Nonintrusive electrical load monitoring

We are developing a sophisticated solution for Nonintrusive load monitoring (NILM) and classifying electrical loads in building automation systems.

The problem

Our client strives to expand his portfolio of services, using opportunities to energy disaggregation. To achieve this, it needs a reliable process for collecting, processing and analysing load information.

The solution

This is a process for analysing changes in the voltage and current going into a house and deducing what appliances are used in the house as well as their individual energy consumption. Powered by Machine Learning, the system collects data from controllers to identify and categorize various load types, such as lighting, HVAC systems, and appliances.

The platform performs detailed analysis of energy consumption, generating actionable insights and interactive reports. These tools allow users to optimize energy usage, reduce operational costs, and enhance sustainability efforts in building management. The opportunities for partnerships with utility companies to enable integration with existing grid management systems and demand-response mechanisms can guarantee our client position of a leader in energy management and IoT services.

Our team was responsible for

Identifying and gathering requirements
UI/UX
Implementing the web user interface
Integration with Machine Learning models

The client

Оne of the leading and fast-growing IoT companies in Bulgaria, known for its innovative smart home and automation products. Long-time partner of Infinno.

Frontend technologies

web UI – Angular

How we used ML

We use tools like NumPy and Pandas to carefully preprocess and engineer features from the raw data, ensuring we extract the most relevant information for training. This allows us to better understand the patterns and relationships within the data before feeding it into our model. To visualize the results and monitor the model’s performance, we use Matplotlib, helping us communicate how well our models are working. For the classification task, we apply deep learning techniques with TensorFlow and Keras, using advanced models like WaveNet, which excels at handling time-series data, to accurately predict appliance usage. Through this approach, we can build robust models capable of making precise predictions for energy consumption.