Short-Term Solar Photovoltaic Power Forecasting Using a Multiple Similar-Day Method Based on Weather Type

With the amendment of relevant laws in recent years, Taiwan's renewable energy is moving towards an open market. Although it increases the freedom and bargaining power of private companies to sell electricity, it also means that companies must prepare their own energy storage equipment and grid dispatch plans. In the field of short-term solar power forecasting, the accuracy of power generation forecasting depends largely on the quality of the predicted radiation in the input variables. In the past, some studies have achieved good results by using the radiation of similar days as the predicted radiation. However, there are still some shortcomings in the method of selecting similar days. In this study, an improved method for similar days is proposed to reduce radiation error due to the sampling error of similar days by using multiple similar days; it is demonstrated that the improved method can reduce the error of radiation prediction (nRMSE) by 10% and the error of power generation prediction (nRMSE) by 2%.