Distributed Photovoltaic Net Load Forecasting in New Energy Power Systems

Abstract

To respond to the demand of achieving carbon peaking and carbon neutrality goals, and to construct a complete “source-grid-load-storage” new energy power system, a distributed photovoltaic net load forecasting model based on Hamiltonian Monte Carlo inference for deep Gaussian processes (HMCDGP) is proposed. First, direct and indirect forecasting methods are used to examine the accuracy of the proposed model and to obtain spot forecasting results. Then, the proposed model is used to perform probability forecasting experiments and produce interval prediction results. Finally, the superiority of the proposed model is verified through the comparative experiments based on the net load data of 300 households recorded by Australia Grid. After obtaining the exact net load probabilistic forecasting results, the photovoltaic production can be fully utilized via power dispatch, which can reduce the use of fossil energy and further reduce the carbon emission.

Keywords

• net load forecasting
• photovoltaic production
• deep gaussian process
• point forecasting
• interval prediction