Forecasting photovoltaic energy using MEWMA models

In any electricity market, both conventional and renewable producers trade their energy in the day-ahead market. Thus, renewable generating companies require a forecasting tool in order to predict the quantity of electricity that they will produce in the short-term future. This paper proposes a forecasting procedure for solar power producers, based on the multivariate exponential smoothing model and employing the solar radiation next-day forecast from an external agent. The model is estimated using the EM algorithm. The proposed method is general and can be straightforwardly implemented in any photovoltaic power producer. A case study is analyzed, using real data from the Spanish energy market, using the MAPE as comparison metric. Results denote that the developed algorithm is both numerically accurate and computationally efficient.