Distributionally Robust Dispatch Of Multi-regional Power Systems Considering The Correlated Variable Wind Power

Wind power has obvious uncertainty.As the proportion of wind power in the power system continues to increase,the uncertainty of the power system operation increases and the balance of safety and economy of the power system is challenged.Therefore,accurate modeling of wind power uncertainty is increasingly important for the dispatch of the power system.In response to this problem,a large number of studies have applied the uncertain decision-making method of operation research to the dispatching problem of the power system,and good results have been achieved.Among them,stochastic programming and robust optimization have been extensively studied.However,stochastic programming requires high accuracy of data prediction.When effective historical data is insufficient,stochastic programming cannot meet the requirements of the security of the power system.Robust optimization is optimized for the worst scenario of uncertain parameters.The worst scenario has a large deviation from the normal value and a small probability of occurrence,which causes the robust optimization method too conservative.Distributionally robust optimization combines the advantages of stochastic programming and robust optimization.It uses ambiguity sets to characterize the characteristics of the probability distribution of uncertain parameters,which can better model the uncertainty of wind power.However,the current moment-based distributionally robust optimization does not properly consider the correlation between wind power,which leads to conservative results.On the other hand,the power generation resources are far from the load centers in the power system,which is difficult for a single regional power system to achieve a balance between supply and demand.In terms of the above problems,the main work of this thesis is as follows:(1)In response to the random volatility of wind power and the correlation between the geographically adjacent wind power a method of constructing a reduced-dimensional ambiguity set based on principal component analysis is proposed.Which performs time-space transformation and dimensionality reduction on wind power to obtain the transformed variable of wind power.The uncertainty of wind power is indirectly considered by describing the uncertainty of the transformed variable.In the description of uncertainty,an ambiguity set characterizing the probability distribution of uncertain parameters based on the first and second moments of the uncertain parameters is established.The effectiveness of the proposed dimensionality reduction ambiguity set is verified by case studies.(2)To deal with the uncertainty of wind power and solve the problem of the power generation resources are far from the load centers in the power system.In this thesis,multiregional power system collaborative optimization is used to absorb wind power and balance the safety and economy of the power system.A distributionally robust dispatch model for multiregional power systems considering the correlation of variable wind power is established.The dispatching model is a two-stage problem model.The first-stage problem is the planned problem before the wind power output is realized,and the second-stage problem is the redispatch problem based on the worst distribution of wind power output after the wind power output is realized.(3)According to the different operation modes of the multi-regional power system,the centralized and decentralized methods are used to solve the distributionally robust dispatch model.For centralized optimization,duality theory is used to transform the model with the expected value-form into a semi-definite programming model,the semi-definite programming model is relaxed,and the delay constraint generation method is used to iteratively solve the relaxed semi-definite programming problem and bilinear problem.The economic efficiency and the solution efficiency of the model proposed in this thesis are verified by case studies.For the multi-regional power system where the information between power systems is private or different power systems belong to different operators,the decentralized optimization method is used to solve the multi-regional power system distributionally robust dispatch problem.The alternating direction multiplier method is used to decouple the multi-regional power system model.Different regions only need to exchange the phase angles of the boundary nodes,which ensures the privacy of the internal data of the regional power system.The convergence and economy of the decentralized distributionally robust dispatch are verified by case studies.