Reinforcement Learning And Fuzzy Neural Network Based Optimized Control For The Coordinated System In Thermal Power Unit

Since the 13 th five-year Plan,with the in-depth strengthening of reform and independent innovation in China’s power industry,ultra(super)critical units have become the main power generation force of thermal power plants in China because of large capacity and high thermal efficiency.The key points and difficulties of the control are mainly reflected in the cooperative work of boiler-steam turbine system and ensuring that the unit meets the requirements of rapid load change and deep peak regulation of the power grid.The improvement of coordinated control quality of supercritical(supercritical)units is of great significance to give full play to the peak regulation and frequency regulation ability so as to ensure the safe and efficient operation of the unit.Therefore,in order to improve the coordinated control quality of supercritical units,this paper deeply studies a method of inverse compensation control and optimization of coordinated system based on reinforcement learning and fuzzy neural network,and proposes a self-learning fuzzy neural network inverse controller(Self-study Fuzzy Neural Network Inverse Controller,SFNNIC)which can ensure that variables track a given expected signal.The controller can realize model self-study when the model is imprecise due to the change of object characteristics.Firstly,the Generalized dynamic fuzzy neural network(GD-FNN)inverse model of load and main steam pressure are designed based on the operation characteristics of supercritical units,and then the coordinated inverse controller following the desired signal is designed by using inverse control theory.Finally,a more detailed coordinated control scheme is proposed by combining reinforcement learning and on-line construction of training set.The control scheme can evaluate the generalization of the GD-FNN model in the process of unit operation,improve the structure and the accuracy of the control model and is more in line with the actual industrial requirements.In the simulation research,taking a 600 MW supercritical unit as the object,the control algorithm is developed based on the MATLAB platform,and a detailed simulation test is carried out with the help of the power plant simulator.The results show that using the designed Self-study Fuzzy Neural Network Inverse Controller(SFNNIC),the speed of unit load following and the stability of main steam pressure are significantly improved,which improve the control effect of supercritical unit coordination system compared with the original PID control of the unit,which effectively improved coordinated control quality of the unit.