Performance Evaluation of Neural Network-Based Short-Term Solar Irradiation Forecasts

Due to the globally increasing share of renewable energy sources like wind and solar power, precise forecasts for weather data are becoming more important. To compute such numerous authors apply neural networks (NN), whereby models became ever complex recently. Using irradiation as an example, we verify if this additional complexity is required in terms forecasting precision. Different NN models, namely long-short term (LSTM) network, a convolutional network (CNN), combinations both benchmarked against each other. The naive forecast included baseline. Various locations across Europe tested analyze models’ performance under different climate conditions. Forecasts up 24 h advance generated compared using goodness fit (GoF) measures. Besides, errors analyzed time domain. As expected, error all increases with rising horizon. Over test stations it shows that combining LSTM CNN yields best performance. However, regarding chosen GoF measures, differences alternative approaches fairly small. hybrid model’s advantage lies not improved but its versatility: contrary or CNN, produces good results