Models And Learning Optimization Methods For Economic Dispatch Of Power System Considering Load Flexibility And Deep Peak Regulation

With the development and construction of technologies and devices such as large-scale energy storage on the load side,flexible load and active distribution network,the characteristics of the load side of the power system are bound to change a lot in the future,and its demand elasticity and dispatchable ability are bound to be constantly displayed.On the other hand,pumped storage power station,deep peak regulation and other technologies will further increase the dispatching capacity and range of the main body of the power side,and the dispatching flexibility of the power side will gradually become prominent.In view of these two points,the source charge flexible resources widely distributed on both sides is the key to the future power grid characteristics,fully tap and use elastic potential on both sides of the source charge,organic dynamic coordination of supply and demand relations,is to promote new energy given,one of the key means of power grid operation efficiency,also is one of the important direction of the smart grid development in China in the future.Therefore,the thesis mainly did the following work:Firstly,from the point of view of the load side,the thesis studies the load flexible,introduces the behavior model of price-based demand response to changes in electricity prices,and gives a hierarchical response behavior model for the incentive-based demand response,and digs deeper The flexible dispatchability of this type of load has established a mathematical model for the economic dispatch of power systems that considers the flexibility of load side,including price-based demand response and incentive-based demand response.When solving the problem,in order to solve the influence of random factors and deal with the continuous state variables in the system,the dueling DDQN(dueling double deep Q-network)algorithm in deep reinforcement learning is introduced,and the deep neural network is used to approximate the value function.The simulation results show that incorporating load side flexible resources into dispatching can reduce the load abandonment rate,and compared with deep Q-network(DQN)algorithm,dueling DDQN has faster convergence speed and better stability.Secondly,from the point of view of the source side,the thesis introduces the conventional peak regulation,deep peak regulation without oil and deep peak regulation with oil,and the corresponding peak regulation model and cost calculation method for thermal power units.Furthermore,the thesis analyzes and studies the adjustment flexibility of thermal power units participating in dispatch operation under the deep peak load regulation mode,and establishes a mathematical model of the economic dispatch problem of the power system that comprehensively considers the bilateral flexibility of the source and load.Then,a source and load flexible resource collaborative scheduling optimization method based on deep reinforcement learning technology is designed.The effectiveness of the proposed model is verified through simulation experiments,and the impact of different peaking capacity and peaking depth on the economics of system operation is analyzed.