Synergistic cyclic optimization strategy for the data screening and forecasting of solar power, Wind power, and electricity load

In response to the challenges posed by the increased penetration of renewable energy systems on the stability of power systems, this paper presents a synergistic cycle optimization strategy for data screening and forecasting of solar power, wind power, and electricity load. Initially, an outlier detection algorithm based on normal distribution is employed to identify outliers within the dataset. Subsequently, adaptive K-nearest neighbors (KNN) interpolation algorithm based on clustered power change rate (AKNN-CPCR) is utilized to effectively interpolate the detected outliers. Following data preprocessing, the processed data is used to train a predictive model. Finally, based on the predictions generated by the model, the parameters of the cycle optimization algorithm are adjusted in reverse to achieve optimal forecasting performance. The effectiveness of the proposed method was rigorously validated using load, wind, and solar datasets across three deep learning models: Convolutional Neural Network Bidirectional Long Short-Term Memory (CNN-BiLSTM), Convolutional Neural Network (CNN), and Bidirectional Long Short-Term Memory (BiLSTM). Experimental results confirm that the proposed algorithm consistently enhances prediction accuracy in all scenarios and datasets, demonstrating its robustness and generalization capability for complex time series forecasting tasks.