Research On Real-time Price Optimization Algorithm Based On Neural Networks

With the development of social economy and the progress of technology,people have more and more dependence on electric power system.At the same time,the dual pressures of resources and the environment make people begin to look at the huge potential of the power system in energy conservation and emission reduction.The development prospect of Smart grid is very broad,and has a huge role in stimulating the economy,has an immeasurable value to ease the pressure on the environment and resources.The development of smart grid in China is still in its infancy,its connotation and extension will continue to enrich.At present China's price is regulated by the government,the electricity price under the control of the government can not truly reflect the relationship between the power supply and demand,and due to the lack of effective adjustment measures,it will cause a huge waste of resources when the power demand fluctuates.Therefore,it is very meaningful to study the real-time price,and the flexible price can bring win-win to users,enterprises and society.Although some researches have been made on the optimization of electricity price,some problems still exist.For example,the optimized price is not the optimal price,which is the convergence of the optimization algorithm is not good.Also,the time of optimizing price is relatively long,that is,the convergence rate of the optimization algorithm is slow.As an intelligent optimization algorithm,neural network has the advantages of parallel computation and fast convergence in solving complex optimization problems.Therefore,it is of significance to introduce the neural network into the real-time price optimization problem.Firstly,in the second chapter,a new neural network model is proposed to solve the nonlinear optimization problem in order to solve the disadvantages of the various methods of solving nonlinear optimization problems.This kind of neural network can be realized by the circuit,and has the advantage of simple calculation.The global solution of the neural network and the existence,uniqueness and boundedness of the neural network is analyzed,and the solution can converge to the feasible region in the limited time,finally the global attractivity of the neural network is analyzed.Two numerical examples verify the validity and superiority of the neural network model.Then,in the third chapter,several models of real time electricity price in smart grid are given,and the general solutions are also given.The fourth chapter uses the sub-gradient based neural network to optimize the real-time price optimization based the maximization interests of users and suppliers in smart grid.According to the comparative analysis with the distributed algorithm and some neural network,the simulation results show that the neural network is correct and effective,and the optimization effect is better.But there are some limitations in the fourth chapter.In the fifth chapter,further research is carried out by using the improved neural network in the second chapter.Through the simulation and comparison,the neural network improved in the second chapter dealing with the problem of real-time price has better effect,also has the nature of global attractive and is more suitable for real-time optimization.Finally,the thesis summarizes the work done,and points out the problems and shortcomings and the problems that need to be studied in the future.