Non-Intrusive Load Monitoring Techniques for Intelligent Energy Management: A Comparative Study of FHMM and LSTM Approach

This study addresses the problem of detecting and segregating electrical loads using two alternative data mining algorithms for Non-Intrusive Load Monitoring (NILM). The initial method makes use of a Factorial Hidden Markov Model (FHMM), which uses hidden state observations to forecast each device's usage. The second method uses a Long Short-Term Memory (LSTM) network and time-windowed data to measure each device's consumption over a given period. Experimental results, measured using the Root Mean Square Error (RMSE) metric, reveal that both techniques successfully address the disaggregation challenge, with low error rates and exact forecasting of individual energy use. The study emphasises the importance of input sequences obtained from mixed signals on model predictions and comparison results. Furthermore, using publically available tools and datasets improves data pretreatment efficiency and facilitates the use of NILM approaches. This study improves NILM methodologies by providing insights into effective modelling strategies and emphasising the importance of readily available resources in the energy disaggregation field. Its purpose is to encourage more research and improvement in this field.