Non-Intrusive Load Monitoring (NILM) based Smart Energy Management System for Households
The conventional smart homes power management and monitoring are the sensor-based techniques, such as motion detector and IR sensors for person counting to perform smart actions for saving power. However, these models have to be implemented separately in all the rooms and areas of the home, that mak
2025-06-28 16:34:16 - Adil Khan
Non-Intrusive Load Monitoring (NILM) based Smart Energy Management System for Households
Project Area of Specialization Electrical/Electronic EngineeringProject SummaryThe conventional smart homes power management and monitoring are the sensor-based techniques, such as motion detector and IR sensors for person counting to perform smart actions for saving power. However, these models have to be implemented separately in all the rooms and areas of the home, that makes it expensive and harder to implement. The aim of this project is to implement sensorless load monitoring and energy management technique along with an efficient machine learning algorithm for appliance classification and identification. Therefore, we are developing a power management system based on Non-Intrusive Load Monitoring (NILM) that can provide us the benefit of controlling all the devices from a single platform and making it smart enough to perform power disaggregation from the aggregated power signal to detect and control the appliances. For this purpose, a hardware setup will be constructed using current transformers and potential transformers to record and observe aggregated power signal that will be processed by a controller, and disaggregated into individual power consumption of appliance loads. The disaggregated power signal information will be used to identify the individual appliances turned on at any particular time. This information will further be used to design a smart controlling device that can restrict any particular appliance for a certain number of time slots and have the capability to shift the shiftable load to off-peak hours. Moreover, the developed system will also be able to generate a detailed daily report regarding power losses, power factor, and harmonic generation.
Project ObjectivesThe main objectives we need to achieve in order to make this project are as follows:
- A hardware setup will be constructed to measure and record aggregated and individual power signal profiles of electric appliances for the energy management system installation.
- A pre-processing algorithm will be designed for the noise filtration of recorded power signal profiles.
- A feature extraction algorithm will be designed to exact important features from the power signal profile, such as mean, variance, standard deviation, and Min/Max values.
- A machine learning-based algorithm will be designed to identify/ classify the individual appliance in real-time by disaggregating the aggregated power signal profile based on the extracted features.
- A centralized mobile application platform will be developed to control and restrict the usage of appliances by determining their time of use. Moreover, the user can assess the record load profile statistically.
Non-Intrusive Load Monitoring (NILM) is a technique to use an aggregated power signal that comes from whole vicinity power monitoring, to make inferences about different individual load in the vicinity. The energy consumption information about individual appliances is much more useful for the customers than the aggregated power output. In this project, a system will be developed that will record the aggregated power signal from a single node and disaggregate it into individual appliance loads and identify the
appliances being plugged in the electric power lanes in real-time. If an appliance is shiftable and turned-on during peak load hours, then the user will be notified through our developed recommender system to use the appliance in off-peak hours. Moreover, our recommender system will be a mobile application that will provide leverage to the user to specify which appliances he wants to restrict and for which time slot. If the appliance restricted appliance is plugged in at the restricted time slot, the system will switch it off adaptively. The targeted objectives of the project are as follows:
- A hardware setup will be constructed to measure and record aggregated and individual power signal profiles of electric appliances for the training purpose. The measuring devices used with this hardware setup will contain current transformers, potential transformers, and micro-controllers.
- The recorded data will be transmitted via Bluetooth module to a computing node for pre-processing, such as noise filtration and harmonic removal.
- A statistical model will be developed to gather important features in the filtered recorded data, such as mean, variance, standard deviation, and Min/Max values.
- The pre-processed data and their extracted features will be used to model different supervised learning classifiers, such as Decision Tree Classifier, Hidden Markov Model, and Neural Networks, and will be tested and compared for accurate load device prediction.
- The most accurate model will be coded into the controller. So that, it can record real-time aggregated data, disaggregate it into individual appliance loads and identify the appliances being used in real-time.
- A mobile application will be developed to provide the user a centralized platform to interact with the system, by switching on/off any plugged-in appliance at any instance, to restrict the usage of specific appliances during specific time slots, and access the recorded data.
- Recommender system will be designed that will notify the user on the mobile application using a Bluetooth Module, if a shiftable appliance is being used during peak hours, and will restrict the usage of user-specified appliances during user-specified time slots.
- The recommender system will also generate a detailed daily report regarding power losses, power factor, and harmonic generation for daily/ monthly assessment.
Some of the major benefits of Non-Intrusive Load Monitoring system with Appliance Restriction are as follows:
- Small and Minimalistic Signature: All components are compact in size and require a small area for integration and installation.
- Easily Integrable: The NILM system in question can easily be integrated into an existing electricity network embedded into a household without wear and tear of walls and wiring routes since it only needs to be installed near a breaker and only requires a power connection for data analytics.
- Data Analytics And Problem Identification: Data obtained from the NILM base system is not only utilized to calculate and monitor the appliance functionality but is also used as feedback to train the system for system training hence data analytic techniques are utilized to identify problems such as when an appliance stops working or when the appliance in question shows rogue behavior.
- Single Module Approach: The NILM based approach stands out when compared with other systems in the market which include individual socket energy monitors produced by tech giants such as XIAOMI and google since our NILM bases system is non-intrusive and only requires intrusion at the circuit breaker level. Using a set of hardware components at a single point of operation nullifies the need for an individual module-based approach giving the NILM approach a significant edge.
- Self-Learning: Data obtained from the NILM system is not only displayed to the user in processed form but is also fed back into the system for training purposes constantly to make the system more efficient. More efficient readings translate into better detection and more savings consequently making the system more efficient.
- User-Friendly: The system is user-friendly as the user just has to deal with the Graphical user interface to restrict appliances, get statistical analysis of power consumption and control appliance usage based on peak hours.
- Electric energy demand is expected to double between 2010 and 2050. However, as the global population continues to grow globally and natural resources are becoming depleted, the electricity production methods being utilized today are not sustainable hence NILM based approach utilizes equipment that produces no emissions and utilizes no components that harm the environment when being utilized having a positive impact on the environment not only by conserving electricity but also by making use of eco-friendly building blocks.
- Sustainable Development: According to the 1987 report “Our Common Future” Of Brundtland Commission, Sustainable development is the development that meets the needs of today, without undermining future generations ' ability to meet their own needs. Not only does this plan aim to ensure sustainable development, but it also helps to conserve energy that is wasted due to our lack of knowledge regarding its consumption.
The hardware requirements for this project consist of potential transformers and current transformers, for data acquisition. The voltage sensor is calibrated at first and checked so that accurate readings may be ensured. For the current transformer calibration, burden resistor calculations are made and wire insulation is kept in mind to obtain accurate current measurements at all times. The output of these sensors is used to record individual data sets of everyday appliances for processing. As the heart of the system, an ARDUINO UNO board will be used mainly due to its compact size, energy efficiency and compatibility with required sensors. The board is connected with a Bluetooth transmission module (HC-05) that is used to establish communication between the software-based mobile application for this project. Using the mobile application, the user will be able to manually control all appliances and monitor statistics and view AI-based suggestions provided by the system. Application-based switch lane control is provided with the aid of a programmable relay connected to the ARDUINO board. To ensure the safety of devices connected to the relay the optoisolator functionality provided within the relay will be utilized. For the software part of this project, a significant amount of code is required and has been written, to establish communication between Bluetooth sensor and mobile application, delivering data to PC for signal filtration and feature extraction. In order to store data, MS Excel is used and a software known as PLX-DAQ will be used to guarantee communication between ARDUINO and MS Excel so that data flow and efficient communication is ensured within the elements of the system. Moreover, for the extraction of individual appliance signatures from a single signal obtained from the power meter at residential establishments, Machine Learning Classifiers/ Algorithms will be used. Classification in machine learning and statistics is a supervised learning approach in which the program learns from the data input provided to it and then uses this learning to classify new observations. We will use different classifiers such as, HMM (Hidden Markov Model), Decision Tree Classifier and Neural Networks for accuracy comparisons and will select the most accurate one for final controller coding.
Final Deliverable of the Project HW/SW integrated systemCore Industry Energy Other Industries Education Core Technology Artificial Intelligence(AI)Other Technologies Internet of Things (IoT)Sustainable Development Goals Industry, Innovation and Infrastructure, Sustainable Cities and Communities, Responsible Consumption and ProductionRequired Resources| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 58400 | |||
| Controller (Arduino Mega)) | Equipment | 2 | 1600 | 3200 |
| Controller (Raspberry Pi) | Equipment | 1 | 10000 | 10000 |
| Transformers (Current, Voltage) | Equipment | 4 | 2500 | 10000 |
| Bluetooth Module | Equipment | 2 | 500 | 1000 |
| Memory Storage for local data saving | Equipment | 1 | 1000 | 1000 |
| Programmable Relay | Equipment | 4 | 800 | 3200 |
| PCB | Equipment | 1 | 5000 | 5000 |
| Android Tablet | Equipment | 1 | 12000 | 12000 |
| Expandable Items (Wires, cables, capacitors, resistors etc.) | Equipment | 1 | 3000 | 3000 |
| Printing and Binding | Miscellaneous | 1 | 3000 | 3000 |
| Stationery | Miscellaneous | 1 | 2000 | 2000 |
| Overheads | Miscellaneous | 1 | 2000 | 2000 |
| Travelling | Miscellaneous | 1 | 3000 | 3000 |