Equivalent Modelling Method for Renewable Power Plant Under the Effect of Multi-scale Factors Based on Machine Learning

In this paper, a multi-generator modelling method for solar and wind power plants under the effect of multi-scale factors is investigated based on machine learning. The key influential factors of the model are selected and classified, of which the historical data of the last two years is captured to build a database. A combined algorithm of K-means++ and Iterative Self-organizing Data Analysis Technique is established to achieve the optimal clustering of the data, so as to separate wind turbines/ photovoltaic cells into subgroups. Thereafter, separated models of generator are built according to the clustering result, which in turn form the equivalent multi-generator model of the entire power plant. The modelling method is validated through case studies of a typical wind farm and a photovoltaic power plant located in the northeast of China, and the result proves that the equivalent model can improve simulation speed effectively while keeping the overall error below 5% in contrast to the detailed model.