A WT-LSTM Approach for Reactive Power Load Forecasting with Active Power Load Integration

To underscore the unique attributes of reactive power load as opposed to active power load, optimize reactive control strategies, enhance voltage quality, and minimize network losses, this paper proposes a reactive load forecasting model using Wavelet Transform-Long Short-Term Memory (WT-LSTM) neural networks, taking into account the randomness and nonlinearity of reactive load while considering active load. This method decomposes the reactive load sequence using wavelet transformation to remove randomness, followed by LSTM modeling and training of the decomposed sequences. The resulting predictions are then reconstructed to obtain the final load forecast. Case study analyses validate the feasibility of this method, indicating its ability to increase the accuracy of reactive power load forecasts, which aids in more rational deployment and management of reactive power in electrical systems.