Research on Non-Intrusive Load Feature Extraction and Load Recognition

Considering the impact of customer electricity usage patterns on load characteristics, this paper carries out a study of a new non-intrusive load monitoring. According to the correlation between users' electricity use behavior and load operation control strategy, this paper analyses the users' electricity use pattern in terms of different use periods and different behavioral habits, and proposes a method to segment the load pattern by weekdays/rest days. Cluster analysis techniques are also used to construct a database of load steady-state power characteristics that takes into account the differences in user electricity usage patterns. Finally, the load decomposition model is constructed by combining multivariate features and conditional judgement, and nonintrusive load decomposition is achieved by genetic optimization. The method proposed in this paper takes the steady state power characteristics as the main identification basis, which makes data acquisition easier and reduces the hardware cost; it also takes into account the influence of the user's power consumption pattern, which improves the effect of load decomposition and has certain practical significance.