Performance Analysis of Nine Disaggregation Algorithms on the DSUALM10 Dataset: A Notebook Study

Non-Intrusive Load Monitoring (NILM) facilitates the disaggregation of appliance-level energy consumption from aggregate electrical signals, offering a scalable method to optimize energy efficiency. This notebook compares the performance of traditional NILM algorithms (Mean, CO, Hart85, Exact FHMM) with deep neural network-based approaches (DAE, RNN, Seq2Point, Seq2Seq, WindowGRU) under various experimental conditions. Factors such as sampling rate, harmonic content, and the application of power filters were analyzed. A key aspect of the evaluation lay in the differing test conditions: while traditional algorithms were evaluated under multiple experimental configurations, deep learning models, due to their high computational cost, were analyzed exclusively under a specific configuration (1-second sampling rate, with harmonic content present, and without the application of power filters).

The results confirm that no universally superior algorithm exists; performance varies depending on the appliance type and signal characteristics. Traditional algorithms are faster and more computationally efficient, making them more suitable for scenarios with limited resources or those requiring rapid response. However, deep learning models, though significantly more computationally expensive, demonstrated higher average accuracy (MAE, RMSE, NDE) and event detection capability (F1-score) in the particular configuration under which they were evaluated. Under these specific conditions, these models excel at detailed signal reconstruction and handling harmonics without the need for filtering. The selection of an optimal NILM algorithm for real-world applications must balance desired accuracy, load types, electrical signal characteristics, and, crucially, the limitations of available computational resources.