blockchain component allows for secure, tamper-proof data logs in real-time, ensuring system integrity.

To illustrate the real-time performance, a time series of temperature data, event detection, and energy consumption adjustments were continuously monitored and updated during the experiments, reflecting how the system responds to changes in the environment and its effectiveness in optimizing energy consumption.

• 3. Algorithmic Procedure: The algorithmic procedure for the proposed system can be described in the following steps:

• Step 1: Data Collection - Real-time data from WSNs (temperature, humidity) are collected.

• Step 2: Predictive Control - Machine learning models (e.g., Random Forest, Support Vector Machines) are used to predict the heating or cooling needs based on historical data.

• Step 3: Event Detection - The system detects heating and cooling events using a combination of predictive models and real-time sensor data.

• Step 4: Blockchain Integration - Blockchain is used to record sensor data and decisions in a secure, tamper-proof manner.

• Step 5: Scheduling & Optimization - Predictive scheduling is applied to optimize energy consumption by adjusting the heating schedule and balancing energy loads.

• Step 6: Feedback Loop - Based on the system’s performance (temperature control and energy savings), adjustments are made, and results are logged in real time.

The specific algorithms used for event detection, predictive scheduling, and blockchain consensus are as follows:

• Predictive Algorithms: Random Forest for event prediction, Neural Networks for dynamic temperature adjustments.

• Blockchain Consensus: Proof-of-Authority (PoA) consensus for validating data integrity and ensuring secure energy trading.

• 4. System Complexity: The system’s complexity can be measured using several parameters, including:

• Computational Complexity: This can be evaluated by analyzing the time complexity of the predictive algorithms (e.g., O(n log n) for Random Forest).

• Energy Consumption: The system’s energy efficiency can be measured by comparing energy consumption (before and after optimization) and calculating the percentage of energy savings achieved.

• Scalability: The scalability of the system is assessed by testing its performance with an increasing number of devices and sensors in the network. A scalability metric can be defined based on system responsiveness and energy savings as the network size grows.

• Latency: The time delay in decision-making (e.g., the time it takes for the system to detect an event and adjust the temperature) can be measured to assess real-time performance.

The system’s complexity is influenced by the machine learning models’ training time, the number of sensors in the WSNs, and the blockchain’s consensus mechanism.

Simulation parameters

The dataset was collected from a real-world smart home setup, equipped with a WSN and a variety of IoT devices. Data was recorded from January to June 2024 across multiple rooms, including the living room, bedroom, and kitchen, to ensure variability in temperature and energy usage patterns. The dataset includes temperature measurements, energy consumption readings, and radiator operational status (on/off) synchronized with real-time external weather data.

The dataset includes the following key features that are crucial for training and testing the predictive models, performing time-shifted analysis, and optimizing energy consumption in Table 3:

Table 3 Dataset features and source.

Total records: 25,920 data points (representing one data point every minute for a period of 18 days). Data is structured in time series, with a total of 6 months of data from multiple rooms, offering a rich set of variations in temperature, energy consumption, and heating events.

To ensure the quality and consistency of the data used for the predictive model and optimization algorithms, the following preprocessing steps were applied:

Missing temperature and energy consumption values were interpolated using a linear interpolation technique to fill in any gaps due to sensor failure or communication issues. For example, if the indoor temperature $\:{\varvec{T}}_{\text{i}\text{n}\text{d}\text{o}\text{o}\text{r}}\:$for the time 12:30 PM is missing, it is linearly interpolated from the neighboring values 12:00 PM and 1:00 PM.

Outliers were identified using the Interquartile Range (IQR) method. Any value higher than 3 times the IQR from the upper quartile (e.g., values > 10 kWh in a 2-hour period) was removed to prevent distortion of the analysis and prediction accuracy.

All temperature values were normalized using Min-Max scaling. For instance, if the range of indoor temperatures is from 18 °C to 28 °C, the normalized indoor temperature is calculated as $\:{T}_{norm}=\frac{{T}_{indoor}-18}{28-18}$.

Temporal features such as time of day (morning, afternoon, evening), weekday/weekend status, and holidays were extracted to enhance the predictive models. These features were used to adjust heating predictions during high-demand periods like weekends or holidays.

The dataset was used for multiple purposes within the framework:

Historical data was employed to train machine learning models (such as LSTM or Random Forest) to predict the indoor temperature and heating durations, providing predictive scheduling for energy-efficient heating control.

Blockchain technology was used to securely log and share data related to temperature and energy usage across decentralized nodes within the smart home network, ensuring tamper-proof and traceable data handling.

The radiator status and event detection were analyzed using the dataset to identify the exact times heating events occur (heat-on, cooling, and heat-off events), which are integral to time-shifted load balancing.

Figure 2 illustrates the neural network model, including the input, hidden, and output layers. The structure of the network is designed to process input data efficiently and optimize system performance through adaptive learning. Additionally, the figure presents the neural network training process using the nntraintool, which provides a comprehensive visualization of the training parameters, performance metrics, and convergence behavior. All essential details of the model, including activation functions, weight adjustments, and training iterations, are clearly depicted, ensuring a thorough understanding of the network’s functionality and optimization process.

Fig. 2

Neural network model with nntraintool.

Scenario: predictive temperature control

Consider a smart home equipped with a WSN for monitoring temperature and a heating system that responds to temperature changes. The AI-powered framework uses historical temperature data, real-time sensor readings, and predictive machine learning algorithms to control the heating system more efficiently. At the same time, the system leverages blockchain to ensure secure data handling.

Step 1

Modeling the Temperature Dynamics

First, the room temperature dynamics need to be modeled based on heat transfer principles. The rate of temperature change in the room can be modeled as:

$$\:\frac{dT\left(t\right)}{dt}=\frac{1}{C}\left(P\left(t\right)-U\left(T\left(t\right)-{T}_{\text{outside\:}}\right)\right)$$

(38)

$\:T\left(t\right)$ is the room temperature at time $\:{t}_{\:}$,$\:\:P\left(t\right)$ is the heating power provided by the system at time $\:{t}_{r}$, $\:{T}_{\text{outside\:}}$ is the outside temperature,$\:\:U$ is the heat loss coefficient (related to insulation), and$\:\:C$ is the heat capacity of the room.

This differential equation governs how the room temperature changes over time based on heating power and heat loss to the environment.

Step 2

Machine Learning for Predictive Control

To optimize energy use, a machine learning model (e.g., a recurrent neural network, RNN) is trained using historical temperature data and heating system responses. The goal is to predict the future temperature $\:T(t+1)$ based on current and past data, allowing for pre-emptive heating adjustments.

Let the inputs to the model at time $\:t$ be $\:X\left(t\right)=\left\{T\left(t\right),P\left(t\right),{T}_{\text{outside\:}}\left(t\right)\right\}$, and the output be the predicted temperature $\:\hat{T}(t+1)$. The trained model minimizes the prediction error:

$$\:\text{\:Loss\:}=\sum\:_{t=1}^{N}\:\left(T\right(t+1)-\hat{T}(t+1){)}^{2}$$

(39)

where $\:N$ is the number of time steps in the historical dataset.

The predictive control algorithm then adjusts the heating power $\:P\left(t\right)$ to maintain the desired temperature $\:{T}_{\text{set\:}}$ within a predefined comfort range $\:\left[{T}_{\text{m}\text{i}\text{n}},{T}_{\text{m}\text{a}\text{x}}\right]$. The control signal is updated as:

$$\:P(t+1)=P\left(t\right)+{K}_{p}\left({T}_{\text{s}\text{e}\text{t}}-T\left(t\right)\right)+{K}_{d}\left(\frac{dT\left(t\right)}{dt}\right)$$

(40)

where $\:{K}_{p}$ and $\:{K}_{d}$​ are proportional and derivative control gains, respectively. This feedback control loop ensures that the heating system responds dynamically to temperature predictions, optimizing energy use.

Step 3

Blockchain for Data Security

To ensure that all data exchanged between smart home devices (e.g., sensors, heating systems) is secure, the system integrates blockchain. Each temperature reading and heating event is recorded as a transaction in the blockchain.

Let $\:{D}_{i}$​ represent the data block for the iii-th transaction, containing temperature data and heating control decisions. A blockchain consensus algorithm (such as Proof of Stake) is used to validate each transaction. The validation condition is:

$$\:{A}_{i}\left(t\right)\ge\:\frac{\left|V\right(t\left)\right|}{2}$$

(41)

where $\:{A}_{i}\left(t\right)$ is the number of validators approving the block, and $\:\left|V\right(t\left)\right|$ is the total number of validators in the network. Once the consensus is reached, the block is added to the blockchain, ensuring tamper-proof and auditable data.

Step 4

Energy Consumption Optimization

The final step is to optimize energy consumption while maintaining room comfort. The cost function for energy consumption over a period $\:T$ is:

$$\:{C}_{E}={\int\:}_{0}^{T}\:P\left(t\right)\cdot\:{P}_{\text{price\:}}\left(t\right)dt$$

(42)

where $\:{P}_{\text{price\:}}\left(t\right)$ is the dynamic energy price at time $\:t$. The optimization problem becomes:

$$\:\text{m}\text{i}\text{n}\text{i}\text{m}\text{i}\text{z}\text{e}{C}_{E}={\int\:}_{\downarrow\:}^{T}\:P\left(t\right)\cdot\:{P}_{\text{price\:}}\left(t\right)dt$$

(43)

subject to:

$$\:{T}_{\text{m}\text{i}\text{n}}\le\:T\left(t\right)\le\:{T}_{\text{m}\text{a}\text{x}},\:P\left(t\right)\le\:{P}_{\text{m}\text{a}\text{x}}$$

(44)

where $\:{P}_{\text{m}\text{a}\text{x}}$ is the maximum power that the heating system can provide.

The AI-powered system adjusts the heating power $\:P\left(t\right)$ to minimize energy costs by scheduling heating during periods of lower energy prices and leveraging the predictive model to avoid unnecessary heating during peak times.

Step 5

Performance Metrics and Evaluation

To evaluate the performance of the system, key metrics such as energy savings, temperature control accuracy, and system scalability are computed. For example, energy consumption can be reduced by 15.8%, as mentioned in the abstract, by optimizing heating schedules and predictive controls. Additionally, the system can reduce computational load by 22% through time-shifted data processing.

The simulations were conducted using MATLAB for predictive temperature control and blockchain implementation. A smart home model was developed, incorporating realistic thermal dynamics, energy consumption profiles, and user preferences. The key parameters for the simulation are as Table 4.

Table 4 Simulation setting.

These simulation settings highlight the robustness and practical feasibility of the proposed framework, providing a clear basis for evaluating its performance.

Figure 3 illustrates the system’s ability to maintain room temperature under varying external conditions, demonstrating the novelty of our predictive AI-driven approach. The machine learning component anticipates temperature fluctuations and adjusts heating in real-time, enabling precise temperature control even as outside temperatures drop from 10 °C to 0 °C. The smooth curves indicate the system’s rapid response to temperature deviations, showcasing the model’s capacity for fine-grained temperature regulation, a significant improvement over traditional thermostat controls that react only after deviations occur.

Fig. 3

Room Temperature vs. Time for Different Outside Temperatures.

Figure 4 emphasizes the energy-efficient nature of the proposed system by displaying the required heating power over time for each outside temperature. The novelty lies in the dynamic power management based on predictive models and time-shifted analysis, which reduces unnecessary energy spikes. Unlike conventional systems that apply heating continuously to counter temperature drops, our system optimally adjusts heating power, minimizing energy consumption while maintaining comfort. The gradual power adjustments across scenarios reveal the system’s ability to adapt its energy output intelligently based on anticipated needs.

Fig. 4

Heating Power vs. Time for Different Outside Temperatures.

The comparison of room and outside temperatures highlights the system’s capability to maintain indoor comfort despite significant external fluctuations in Fig. 5. The novelty of this framework is in the integration of WSNs with AI and blockchain, which enables real-time environmental monitoring and secure data handling. The system responds predictively to outside temperature changes, reducing heating power when external temperatures are higher, and increasing it when external temperatures drop, thereby optimizing energy efficiency. This dynamic response mechanism showcases how the system outperforms traditional static thermostat controls.

Fig. 5

Room and Outside Temperature Comparison.

The bar chart in Fig. 6 showcases the overall energy efficiency of the system, with a clear trend showing increased consumption as outside temperatures decrease. The novelty of the approach is evident in how it minimizes energy consumption through predictive analysis and intelligent scheduling. Even though more heating power is required in colder conditions, the energy usage is optimized due to the system’s ability to foresee heating demands and avoid overcompensation. The blockchain component ensures secure and transparent monitoring of energy usage, further enhancing the system’s efficiency by allowing decentralized energy management in a smart home environment.

Fig. 6

Energy Consumption vs. Outside Temperature.

Figure 7 depicts the living room radiator temperature over time, showcasing how the temperature dynamically adjusts as external conditions change. This figure highlights the impact of predictive temperature control, comparing power consumption with and without time-shifted analysis at varying outside temperatures. The proposed AI-powered system effectively manages the radiator’s heat output based on predictive algorithms, ensuring optimal comfort while minimizing energy consumption. By anticipating temperature variations and adjusting the heating schedule accordingly, the system reduces energy use during periods of low demand. This is especially evident in the temperature curves, where the power consumption with time-shifted analysis demonstrates smoother transitions and fewer peaks.

The method’s superiority stems from its ability to balance energy efficiency and comfort. Traditional non-predictive systems react only after significant temperature changes occur, leading to more energy being consumed to restore the desired room temperature. In contrast, the predictive method preemptively adjusts the heating output, resulting in a more stable and efficient control. Additionally, the integration of blockchain technology enhances system security without compromising performance. By securing sensor data in a decentralized manner, the system eliminates vulnerabilities that may exist in conventional smart home setups, ensuring trust and transparency in data handling.

Ultimately, this combined approach of predictive AI, time-shifted analysis, and blockchain demonstrates significant improvements over traditional methods, with lower energy consumption, higher system security, and a smoother user experience.

Fig. 7

Living room radiator temperature over time.

Figure 8 illustrates the heat-on event detection process, where the system accurately identifies when the radiator begins to heat the room. This event detection is key to optimizing energy use, as it allows the system to adjust heating schedules based on real-time and predicted temperature needs. The proposed method’s superiority lies in its combination of machine learning and time-shifted analysis. By leveraging AI algorithms, the system not only detects heat-on events but also predicts future heating needs, minimizing unnecessary energy use. Unlike traditional systems that activate heating based purely on current temperature, this approach proactively schedules heat-on events in advance. Additionally, the integration of blockchain ensures that the heat-on event data is securely logged, preventing unauthorized tampering and adding an extra layer of trust to the system. This hybrid approach, combining event detection, predictive scheduling, and secure data management, results in a more efficient and reliable temperature control system that outperforms conventional smart home solutions.

Fig. 8

Heat-on event detection.

Figure 9 displays the detection of scheduled heat-on events, where the system identifies and activates heating based on predefined schedules. This is achieved through predictive algorithms that optimize heating times according to historical data and anticipated temperature changes. The superiority of the proposed method lies in its ability to precisely control heating schedules while adapting to real-time conditions. Unlike conventional systems that rely on fixed schedules, the AI-powered framework dynamically adjusts heat-on timings to better match energy demand, reducing wastage. The system also leverages time-shifted analysis to further improve efficiency by minimizing peak energy loads. Moreover, blockchain integration ensures that all scheduled events are securely recorded, preventing unauthorized alterations and enhancing transparency. This combination of predictive control, time-shifted analysis, and secure scheduling significantly outperforms traditional thermostat systems, offering enhanced energy efficiency, reliability, and trust.

Fig. 9

Detection of scheduled heat-on events.

Figure 10 illustrates the time delay before the radiator starts to warm up. This figure highlights the system’s predictive capability to preemptively activate heating based on forecasted temperature needs. The proposed method excels by minimizing the time between detecting the need for heating and initiating the warming process. By forecasting temperature changes and adjusting the radiator’s operation in advance, the system ensures a more consistent and efficient indoor climate. This proactive approach contrasts with traditional systems that often react too late, leading to less optimal energy use. The combination of advanced predictive algorithms and secure, real-time adjustments leads to faster responses and reduced energy consumption, showcasing the system’s superior performance and efficiency.

Fig. 10

Time delay before radiator activation, showing efficient preemptive heating.

Figure 11 illustrates the detection of radiator hot events, where the system identifies when the radiator reaches high temperatures. This feature is crucial for maintaining optimal heating performance and preventing overheating. The proposed method stands out by providing accurate and timely detection of hot events, ensuring that heating levels are adjusted promptly to avoid excessive temperatures. This proactive management prevents energy wastage and enhances safety. By integrating predictive algorithms and real-time monitoring, the system not only optimizes energy efficiency but also improves overall reliability compared to traditional heating controls that may lack such precision.

Fig. 11

Radiator hot event detection, ensuring optimal heating and preventing overheating.

Figure 12 depicts radiator cooling event detection, illustrating how the system identifies when the radiator’s temperature decreases. The method’s advantage lies in its precise detection and timely response to cooling events, allowing for effective management of temperature transitions. By accurately monitoring and adjusting the radiator’s cooling process, the system enhances energy efficiency and maintains a stable indoor climate. This proactive capability surpasses traditional systems, which may react sluggishly to cooling needs, ensuring better comfort and reduced energy consumption.

Fig. 12

Radiator cooling event detection.

Figure 13 displays the temperatures of the living room and radiator, highlighting how the system maintains optimal comfort. This figure demonstrates the system’s ability to synchronize radiator output with living room temperature, ensuring consistent and efficient heating. The advanced predictive control not only keeps the room temperature stable but also reduces energy usage by adjusting radiator settings proactively. This approach provides superior performance compared to traditional systems, which may result in fluctuating temperatures and higher energy consumption.

Fig. 13

Living room and radiator temperatures, showcasing efficient and consistent temperature control.

Figure 14 shows room temperatures during a single heating event, illustrating the system’s precise control.This figure highlights how the proposed method maintains stable room temperatures throughout the heating process. The system’s predictive algorithms ensure that temperature changes are smooth and well-managed, avoiding abrupt fluctuations. This capability improves comfort and energy efficiency compared to traditional systems, which may struggle to regulate temperature consistently during heating events.

Fig. 14

Room temperatures during one heating event.

Table 5 is a table comparing the proposed methods in the article against traditional temperature control methods. This table highlights key features and performance metrics to illustrate the advantages of the proposed approach.

Table 5 Comparison of proposed method and traditional temperature control systems.

The proposed method, leveraging predictive AI, time-shifted analysis, and blockchain, demonstrates significant improvements over traditional temperature control systems. It offers enhanced energy efficiency, better temperature stability, faster response times, and improved data security. These features collectively contribute to a more effective and reliable smart home temperature management solution.

Table 6 is a numerical table comparing the performance of the proposed method with traditional temperature control methods. The table includes various metrics to showcase the improvements offered by the proposed approach.

Table 6 Numerical comparison of the proposed method versus traditional temperature control systems.

This numerical comparison illustrates the proposed method’s superior performance across several key metrics. The method achieves significant reductions in energy consumption, improved accuracy in event detection, and enhanced reliability in scheduled heating. It also provides better temperature stability and faster response times compared to traditional methods.

The simulation results in Fig. 15 clearly demonstrate the superior performance of the proposed AI-Powered Blockchain Framework for Predictive Temperature Control compared to traditional thermostat and PID control methods. The first graph shows that the proposed method achieves more precise temperature regulation, closely tracking the desired temperature with minimal fluctuations. In contrast, the thermostat and PID methods exhibit higher deviations. The second graph highlights the energy efficiency of the proposed framework, consuming significantly less energy due to its predictive and adaptive capabilities. This reduction in energy consumption, combined with improved temperature control accuracy, underscores the effectiveness of the proposed system in optimizing energy usage while maintaining comfort, making it an ideal solution for modern smart homes.

Fig. 15

Comparison of temperature regulation and energy efficiency among control methods.

Algorithmic complexity analysis of proposed approach

To analyze the algorithm complexity for the proposed AI-powered blockchain framework for predictive temperature control in smart homes, we need to evaluate both the time complexity and space complexity of the key components involved. Here is a general approach to analyzing algorithm complexity for the proposed framework:

• I. Data Collection and Preprocessing.

• Time Complexity: Collecting data from WSNs involves reading sensor data periodically, which is typically O(n) for n sensors.

• Space Complexity: Storing sensor data requires O(n) space, as data from each sensor must be stored temporarily for processing.

• II. Machine Learning Model for Predictive Temperature Control.

• Training the Model: The time complexity for training a machine learning model (e.g., decision trees, SVMs, or neural networks) is usually dependent on the number of data points (m) and features (d). For a model like neural networks, the complexity can be O(m.d.k), where k is the number of epochs. For tree-based models, the complexity is generally O(m.log(m)).

• Inference Complexity: The time complexity for real-time predictions is generally O(d), where d is the number of features used in the model. This is the time taken to make predictions after the model has been trained.

• III. Blockchain Integration for Secure Data Logging.

• Transaction Processing: Each data transaction to the blockchain (e.g., recording sensor readings or control actions) can be processed in O(1) time for each transaction. However, as the blockchain grows in size, block verification and consensus can introduce additional computational complexity. The time complexity for consensus protocols (e.g., Proof-of-Work, Proof-of-Stake) is typically O(log(n)) where n is the number of nodes in the network.

• Space Complexity: The space required for blockchain storage is proportional to the number of blocks and transactions. For a decentralized system, this grows with the number of blocks, making the space complexity O(b) where b is the number of blocks in the chain.

• IV. Real-time Event Detection and Adjustments.

• Event Detection: Detecting heat-on or cooling events involves analyzing the sensor data in real-time. The complexity of event detection is typically O(n), where n is the number of sensors. If using a more sophisticated model for anomaly detection, the complexity could increase to O(n.d).

• Adjustment Calculation: The time required to adjust temperature settings or initiate control actions is generally O(1) for each event.

• V. Energy Consumption and Network Optimization.

• Energy Consumption Estimation: Calculating energy consumption reduction involves estimating the energy saved through predictive adjustments. This is typically O(1) for each prediction, as it requires only the current state of the system.

• Network Lifetime Analysis: The complexity of analyzing network lifetime depends on the number of edge devices (n) and their energy consumption rates. For systems with multiple devices, the complexity could be O(n) or higher, depending on the communication and processing requirements.

Overall Algorithm Complexity.

• Time Complexity: The overall time complexity of the system can be considered a combination of the complexities of the individual components. The overall time complexity is influenced by the sensor data collection (O(n)), real-time machine learning inference (O(d)), blockchain transaction processing (O(1) or O(log(n))), and event detection (O(n)).

• Space Complexity: The space complexity primarily depends on the data storage for sensor readings (O(n)) and the size of the blockchain (O(b)), resulting in an overall space complexity of O(n + b).

This analysis gives an overview of the computational complexity of the major components involved in the AI-powered blockchain framework. The framework is designed to be efficient, with optimizations for both real-time control and secure data logging.

Results and discussion

This paper presents a novel AI-powered blockchain framework for predictive temperature control in smart homes, integrating wireless sensor networks and time-shifted analysis. The proposed system demonstrates significant advancements in energy efficiency, temperature stability, and data security compared to traditional methods. The key innovations of the framework include dynamic detection of heating and cooling events, predictive scheduling, and secure data handling through blockchain technology. Performance evaluations reveal that the system reduces energy consumption, achieves higher accuracy in event detection, and provides reliable temperature control with minimal fluctuations. The time-shifted analysis further enhances efficiency by reducing peak-time computational loads. Overall, the proposed method offers a robust and efficient solution for smart home temperature management, addressing the limitations of conventional systems. By combining advanced AI algorithms, secure blockchain integration, and real-time data processing, this approach sets a new standard in optimizing energy use and improving user comfort in smart homes.

While the proposed AI-powered blockchain framework demonstrates significant potential for predictive temperature control in smart homes, there are several avenues for future research and improvements. One important direction is the enhancement of the framework’s scalability, particularly when dealing with larger networks of IoT devices and edge nodes. The system’s ability to efficiently handle a growing number of sensors and devices, while maintaining real-time performance and data integrity, will be crucial for large-scale deployments.

Additionally, future work could explore the integration of more advanced machine learning models, such as deep reinforcement learning, to further optimize predictive control and improve energy efficiency. This would allow the system to continuously learn from real-time data and adapt to environmental changes more effectively.

Another area for further investigation is the incorporation of more robust cybersecurity measures. Although blockchain technology provides a secure data logging mechanism, potential vulnerabilities may still exist in other components of the system. Future research could focus on enhancing the security protocols, such as incorporating secure multi-party computation or homomorphic encryption to protect sensitive data throughout the network.

Lastly, real-world testing and validation of the framework in diverse smart home environments are essential. While the current results are promising, future work should include experimental validation in various settings to ensure the practical applicability and robustness of the system. This would also involve assessing the system’s performance under different energy consumption scenarios, network conditions, and user behaviors.

By addressing these areas, the framework could be refined and better adapted for widespread adoption in smart home systems, contributing to more energy-efficient and secure home automation solutions.