Characteristic Portrait Construction of Voltage in Substation Areas Using ISSA-Based Cluster Prediction

To effectively utilize power data for monitoring and issuing warnings in distribution-substation areas, a clustering prediction model was developed by integrating an improved salp swarm optimization algorithm. First, power characteristics related to the target were collected using original data mining and dimensionality reduction techniques. A voltage-characteristic label model for the substation areas was then developed, focusing on three dimensions: safety (S), stability (S), and economy (E). Next, because the improved salp swarm algorithm (ISSA) has better optimization effect and less iteration time, the fusion ISSA was combined with the K-means clustering algorithm to analyse the data, while the ISSA-BP neural network algorithm was used for time-series prediction. This approach generated a voltage profile and detected abnormal conditions in the substation areas. Finally, the model was validated using real data from a substation area in Shanghai. The results demonstrated that the proposed algorithm was more efficient, with improved prediction accuracy and performance metrics.