Application Of Neural Network Algorithm In Energy Dispatch And Demand Response Problems

Smart grid is the development trend of the future grid.It uses two-way power and information flow to build an automated distributed advanced energy transmission network,which complements and enhances the traditional power grid.Smart grid uses information,two-way,safe communication technology and computing intelligence as the core,including power generation,transmission,distribution,substation,consumption and other functions,to achieve clean,safe,reliable,flexible,efficient and sustainable power system.In this paper,two energy scheduling strategies are designed for the optimization of energy dispatching in smart grid,and the corresponding algorithms are proposed to find the optimal solution.The first one is a demand response management strategy based on real-time electricity price.Demand response is the reward that end users pay to alter their normal patterns of electricity use as prices change over time,or to reduce electricity consumption when wholesale market prices are high or system reliability is threatened.By facilitating user interaction and response,demand response brings economic benefits to users and utilities and makes the entire grid more stable and reliable.There are three types of user loads in the strategy designed in this paper: base load,elastic load and energy storage device.The base load consumes a fixed amount of power at each time period and cannot be adjusted.Elastic load has its own expected power consumption,the more power actually consumed,the higher the user satisfaction.In order to reflect the relationship between power consumption and user satisfaction,a satisfaction function is constructed.Energy storage devices can store electricity when the real-time electricity price is low and release it for users to use when the price is high,bringing some benefits to users.The zero-one variable is used to distinguish the charging and discharging states of energy storage devices.The final objective function is a nonconvex optimization problem.In order to find the global optimal solution of the problem,an inertial neural network is proposed and combined with particle swarm optimization algorithm.Simulation examples and comparison experiments show that the inertial neural network has faster convergence speed,the combined algorithm can find the global optimal solution of the problem,the demand response management strategy can make the electricity consumption in each time period smoother,so as to improve the efficiency of the smart grid.The second one is a fully distributed energy scheduling strategy considering both user side and generator side.In recent years,distributed optimization problem has been widely studied.The objective function for this type of problem is the sum of multiple objective functions,each of which represents the contribution of an agent to the global objective function.Due to the advantages of distributed algorithm in large-scale optimization compared with centralized algorithm,and the low computational cost and good performance of each agent or node,many distributed algorithms have been proposed to effectively solve the convex optimization problem in multi-agent networks.The generator side designed in this paper includes traditional power grid,distributed generator and new energy,which has their own price function.The user side has added electric vehicles.Electric vehicles have the advantages of low emission and efficient emission reduction,and will become an important part of the family in the future.When it is parked at home,it can be used as an energy storage device to charge and discharge,thus gaining some benefits,but it has its own power demand for the next trip itself,so it is different from the model of ordinary energy storage devices.In order to solve the nonconvex optimization problem of the user side,a dynamic algorithm combining neural network algorithm and differential evolution algorithm is proposed.In order to solve the convex optimization problem of the generator side,a distributed algorithm is proposed to deal with general constraint problems.Simulation examples and comparison experiment show the combination of the dynamic algorithm can find global optimal solution effectively,and has faster convergence speed,the distributed algorithm can effectively find the global optimal solution,distributed energy scheduling policy can impact on the power consumption peak peel,and maximize the protection of the privacy on the user side and generator side.