Active power distribution network real-time random optimization scheduling method based on reinforcement learning

Abstract

The invention provides an active power distribution network real-time random optimization scheduling method based on reinforcement learning. The method comprises the following steps: establishing a mathematical model of the active power distribution network; proposing a dynamic planning formula of the original problem, constructing a value function for representing the influence of the current decision on all subsequent time periods according to the characteristics of the electric vehicle in the active power distribution network, and avoiding the calculation of an expected value by utilizing astate value function after the decision; training a value function to obtain an approximate value function; and performing recursive solution on the random optimization scheduling problem in the real-time scene of the active power distribution network by using the trained approximate value function so as to obtain the approximate optimal decision of each time period. During real-time operation, on one hand, the scheduling income can be more effectively improved, and on the other hand, a peak clipping and valley filling effect is also achieved. When random factors in the environment change more violently, the method is still applicable and can be close to the optimal solution, the calculation time in the real-time scheduling process is not affected, and high robustness is achieved.

Description

technical field [0001] The invention relates to the field of optimal scheduling of active distribution networks in power systems, in particular to a method for real-time random optimal scheduling of active distribution networks based on reinforcement learning. Background technique [0002] As distributed generators (Distributed Generator, DG), energy storage devices (Energy Storage, ES) and flexible loads (Flexible Load, FL) are widely connected to the power grid, the traditional distribution network is gradually becoming an active distribution network (ActiveDistribution Network, ADN) transformation. Different from the one-way flow of energy in the traditional distribution network, the word "active" in ADN now has a two-way flow of energy, that is, not only the main network can transmit power to the distribution network, but also the DG, ES, and FL in the distribution network can be reversed Transfer power to the main network. The continuous development of ADN brings a se...

AI Technical Summary

Problems solved by technology

However, the above method is designed for the value function of a single energy storage device, and cannot be simply applied to a large number of electric vehicles.

In terms of computational efficiency, the successive projection approximation method is used to update the value function (Liu Cuiping, Lin Shunjiang, Liu Mingbo, Jian Ganyang, Lu Wentian. Application of approximate dynamic programming algorithm to solve security-constrained stochastic dynamic economic scheduling problems [J]. Electric Power System Automation, 2016 ,40(22):34-42.SALAS D F,POWELL W B.Benchmarking ascalable approximate dynamic programming algorithm for stochastic control of grid-level energy storage[J].Informs Journal on Computing,2018,30(1):106-123 .), and use the method of projection operation to ensure that the slope is monotonously decreasing, thereby restoring the concavity of the value function, but this method has a slow convergence speed and is too dependent on the initial slope

Images