Thermostatically Controlled Loads Regulation Strategy In Smart Grid Based On Neural Network

Nowadays,facing the new situation of increasingly scarce power resources,smart grid construction has become the direction of power system reform in the world.Smart grid is based on the rapid development of automation technology,communication technology and new energy technology,realizing intelligent grid dispatching and information sharing to ensure the safe,reliable,economic,efficient,clean and environmentally friendly operation of power system.Frequency regulation service of power grid is closely related to the operation of power market,the safety of power generation side and power consumption side equipment.Demand-side response has changed the user-side power consumption mode,optimized the allocation of power resources,and provided a new idea for stabilizing power grid frequency and ensuring power system security.Based on the operation characteristics and control methods of thermostatically controlled loads and the advantages of neural network in control and optimization,this paper studies the control methods and dispatching strategies of demand side thermostatically controlled loads in frequency regulation service of smart grid,which is of great significance for the further development of smart grid and the construction of auxiliary service market.The main research contents are as follows:Firstly,an on-line real-time optimized controller based on back propagation(BP)neural network is designed.By using the temperature setting value adjustment strategy of thermostatically controlled loads,the frequency regulation signal can be effectively tracked and the auxiliary service of frequency regulation can be improved.Secondly,a frequency regulation strategy of thermostatically controlled loads based on fuzzy neural network is proposed.A fuzzy neural network controller is designed based on the fuzzy information expression and logical reasoning ability of fuzzy control and the self-learning ability of neural network.Particle swarm optimization(PSO)and BP algorithm are combined to optimize the initial values of connect weights and adjustable parameters of membership function in the training process of the fuzzy neural network,avoiding the local minimum and accelerating the convergence speed.Finally,a hybrid control strategy of thermostatically controlled loads considering user satisfaction is studied to provide frequency regulation service.The control strategy divides thermostatically controlled loads into two clusters according to the control methods of thermostatically controlled loads.Based on the cerebellar model articulation controller(CMAC)neural network,least mean square(LMS)algorithm is used to optimize the frequency modulation distribution signal in real time.Combining the advantages of switch priority control and temperature setting value adjustment,the consumption of thermostatically controlled loads in demand side can track the frequency modulation signal of power grid.The control strategy also uses the fuzzy comprehensive evaluation method and the comprehensive evaluation function to real-time evaluate the control effect of the system,and feedback the results to the control system,so that the control system can improve the tracking accuracy of thermostatically controlled loads and optimize the user satisfaction.