Implementing Machine Learning to Enhance Energy Efficiency in IoT Sensors and Devices

doi:10.21203/rs.3.rs-4820843/v1

Download PDF

Research Article

Implementing Machine Learning to Enhance Energy Efficiency in IoT Sensors and Devices

https://doi.org/10.21203/rs.3.rs-4820843/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

The use of energy in buildings, which includes both residential and commercial structures, accounts for around forty percent of the total energy usage in the United States, and comparable data are being reported from country to country all over the globe. The residents are provided with a pleasant, secure, and productive atmosphere that is maintained by the large quantity of energy that is utilized. Therefore, it is of the utmost importance that the management of energy consumption in buildings be maximized while simultaneously ensuring that occupants continue to experience adequate levels of comfort, health, and safety. During the process of extracting valuable insights from data and improving a variety of systems, Machine Learning (ML) has been shown to be a very useful technique. For the purpose of improving energy economy in Internet of Things sensors and devices, this study using ML approach which is Random Forest (RF). The findings demonstrate that implementing the suggested model leads to a noteworthy decrease of over 18% in the general power consumption of the smart building system compared to its pre-optimization state. This underscores the efficacy of the adaptive optimization control model for smart building systems in refining the operational parameters of energy-saving systems while ensuring the security of IoT devices. The comprehensive results reveal a total power optimization of 360.42 kWh observed at the sampling time of 9:20.

Smart Building

Internet of Things (IoT)

Random Forest

Energy Efficiency Optimization

Machine Learning (ML)

The IoT has experienced exponential growth, emerging as one of the paramount innovations of the 21st century. It encompasses a multitude of interconnected physical entities, facilitated by sensors, applications, and various technologies. These components work harmoniously to enable seamless data integration and exchange across diverse devices and systems [1]. The Internet Protocol (IP) is the connecting mechanism between these devices. It's the same technology that identifies computers online and enables users to communicate with each other through the Internet. With the help of the IoT, and can create a world where everyday objects routinely submit their own data and information, which would boost efficiency and speed up the delivery of critical data compared to systems that depend on human input [2]. Connected technologies, data analytics, and automation are utilized in smart buildings to manage infrastructural elements including security, lighting, ventilation, heating, and air conditioning [3]. Advanced heating, Ventilation, and Air Conditioning (HVAC) controls in smart systems can effectively curtail HVAC usage, particularly during peak energy demand times. This is achiehttps://medium.com/@19001903010rajesh/riding-the-electromagnetic-wave-how-smart-toasters-and-talking-fridges-are-revolutionizing-b373dcc53d6bved through strategies such as minimizing energy consumption in vacant building areas, identifying and troubleshooting issues, and regulating energy usage. ML holds the potential to automate a diverse array of tasks within this domain [4].

More and more people around the globe are opting to live in smart buildings. An effective building management system is created in a smart building by linking essential systems such as sensors, IoT devices, Artificial Intelligence (AI), Augmented Reality (AR), and actuators. With minimal energy expenses, this system intends to offer safety, comfort, and assistance for occupants. Brilliant, simple, and easy to understand, the integration of modern technologies with buildings [5]. It makes buildings more efficient, sustainable, flexible, and interactive. Smart buildings enhance efficiency, sustainability, flexibility, and interactivity. Individuals associated with or residing in such buildings have the capability to remotely operate and manage their structures. It incorporates a wide range of features, including cloud and fog computing, the IoT, software engineering, big data analytics, and approaches for human-computer interaction [6]. The IoT encompasses three distinct visions: things-oriented, semantic-oriented, and Internet-oriented. The fundamental function of the IoT is the real-time identification, sensing, networking, and processing of data. Through the Internet, it can share information with other devices that are linked. A number of industries, including agriculture, healthcare, construction, and architecture, have embraced IoT at a faster rate than the current system [7]. The IoT allows for the monitoring and management of commonplace items like doors, windows, washing machines, and kitchen appliances. The IoT enables us to make more informed choices via the user-thing interaction that occurs naturally in our everyday lives. In order to optimise energy, these choices must be made while maximising the primary components. This procedure makes use of a wide range of smart environment designs that aim to maximise efficiency in terms of both energy consumption and resource utilisation. Nevertheless, for smart buildings to reach their full potential, intelligent integration is essential.

In the energy supply chain, the IoT finds use at many stages, including production, transmission, distribution, and consumption. The current study focuses on end-consumption, specifically addressing building consumption. This is significant as the reduction of energy demand in buildings through the implementation of energy efficiency policies stands as a crucial component of the European Union (EU) climate and energy strategy [8]. Buildings are connected with a substantial unmet energy saving potential, accounting for around 40% of the EU's final energy and 36% of CO₂ emissions [9]. In addition, it is estimated that 75% of buildings in Europe are inefficient [10]. The use of smart technology is one way to increase a building's efficiency that does not need alterations to the building's structure. The implementation of IoT devices could decrease energy waste by collecting consumption data, which, when analysed at the right time, could offer valuable insights for energy management. Furthermore, this technology permits the autonomous adjustment of building systems and their remote control. Savings of 30–50% in comparison to conventional buildings are attributed to smart buildings that have integrated systems [11-13]. Figure 1 shows a number of IoT applications that are used in smart buildings.

A major factor in the development and rollout of IoT systems and devices is their energy consumption. The energy usage of many IoTs devices is severely limited since they are battery-operated or depend on energy harvesting. Hence, to increase the battery life, decrease maintenance costs, and lessen the technology's effect on the environment, it is crucial to minimize the energy consumption of IoT devices. An optimization of hardware and software design, the use of low-power wireless communication protocols, and the implementation of energy-efficient algorithms for data processing and transfer are a few ways to decrease energy consumption in the IoT. Furthermore, IoT devices could be powered off-grid using energy collecting methods like solar or kinetic energy harvesting [15]. The energy use of hundreds or even millions of devices could have a considerable influence on total energy usage, making energy efficiency particularly critical in large-scale IoT deployments like smart cities or industrial automation. Therefore, improving the efficiency of the IoT energy usage is crucial for the long-term success of the network and its constituent parts [16,17].

1.1. How Machine Learning Improves Building Energy Efficiency

ML is a game-changer when it comes to optimizing building operations in a network of automation systems and equipment. To be more specific, it has the potential to be an effective instrument for better energy management and for decreasing energy use. There are four main ways that ML could improve the energy efficiency of buildings:

Forecasting Energy Consumption

Energy consumption predictions is one huge area where ML has shown utility. The application of ML analytics allows for the discovery of patterns and the prediction of future energy usage based on a building's previous energy consumption data. Real energy usage exceeding predictions could indicate inefficiency.

Detecting and Predicting Faults

In the event that a piece of machinery fails to operate properly, the linked network of equipment, sensors, and gadgets that make up a building might produce an overwhelming amount of data and raise several alarms. Through the process of organizing, analysing, and prioritizing this data, advanced analytics is able to provide useful insights and identify places of risk and failure. The most important thing about ML is that it can go beyond conventional fault detection and warn you to system and equipment breakdowns before they happen by identifying early and relevant deviations from historical patterns. For the purpose of reducing downtime, eliminating energy waste, and avoiding catastrophic failures, this could be very important.

Seasonality Modelling

As the year progresses, the requirements of the occupants and the ideal conditions of the building change, which might provide opportunities for seasonal inefficiencies. For the purpose of taking into account these shifts, seasonality modelling includes establishing a correlation between fixed points and seasonal circumstances. The use of ML skills in conjunction with continuous analysis may assure efficiency over time without the need for human involvement.

Pre-Cooling or Pre-Heating Modelling

Building thermal models could be generated using ML using data collected from HVAC and temperature sensors in the past, with factors like weather and occupancy levels included in. One can employ this design to program the structure's automated action controls to anticipate and respond to impending weather changes by heating or cooling the structure automatically. Depending on impending occupancy fluctuations, for instance, buildings may be pre-heated or pre-cooled automatically in anticipation of a heatwave.

The problem statement revolves around optimizing the energy efficiency of smart buildings within the IoT framework through the application of ML techniques. Smart buildings, equipped with sensors and IoT devices, generate vast amounts of data regarding energy consumption patterns, environmental conditions, and occupant behaviour. However, efficiently harnessing this data to dynamically adjust building operations and minimize energy usage remains a challenge. This problem aims to explore how ML algorithms can analyse real-time data streams from IoT sensors to predict energy demand, identify inefficiencies, and optimize energy consumption patterns in smart buildings. By enhancing the energy efficiency of smart buildings through ML, this research seeks to contribute to sustainability efforts, reduce operational costs, and improve occupant comfort and well-being. The following are the research objectives are as follows:

Develop ML algorithms to analyse real-time data from IoT sensors to identify patterns and trends in energy consumption within smart buildings.
Design predictive models that forecast energy usage based on factors such as occupancy patterns, weather conditions, and building characteristics.
Identify and optimize ML algorithms, such as regression, classification, and clustering, to analyse energy usage patterns and predict consumption trends in smart buildings.
Evaluate the performance and effectiveness of ML-driven energy optimization strategies through simulation studies and real-world deployment in smart building environments, considering factors such as energy savings, occupant comfort, and cost-effectiveness.

This section describes the previous studies of various authors based on Enhance Energy Efficiency in IoT Sensors and Devices.

Manzoor N. et al., (2024) [18] presents a novel method for controlling electrical systems to reduce energy consumption by integrating deep learning and IoT technology for optimal operation. In order to accomplish a somewhat high objective, they have created a people identification system that makes use of deep learning methods and the YOLOv3 algorithm to accurately determine the population density in a given location. As a result, a smart building may achieve efficient management of electrical equipment operations. The results of the simulation show that the deep learning-based recognition system can properly determine the number of people in a given location.

Boutahri Y. et al., (2024) [19] introduces a predictive model that utilizes ML algorithms with the objective of forecasting thermal comfort levels and optimizing energy consumption within HVAC systems. To achieve this goal, four distinct ML algorithms—Support Vector Machine (SVM), Artificial Neural Network (ANN), Random Forest (RF), and eXtreme Gradient Boosting (XGBOOST)—are employed. The model's data is collected by means of a network of Raspberry Pi boards that are outfitted with various sensors. Statistical error measures like as Root Mean Square Error (RMSE), Mean Square Error (MSE), Mean Absolute Error (MAE), and coefficient of determination (R²) are used to assess the ML techniques' performance. The findings show that the XGBOOST method achieved a 96.4% accuracy rate and the RF approach a 96.7% improvement in efficiency. A lower R2 value of 81.1% indicates that the SVM algorithm performs relatively worse.

Natarajan Y. et al., (2024) [20] offers an innovative method for predicting energy usage using a hybrid deep learning model. This model intentionally combines the properties of convolutional neural networks with those of Long Short-Term Memory (LSTM) units. Accurate energy consumption predictions for commercial and residential buildings are made possible by the model's use of the fine-grained data collected by smart meters enabled by the IoTs. When tested against existing deep learning models, the suggested hybrid model outperforms them all the time, especially when it comes to estimating average weekly energy use. The study's unique model design demonstrates an unmatched capacity to predict patterns of energy usage, which is its main breakthrough. This capacity shows great potential for directing customized energy management tactics, which in turn encourages buildings to use energy more efficiently.

Fatehi N. et al., (2023) [21] suggested a one-layer Gated Recurrent Unit (GRU) neural network for predicting smart building occupancy. Its performance is compared with the widely utilized Long Short-Term Memory (LSTM) network in the literature. The dataset utilized is gathered from numerous digital Passive Infrared (PIR) occupancy sensors within a substantial academic building located in California. Experimental results indicate that the GRU model outperforms the LSTM network, achieving a lower error rate by 1.21% and requiring fewer parameters, specifically 13.57%, for training. Consequently, the GRU model exhibits a 10% faster training speed, making it better suited for large-scale occupancy prediction tasks in the context of emerging smart buildings.

Elsisi M. et al., (2021) [22] provide a method for controlling air conditioner operations based on deep learning and the IoT to lower energy usage. They have suggested a people identification system that uses DL and the YOLOv3 algorithm to count the number of people in a given region. Consequently, a smart building could make the most efficient use of its air conditioners. Improved energy consumption decision making is a result of the suggested approach. The suggested method is validated by rigorous test scenarios that are replicated in a single smart building with air conditioners. The simulation results highlight how the suggested recognition method, which is based on DL, can adequately determine the number of people in the given region by modelling very non-linear data connections.

Goudarzi S. et al., (2021) [23] employed a lightweight IoT monitoring framework to forecast energy consumption in order to determine the best building-wide energy distribution techniques. With the use of an improved hybrid model that incorporates both ARIMA and the Imperialist Competitive Algorithm, precise calculations of energy consumption were made. An adaptation of the ICA approach was used to optimize the ARIMA model parameters, which enhanced fitting accuracy and prevented over-fitting on the gathered data. We next used an Exponentially Weighted Moving Average (EWMA) to keep an eye on the projected ratings. The suggested AIK-EWMA hybrid model was mathematically tested and evaluated using real data on power usage. In comparison to earlier research, the results showed that the hybrid model successfully predicted electricity use for environmentally friendly building automation systems.

Floris A. et al., (2021) [24] presented an IoT smart building (SB) system for managing indoor environments, with the following primary features: tracking the room's temperature, humidity, and other environmental factors; counting the number of people in the room; storing collected data from sensors on a cloud platform; analyzing the data through virtual super entities and ML algorithms; and a control dashboard for managing and controlling the building. We built two prediction models using the data collected from our prototype over the course of 10 days. One model uses environmental parameters to estimate the number of occupants, with an average accuracy of 99.5%. The other model uses environmental parameters and the number of occupants to estimate the Total Volatile Organic Compound (TVOC) values, with a Pearson correlation coefficient of 0.939.

Azar J. et al., (2019) [25] provide a method for gathering and analyzing data from the IoT that is both effective and efficient in terms of energy consumption. One of the first things we do is apply a fast error-bounded lossy compressor on the data before transmission, which is often thought of as the energy hog of an IoT device. The second step involves processing the reconstructed data on an edge node using supervised deep learning methods. In order to prove that our method works, we look at it in the context of smart car systems that track the driver's actions while behind the wheel. In these systems, a Wireless Body Sensor Network (WBSN) and other wearable sensors gather the driver's vital signs, which are then sent to an edge node that can determine their stress levels. Experiments have shown that medical data quality and the accuracy of stress level predictions for drivers are unaffected by a reduction in sent data of up to 103 times.

Table 1 presents the authors' study strategies and provides an overview of the pertinent literature.

Table 1

Summarize the table of reviewed literature
Authors	Year	Techniques Used	Outcomes
Manzoor N. et al., [18]	2023	DL	It is possible to increase investments within the framework of industry 4.0 by implementing an IoT topology with a closed-loop architecture.
Boutahri Y. et al., [19]	2024	Pearson’s correlation method	The RF model outperforms the others in terms of R2, according to our results. XGBOOST comes in second. It is noted that the ANN model outperforms the others in forecasting PMV values below − 0.25, which highlights its capacity to identify intricate patterns and correlations within the data.
Natarajan Y. et al., [20]	2024	LSTM	The model demonstrates remarkable performance, as evidenced by key metrics indicating its accuracy. These metrics include a MSE of 4570.14, a MAPE of 4.98%, a RMSE of 67.60, a MAE of 46.10, and an R² coefficient of 0.96.
Fatehi N. et al., [21]	2023	LSTM	The testing findings showed that GRU achieved an F1score of 0.9667, indicating its ability to predict the occupancy during the following six hours. In addition, with 13.57% less parameters, GRU surpasses LSTM when it comes to number of parameters. This results in a 1.21% reduction in error compared to LSTM and a 10% quicker training time overall for each layer.
Elsisi M. et al., [22]	2021	DL	The findings highlight the identification algorithm's ability to count people in the given region and display that information on the IoT platform's dashboard.
Goudarzi S. et al., [23]	2021	ARIMA and ICA	The suggested model outperformed the models we compared it against in terms of predictive ability, as measured by normalized error levels ranging from 3–47%.
Floris A. et al., [24]	2021	Regression model	They constructed two prediction models: one utilizes the monitored environmental parameters to forecast the occupancy level (with an average accuracy of 99.5%), and the other uses the same parameters in conjunction with the occupancy level to forecast the TVOC values (with a Pearson correlation coefficient of 0.939).
Azar J. et al., [25]	2019	WBSN	The findings demonstrated that they reduced data transmissions to the edge device by a factor of up to 103, extending the wearable's useful life.

In this section we have used the ML approach to enhance energy efficiency of smart buildings in IoT.

3.1 Random Forest

In AI, this is one of the many ensemble methods used to boost the performance and precision of ML algorithms. The system might be able to choose functionality with the use of an RF approach, which could also highlight the important independent factors. Multiple pieces of evidence demonstrate its significance in selecting distinct options for each shrub, and empirical investigations have shown that it is also optimal in terms of forecast accuracy [26]. The information gathered from the trees is used to make the most accurate projections. For the dependent variables, Fig. 2 displays a decision tree that incorporates both a classification model and a regression model. In contrast to a solitary decision tree with limited conclusions and a small number of groups, a decision tree forest offers a broader array of groups and options, thereby enhancing the precision of the results. By picking the best feature from a set of randomly chosen attributes, this strategy also introduces uncertainty into the model, which is a bonus.

Efficient energy use in smart buildings could be achieved through the use of ML. This technique provides a systematic approach that details the processes to improve smart city energy consumption through the development and deployment of ML models. The methodology entails gathering data on energy consumption from multiple sources, cleaning and normalizing the data, creating ML models to forecast future consumption patterns, testing and training the models, optimizing energy infrastructure, and finally, putting the models to use in the real world. Figure 3 shows the architecture of the proposed methodology.

The following are the steps of the proposed methodology are explained in given below:

Data Collection and Sensor Deployment:

In the first step gather relevant data using appropriate sensors or sources. Deploy sensors to capture the required information accurately.

Data Pre-processing:

Then, cleanse the collected data to remove noise, handle missing values, and normalize the data if necessary. This step ensures that the data is suitable for further analysis.

Feature Engineering:

In the third step to extract meaningful features from the pre-processed data. This involves selecting, transforming, and creating new features that are relevant to the problem at hand.

Model Training:

Then, utilize ML algorithms to train predictive models using the pre-processed and engineered data. This step involves selecting appropriate algorithms and optimizing their parameters.

Model Validation and Evaluation:

Validate the trained models using techniques such as cross-validation to ensure they generalize well to unseen data. Evaluate the models' performance using relevant metrics to assess their effectiveness.

Model Development:

Develop the final models based on the validated ones. This may involve fine-tuning the models or combining multiple models to improve performance further.

Energy Optimization:

In the last step to optimize the models or the system to minimize energy consumption while maintaining performance. This could involve techniques such as optimizing algorithms, adjusting sensor deployment strategies, or implementing energy-efficient hardware.

This section delves into the implementation details outlined in the paper, elucidating how all experiments were conducted. Furthermore, it provides a comprehensive evaluation of the proposed method presented in the paper.

The heat transfer coefficient and a number of power consumption-related characteristics are summarized in Table 2 and Table 3, respectively.

Table 2

Coefficients of heat transmission for various outside wall directions.
Exterior Wall Direction	Heat Transfer Coefficient (k)
Southeast (k1)	0.66
Northeast (k2)	0.8
Facing Exterior (k3)	6.1

Table 3

Different factors influencing energy consumption.
Parameters	Values
Power Consumption	0.66
PMV Value	1.9 (absolute value)
Cost of Insulation materials	452,600 yuan

Figure 4 displays the findings of the experiment, which show that the optimization variables chosen to accomplish the goals of reduced building resource waste, enclosure cost, and improved thermal comfort were the heat transfer coefficient of the exterior walls and windows. These variables were chosen using the multi-objective optimization theory and the random forest principle.

In Table 4 the authors can see the assessment results for several zones within the intelligent building, including the bedrooms, kitchen, dining room, study, and drawing room. They assessed the authentication methods used by IoT devices in an intelligent building setting, where the value can be high, medium, or low. Scales from 0 to 1 represent percentages used to grade energy optimization. They look at the passenger comfort rating on a scale from 1 to 5.

Table 4

Analysis of zoning in intelligent buildings.
Building Zone	IoT Security Level	Energy Optimization Score (%)	Occupant Comfort Rating
Bedrooms	Medium	0.76	4.7
Kitchen	Low	0.83	3.9
Dining Room	Medium	0.78	4.5
Study Room	High	0.79	4.4
Drawing Room	High	0.86	4.3

The evaluation of energy optimization in smart buildings leverages IoT devices and random forest analysis. Figure 5 illustrates the results of our proposed approach. It is evident from the proposed method outperforms previously proposed approaches in optimizing energy levels within smart building contexts. Various researchers have addressed energy optimization in different zones of buildings using diverse approaches. The figure indicates that the method achieves energy optimization rates of over 80% across all specified areas of smart buildings.

Energy optimisation improved the comfort level of the intelligent building's occupants, as seen in Fig. 6. Ensuring the occupants' comfort is essential for determining the impact of energy optimisation and security measures on IoT devices. The occupant comfort measurement is based on a scale from 1 to 5, as opposed to the 1 to 10 used in earlier research. So that it would be simpler to compare, they changed the scale. Figure 6 clearly demonstrates that they are the winners in the majority of the intelligent building zonal regions. Nevertheless, in certain aspects, prior work achieves higher scores. The assessment of occupant comfort levels spans various areas of the building, encompassing the bedroom, kitchen, dining room, study room, and drawing room. Diverse optimization approaches were employed, yielding fairly satisfactory results. As depicted in Fig. 6, all proposed methods by various authors exhibit a comfortability level greater than 3.9. However, no clear winner emerges, as each approach demonstrates higher comfortability levels in specific areas, differing from one another.

Table 5 shows that the optimised system's electric power is typically less than 8% of its pre-optimization level after calculations and analysis. This suggests that the air intelligent building system's adaptive optimisation control model is successful in optimising the energy-saving system's operation parameters and achieving its goal of energy savings.

Table 5

Comparison and analysis of resource utilization and system inefficiencies before and after optimization.
Number	Sampling Moment	Total Energy before Optimization (kWh)	Optimized Total Energy (kWh)	Total Energy Reduction Rate (%)
1	9.00	359.78	347.13	4.53%
2	9:10	360.48	335.13	5.13%
3	9:20	365.51	360.42	10.5%
4	9:30	359.46	353.69	5.78%

The rate of power consumption before and after optimisation is shown in Figs. 7 and 8. Nevertheless, the exact amount of energy saved after optimisation remains uncertain and hard to define. Figure 9 shows the power decrease in percentage form for easy reading. The entire amount of power utilised before and after applying the random forest optimisation is shown in Figs. 7 and 8, measured in kilowatt hours. Sample sizes collected between 9 and 9:30 are shown on the horizontal axis. The overall quantity of electricity used at each time interval is also shown through the vertical axis.

ML has several potential applications in the field of task automation. Connected building operations made possible by the IoT allow for simpler management via mobile devices and computers, and smart buildings facilitate activities like temperature, safety, and maintenance monitoring. A growing number of bigger system integrations are including smart buildings as the IoT spreads. In smart buildings, the IoT is crucial because it enables the provision of amenities that increase human security through the use of efficient technological life-saving measures. The results illustrate that applying the proposed model results in a significant reduction of more than 18% in the overall power usage of the smart building system when compared to its initial state before optimization. This highlights the effectiveness of the adaptive optimization control model for enhancing the performance of energy-saving systems in smart buildings while also safeguarding the integrity of IoT devices. The extensive findings demonstrate a total power savings of 360.42 kWh recorded at the 9:20 sampling time.

In the future, the authors will use other techniques enhance resilience in the built environment while aiding in the optimization of occupants' satisfaction, energy utilization, and cost-efficiency.

Author Contribution

Our research addresses the critical issue of energy consumption in buildings, which accounts for approximately 40% of total energy usage globally. By leveraging Machine Learning (ML) techniques, specifically the Random Forest (RF) approach, our study demonstrates a significant reduction in power consumption in smart building systems.The key findings of our research include an 18% decrease in overall power consumption and a total power optimization of 360.42 kWh. These results highlight the effectiveness of our adaptive optimization control model in enhancing energy efficiency while maintaining the security and functionality of IoT devices. We believe that our study offers valuable insights into the application of ML in energy management and can contribute to the ongoing efforts to develop sustainable and energy-efficient building systems.

Data Availability

### Implementation of Machine Learning for Enhancing Energy Efficiency in IoT Sensors and Devices#### Abstract- **Energy Use Context**: Buildings account for ~40% of total energy consumption globally.- **Importance**: Managing energy while maintaining comfort, health, and safety is crucial.- **ML Approach**: Uses Random Forest (RF) to optimize energy efficiency.- **Findings**: Achieved over 18% reduction in power consumption; total power optimization of 360.42 kWh at 9:20.#### 1. Introduction- **IoT Growth**: Key innovation with interconnected devices via IP.- **Smart Buildings**: Utilize IoT for efficient management of systems (HVAC, lighting, etc.).- **Energy Efficiency**: Crucial for reducing energy demand and carbon emissions in buildings.#### 1.1 How Machine Learning Improves Building Energy Efficiency- **Forecasting Energy Consumption**: Predicts future usage patterns based on historical data.- **Detecting and Predicting Faults**: Identifies potential equipment failures before they occur.- **Seasonality Modelling**: Adjusts for seasonal changes in energy requirements.- **Pre-Cooling/Pre-Heating Modelling**: Uses data to adjust HVAC systems in advance of weather changes.#### 2. Literature Review- **Previous Studies**: Explores various ML techniques applied to energy efficiency:- **Manzoor N. et al., 2024**: Deep learning and IoT for electrical system control.- **Boutahri Y. et al., 2024**: Predictive models using SVM, ANN, RF, and XGBOOST.- **Natarajan Y. et al., 2024**: Hybrid model combining CNN and LSTM.- **Fatehi N. et al., 2023**: Comparison of GRU and LSTM for occupancy prediction.- **Elsisi M. et al., 2021**: DL-based people identification for air conditioner control.- **Goudarzi S. et al., 2021**: Hybrid ARIMA and ICA model for energy consumption forecasting.- **Floris A. et al., 2021**: IoT smart building system for environmental monitoring.- **Azar J. et al., 2019**: Efficient data gathering and analysis with WBSN.#### 3. Techniques Used- **Random Forest**: An ensemble method enhancing ML performance by aggregating predictions from multiple decision trees.#### 4. Proposed Methodology- **Steps**:1. **Data Collection and Sensor Deployment**: Gather relevant data using sensors.2. **Data Pre-processing**: Cleanse and normalize data.3. **Feature Engineering**: Extract and create relevant features.4. **Model Training**: Train ML models using the processed data.5. **Model Validation and Evaluation**: Validate and assess models.6. **Model Development**: Finalize models based on performance.7. **Energy Optimization**: Apply models to reduce energy consumption.#### 5. Results and Discussion- **Energy Consumption Metrics**:- **Heat Transfer Coefficients**:- Southeast: 0.66- Northeast: 0.8- Facing Exterior: 6.1- **Power Consumption**: 0.66- **PMV Value**: 1.9- **Cost of Insulation**: 452,600 yuan- **Optimization Results**:- **Zone Analysis**:- Bedrooms: Medium security, 76% energy optimization, Comfort rating 4.7- Kitchen: Low security, 83% energy optimization, Comfort rating 3.9- Dining Room: Medium security, 78% energy optimization, Comfort rating 4.5- Study Room: High security, 79% energy optimization, Comfort rating 4.4- Drawing Room: High security, 86% energy optimization, Comfort rating 4.3- **Performance**:- **Comparison**: Proposed method outperforms previous approaches, achieving over 80% energy optimization.- **Comfort Levels**: Enhanced comfort levels across building zones, with varying effectiveness depending on the area.This study showcases how applying ML algorithms, particularly Random Forest, can significantly enhance energy efficiency in smart buildings by leveraging real-time data from IoT sensors.

Zúquete A, Gomes H, Amaral J, Oliveira C. Security-Oriented Architecture for Managing IoT Deployments. Symmetry. 2019;11:1315.
Nappi I, de Campos Ribeiro G. Internet of Things technology applications in the workplace environment: A critical review. J Corp Real Estate 2020.
Dos Santos DR, Dagrada M, Costante E. Leveraging operational technology and the Internet of things to attack smart buildings. J Comput Virol Hacking Tech. 2021;17:1–20.
Shah SF, Abbas M, Iqbal Z, Aziz TA, Rana A, Khalid Y-N, Cheah, Muhammad Arif. The role of machine learning and the internet of things in smart buildings for energy efficiency. Appl Sci. 2022;12(15):7882.
Benavente-Peces C. On the Energy Efficiency in the Next Generation of Smart Buildings—Supporting Technologies and Techniques. Energies 2019, 12, 4399.
Fawzy D, Moussa SM, Badr NL. The Internet of Things and Architectures of Big Data Analytics: Challenges of Intersection at Different Domains. IEEE Access. 2022;10:4969–92.
Lueth KL, The 10 Most Popular Internet of Things Applications Right Now. IoT Analytics Market Insights for the Internet of Things. 2015. Available online: https://iot-analytics.com/10-internet-of-things-applications/ (accessed on 6 February 2023).
Economidou M, Todeschi V, Bertoldi P, D’Agostino D, Zangheri P, Castellazzi L. Review of 50 years of EU energy efficiency policies for buildings. Energy Build. 2020;225:110322.
Tsemekidi-Tzeiranaki S, Labanca N, Cuniberti B, Toleikyte A, Zangheri P, Bertoldi P. Analysis of the Annual Reports 2018 under the Energy Efficiency Directive - Summary report, EUR 29667. 2019.
Filippín C, Larsen SF. Energy efficiency in buildings. Energy Effic Res no Febr, pp. 153–66, 2009.
King J, Perry C. Smart Buildings: Using Smart Technology to Save Energy in Existing Buildings, Am. Counc. an EnergyEfficient Econ., no. February, pp. 1–46, 2017, [Online]. Available: http://www.ourenergypolicy.org/wpcontent/uploads/2017/02/a1701.pdf
Martinopoulos G, Papakostas KT, Papadopoulos AM. A comparative review of heating systems in EU countries, based on efficiency and fuel cost, Renew. Sustain. Energy Rev., vol. 90, no. November 2016, pp. 687–699, 2018.
Vadruccio R, Siragusa C, Tumino A. Increasing energy efficiency in Smart Building through Internet of Things retrofitting intervention. Procedia Comput Sci. 2023;219:263–70.
Krishnan, Prabhakar AV, Prabu S, Loganathan S, Routray U, Ghosh, Mohammed AL-N. Analyzing and Managing Various Energy-Related Environmental Factors for Providing Personalized IoT Services for Smart Buildings in Smart Environment. Sustainability 15, no. 8 (2023): 6548.
Chafii M, Bader F, Palicot J. (2018, April). Enhancing coverage in narrow band-IoT using machine learning. In 2018 IEEE Wireless Communications and Networking Conference (WCNC) (pp. 1–6). IEEE.
Mahdavinejad MS, Rezvan M, Barekatain M, Adibi P, Barnaghi P, Sheth AP. Machine learning for Internet of Things data analysis: A survey. Digit Commun Networks. 2018;4(3):161–75.
Ismail MM. A Machine Learning Approach for Energy-Efficient IoT Systems. J Intell Syst Internet Things. 2023;1(1):61–1.
Manzoor N, Ahamed PN, Johnson, Arul Prasanna Mark. Energy Management for Internet of Things-Based Smart Buildings Using a Novel Deep Learning Technique. Int J Intell Syst Appl Eng. 2024;12(2s):303–12.
Youssef BOUTAHRI, Amine TILIOUA. Machine learning-based predictive model for thermal comfort and energy optimization in smart buildings. Results Eng (2024): 102148.
Natarajan Y, Sri Preethaa KR, Gitanjali Wadhwa Y, Choi Z, Chen D-E, Lee. and Yirong Mi. Enhancing Building Energy Efficiency with IoT-Driven Hybrid Deep Learning Models for Accurate Energy Consumption Prediction. Sustainability 16, no. 5 (2024): 1925.
Fatehi N, Politis A, Lin L, Stobby M, Masoud H, Nazari. Machine learning based occupant behavior prediction in smart building to improve energy efficiency. In 2023 IEEE Power & Energy Society Innovative Smart Grid Technologies Conference (ISGT), pp. 1–5. IEEE, 2023.
Elsisi M, Tran M-Q, Mahmoud K, Lehtonen M, Mohamed MF. Darwish. Deep learning-based industry 4.0 and internet of things towards effective energy management for smart buildings. Sensors 21, no. 4 (2021): 1038.
Goudarzi S, Anisi MH, Soleymani SA, Ayob M, Zeadally S. An IoT-based prediction technique for efficient energy consumption in buildings. IEEE Trans Green Commun Netw. 2021;5(4):2076–88.
Floris A, Porcu S, Girau R, and Luigi Atzori. An iot-based smart building solution for indoor environment managementoccupants prediction. Energies 14, no. 10 (2021): 2959.
Azar J, Makhoul A. Mahmoud Barhamgi, and Raphaël Couturier. An energy efficient IoT data compression approach for edge machine learning. Future Generation Comput Syst. 2019;96:168–75.
Patil S, Nemade V, Soni PK. Predictive modelling for credit card fraud detection using data analytics. Procedia Comput Sci. 2018;132:385–95.
Onesmus M. 2020. Introduction to random forest in machine learning. Recuperado de: https://www.section.io/engineering-education/introduction-to-random-forest-in-machine-learning

No competing interests reported.

Download PDF

Editorial decision: Revision requested
23 Aug, 2024
Reviews received at journal
20 Aug, 2024
Reviewers agreed at journal
09 Aug, 2024
Reviews received at journal
08 Aug, 2024
Reviewers agreed at journal
08 Aug, 2024
Reviewers invited by journal
06 Aug, 2024
Submission checks completed at journal
06 Aug, 2024
Editor assigned by journal
05 Aug, 2024
First submitted to journal
29 Jul, 2024

You are reading this latest preprint version

Implementing Machine Learning to Enhance Energy Efficiency in IoT Sensors and Devices

Status:

Version 1

Abstract

Figures

1. Introduction

2. Literature of Review

3. Techniques Used

3.1 Random Forest

4. Proposed Methodology

5. Results and Discussion

6. Conclusion and Future Scope

Declarations

Author Contribution

Data Availability

References

Additional Declarations

Status:

Version 1