Online Solar Radiation Forecasting under Asymmetric Cost Functions

Grid operators are tasked to balance the electric grid such that generation equals load. In recent years renewable energy sources have become more popular. Because wind and solar power are intermittent, system operators must predict renewable generation and allocate some operating reserves to mitigate errors. If they overestimate the renewable generation during scheduling, they do not have enough generation available during operation, which can be very costly. On the other hand, if they underestimate the renewable generation, they face only the cost of keeping some generation capacity online and idle. So overestimation of resources create a more serious problem than underestimation. However, many researchers who study the solar radiation forecasting problem evaluate their methods using symmetric criteria like root mean square error (RMSE) or mean absolute error (MAE). In this paper, we use LinLin and LinEx which are asymmetric cost functions that are better fitted to the grid operator problem. We modify the least mean squares (LMS) algorithm according to LinLin and LinEx cost functions to create an online forecasting method. Due to tracking ability, the online methods gives better performance than their corresponding batch methods which is confirmed using simulation results.