Accurate forecasting method of new energy power in multi-temporal and spatial scales by CNN-BiLSTM

To address the issue of disregarding weather factors in the current prediction process for new energy power, which leads to inaccurate predictions for new energy power stations, a novel approach is proposed. This method utilizes a combination of CNN-BiLSTM models to achieve accurate predictions across multiple time and spatial scales. According to the known data, the data gaps are filled, and the interval average of time series is calculated. On this basis, the data of multi-time and space scale accurate prediction of new energy power are normalized. Combining CNN and BiLSTM, the specific steps of accurate prediction of new energy power in multiple time and space scales are designed. The experimental findings demonstrate that this approach exhibits excellent predictive performance, outperforming other methods in terms of two evaluation metrics: mean square error and average absolute error rate for the accurate forecasting of new energy power across multiple time scales.