A Spatial Forecasting Model for Solar PV Generation and Its Application in Household Electricity Management

A time-series data that depends on sunlight, such as solar photovoltaic generation output, principally consists of double sinusoidal components since solar irradiance reaches the Earth's surface while it spins and orbits. In addition, a part of solar energy is absorbed in the atmosphere by water vapour, dust, and ozone. In other words, geographical coordinates and regional weather conditions influence the data patterns. Therefore, this study focuses on developing a novel forecasting model to capture double seasonal patterns by adopting the spatial information of the data such as sunshine duration, cloudiness, and geographical coordinates. To explain an intra-day cyclical movement, a sine wave function with a predicted magnitude is considered to be integrated as a main part of the proposed model. Besides, an additive seasonal exponential smoothing model, which is a classical decomposition approach widely used for short-term forecasts, plays a role in adjusting the step-by-step error of forecasting over the daily pattern of the sine wave. Aggregated (intra-day to daily) data and regional daily weather-related variables are a real-world data set used in an empirical analysis by a regression model. The numerical results showed how the performance of the proposed model at different time horizons (e.g., one-step and one-period ahead forecasts) compared with existing models by the mean error (ME), mean absolute error (MAE), and root mean square error (RMSE). Moreover, we present an application of the proposed forecasting model using a stochastic programming (SP) model to optimise electricity usage in a household with a photovoltaic system and an electric vehicle (EV). The several types of solar panel electricity systems, including on-grid and hybrid, with EV battery incentive schemes will be developed by the two-stage SP model to investigate how to balance day-ahead electricity supply and demand and minimise daily electricity costs.