Short-Term Load Forecasting in a microgrid environment: Investigating the series-specific and cross-learning forecasting methods

Abstract A reliable and accurate load forecasting method is key to successful energy management of smart grids. Due the non-linear relations in data generating process availability issues, remains a challenging task. Here, we investigate application feed forward artificial neural networks, recurrent networks crosslearning methods for day-ahead three days-ahead forecasting. The effectiveness proposed evaluated against statistical benchmark, using multiple accuracy metrics. test sets are high resolution multi-seasonal time series electricity demand buildings Belgium, Canada UK from private measurements open access sources. Both FFNN RNN show competitive results on benchmarking datasets. Best varies depending metric selected. use cross-learning fitting global model has an improvement final accuracy.