Research On Two-layer Scheduling Of Energy Storage Battery Cluster For Load Smoothing

In recent years,with the continuous improvement of power supply and distribution requirements,energy storage technology has become increasingly indispensable for power grid.Energy storage technology plays an important role in regulating the frequency of electric power,suppressing the fluctuation of renewable energy grid connection and improving the power quality.This paper studies the two-layer scheduling problem of energy storage battery cluster.The upper-layer scheduling object is the whole energy storage battery system,which is used to generate the low-level scheduling instructions.The lower-layer scheduling object is each battery module,which makes each battery module track the upper-layer instructions.The main contents of this paper are as follows:The scheduling of energy storage battery cluster is divided into two layers.T The first layer adopts the deep reinforcement learning method to make decisions and studies the overall scheduling of the energy storage battery cluster.This layer outputs the instructions for the entire energy storage system.The second layer designs a collaborative control algorithm,which dispatches each energy storage unit in the cluster with a distributed strategy to complete the instructions sent by the upper layer.For the upper scheduling,the grid side lithium-ion battery energy storage scheduling problem with load suppression as the main objective is studied.Firstly,the mathematical model of lithium-ion battery energy storage was built,and the objective function was designed considering the construction cost and the depreciation cost of the battery.Then a Markov model is established for the problem,and appropriate state space,action space and reward function are selected.A simulation experiment on Py Torch platform was conducted to verify the effectiveness of the deep reinforcement learning algorithm.In the low-level scheduling,the detailed allocation and coordination of the upper scheduling instruction,namely the total charge-discharge power of the stored energy,among the battery modules was studied.Because of the large capacity lithium-ion battery is usually made up of many lithium-ion battery module,each module for manufacturing process and series-parallel state cannot exactly the same,cause the charged state of imbalance,appear even part of the battery power module is empty or full the extreme situation of electricity,the battery life and upper an adverse effect on the stability of the algorithm.Therefore,the initial assignment algorithm based on consistency and the online distributed adjustment algorithm are used to allocate and track the upper scheduling tasks.Through mathematical theoretical derivation,the correctness of the proposed algorithm is proved,and the effectiveness of the control algorithm is verified by MATLAB simulation experiment.To sum up,this paper studies the two-tier scheduling problem of energy storage battery cluster.The upper scheduling adopts deep reinforcement learning algorithm to make decisions,and the lower scheduling uses the consistency algorithm to reasonably perform the upper scheduling tasks.The effectiveness of the algorithm is verified through simulation experiments.