Probabilistic net load forecasting based on sparse variational Gaussian process regression

The integration of stochastic and intermittent distributed PVs brings great challenges for power system operation. Precise net load forecasting performs a critical factor in dependable operation and dispensing. An approach to probabilistic net load prediction is introduced for sparse variant Gaussian process based algorithms. The forecasting of the net load is transferred to a regression problem and solved by the sparse variational Gaussian process (SVPG) method to provide uncertainty quantification results. The proposed method can capture the uncertainties caused by the customer and PVs and provide effective inductive reasoning. The results obtained using real-world data show that the proposed method outperforms other best-of-breed algorithms.