Research on a single-phase multi-circuit energy meter scheduling scheme based on machine learning algorithm

The strong stochasticity of photovoltaic (PV) power generation makes the designing of grid scheduling schemes more challenging. To make scheduling more forward-looking, it is crucial to accurately predict PV power generation. Therefore, this paper presents a scheduling scheme for predicting PV power using single-phase multi-circuit energy meters as a priority. Firstly, from the point of view of energy meter hardware, multiloop can be calibrated by vector sums to guarantee accurate metering of electrical energy on all three sides. Secondly, a PV management module is added to capture multidimensional data related to PV power. Considering the relevance of meteorological data and the available computational resources, the proposed scheme applies the radial basis function-particle swarm optimization (RBF-PSO) neural network to predict PV power from DKASC open-source data to verify the feasibility of the algorithm. The prediction results within 3 days, running time and prediction curves are comprehensively analyzed. It can be found that the RBF-PSO neural network has a higher practical application value than CNN and GRU models, which further validates the rationality and feasibility of the scheduling scheme.