Accurate source-load prediction results are important for new energy consumption and improving grid stability. The traditional BP algorithm has defects such as easy to fall into local optimum and premature maturity, and the genetic algorithm is to optimise the BP parameters to improve the prediction accuracy. The GA-BP is used to establish PV power generation prediction model and household load prediction model respectively. Comparative simulation analysis reveals that the proposed genetic algorithm optimised BP has a smaller prediction error than the traditional BP and PSO-BP, which verifies the reliability of the proposed method.