Non-intrusive load monitoring for industrial production line based on GMM-FHMM

Non-intrusive load monitoring plays a significant role in supporting applications such as load forecasting and demand response. To address the challenge of independently decomposing sub-equipment in industrial load users with production lines， a non-intrusive load monitoring （NILM） scheme is proposed， using the production line as the decomposition unit， based on the interlinked operation of equipment within the line. A factorized hidden Markov model （FHMM）， based on Gaussian mixture model （GMM）， is employed to achieve a fine-grained representation of load at the production line level. Additionally， a time-segmented state transition probability decomposition model is developed， leveraging the stable overall load patterns of industrial production lines， to further enhance load monitoring accuracy. Experimental results demonstrate that the proposed model significantly improves performance in both multi-state modeling and time segmentation， with load monitoring error metrics on some production lines ultimately reduced by nearly 20%.