Harmonic Prediction Based on PCA-ELM-LSTM Hybrid Model

With the advancement of new power system construction, the harmonic sources of power grid are more dispersed and coupling is stronger. The basis for effective analysis and governance of power grid harmonics is data, while the installation of special harmonic monitoring nodes is very small, so the gap of harmonic data is large, leading to difficulty in harmonic problem analysis. In order to obtain harmonic data covering the entire network, this paper constructs a hybrid prediction model based on PCA-ELM-LSTM, which is based on harmonic universal measurement and online monitoring data. This model combines historical information memory and output stability, specifically based on LSTM learning. Subsequently, the input layer and hidden layer results of LSTM are given to an ELM, which performs secondary learning. The learning results of these ELMs are then output to the final ELM, obtain load prediction results through the third learning of the final ELM; To improve the quality of input data and alleviate dimensional disasters, PCA is used for data dimensionality reduction processing; Finally, the harmonic data of a certain power grid was validated, and the results showed that the PCA-ELM-LSTM hybrid model designed in this paper had more accurate harmonic prediction results.