DLinear photovoltaic power generation forecasting based on reversible instance normalization

Accurate and timely PV power generation forecasting is of great significance for maintaining the stability of the power system. In order to improve the forecasting accuracy and forecasting speed of PV power generation, this paper proposes a DLinear PV power generation forecasting model based on reversible instance normalization - RevIN-DLinear. The model firstly uses normalization and denormalization to alleviate the distribution shift problem of PV data over time, and secondly combines the advantages of the DLinear network model: the time series analysis method of seasonal-trend decomposition improves the predictability of PV data; the linear network to directly performs multi-step forecasting to avoid error accumulation. The experimental results show that the forecasting accuracy of the model is higher than that of the original DLinear and the advanced Transformer series models, and the speed is much better than that of the Transformer series models.