Abstract: Aiming at the problem that large-scale access to the grid by distributed photovoltaic and electric vehicles (EVs) will cause an impact on the power system, we develop a distributed PV-EV complementary scheduling model.With the goals of smoothing the fluctuation of PV grid connection and increasing the economy of EV users, the factors such as the stochasticity of PV output, load power fluctuation, stochasticity of EV access time and power, real-time tariffs, battery aging cost and so on are taken into consideration, and an improved proximal policy optimization algorithm with stochastic perturbation of the gradient is proposed to solve the problem. Meanwhile, the model objective function is adjusted to obtain 2 real-time operation strategies based on different optimization objectives. It can be seen from the arithmetic example that the real-time scheduling strategy can effectively smooth outlet power fluctuations, and the scheduling effect is improved by 3.48% compared with the traditional PPO algorithm. Strategy 1 takes the user's travel demand and smoothing the power fluctuation of the parallel network as the primary objectives, which can ensure the user's 24 h demand, and at the same time, the power stabilization rate of the parallel network reaches 91.84%. Strategy 2 takes the user's economic benefit as the primary optimization objective, and the EV benefit of participating in dispatching for the whole day can reach RMB 82.6, which can encourage the users to participate in dispatching.