Investigating the influence of switching from load forecasting to net load forecasting on the forecasting accuracy

Abstract

Forecasting the electric load is important in the energy sector for ensuring a stable and economic power system operation. However, the increasing integration of non-dispatchable renewable energy sources, such as wind and solar power, into the grid has created a need for predicting the net load, that is, the amount of load remaining after renewable energy sources cover a portion of the total energy demand. The scientific literature contains numerous forecasting techniques and methods that can be used to forecast the load or the net load.  Electricity companies may rely on a particular load forecasting method and apply it to net load forecasting as well. Therefore, this paper investigates whether this approach yields similar forecasting accuracy.

The research utilizes Adaptive Neuro Fussy Inference Systems (ANFIS) as an advanced artificial intelligence-based forecasting technique to predict the load and the net load using real historical data from the Spanish electricity grid which is characterized by a significant contribution of wind and solar power. The study also uses the root mean square error (RMSE) and mean absolute error (MAE) as evaluation criteria. The results show that, when using the same forecasting method, the accuracy of net load forecasting is lower than that of load forecasting. Therefore, when planning to integrate additional wind and solar power into the electricity grid, electricity companies should not rely on the same method for forecasting the net load but should take the necessary measures to restore the accuracy of net load forecasting to the desired level