Research On Multi-source Coordinated Optimal Dispatch For Power System Based On Data-Driven Robust Optimization

With the large-scale grid-connection of wind power and photovoltaic power generation,its inherent uncertainty brings great challenges to the scheduling decision of power system.The scheduling scheme obtained by the traditional optimal scheduling method can no longer meet the practical needs,either economically or practically.Therefore,it is necessary to model the uncertainty of intermittent power supplies,study the optimal scheduling methods that suitable for the grid-connection of large-scale wind power and photovoltaic,and give a feasible scheduling scheme.In addition,as a kind of efficient and clean energy,nuclear power is developing rapidly in China.However,a large number of nuclear power plants are connected to the network as a basis load power source,which brings huge peak-load operation pressure to the power grid and seriously squeezes the power generation space of the other power sources,which is not conducive to the absorption of large-scale renewable energy.Taking Guangxi province as an example,with the large-scale access of wind and photovoltaic power in Guangxi power grid,as well as the slow growth of electricity demand in Guangxi,coupled with the absence of pumped-storage power plants to assist the operation of Fangchenggang nuclear power plant,the wind,hydro,thermal and nuclear power in this region have prominent contradictions,and wind and water spillage occur from time to time The large-scale grid-connection and absorption of wind and photovoltaic power and the safe operation of nuclear power have put forward new requirements for the flexibility and regulatory capacity of power systems.Therefore,it is necessary to discuss the coordinated and optimal scheduling problems of multiple power sources with different characteristics,including nuclear power.In this paper,unit commitment(UC)is taken as the breakthrough point to study the coordinated optimization scheduling problems of multiple power sources,including wind,photovoltaic and nuclear power,hydro and thermal power.There are mainly the following four parts:1)The data-driven robust optimization method is studied in this paper.The kernel density estimation and robust kernel density estimation,which are non-parametric statistical methods that can mine the probability density and distribution information from big data,are analyzed.The key technologies to realize robust kernel density estimation,such as robust loss function,kernel function smoothing parameter selection,and kernelized iteratively reweighted least squares algorithm,are studied in depth.According to the obtained distribution information,a data-driven uncertainty set is constructed,and then,a general framework of data-driven robust optimization model for power systems is established.2)In view of the conventional unit commitment is non-convex and its globally optimal solution is hard to be obtained in finite time,an equivalent convex quadratic programming model for unit commitment is proposed.Based on two kinds of 0-1 variables,an equivalent linear expression of the start-up cost considering cold and hot start-up cost is constructed to make it simpler and easier to calculate.Through a set of special inequalities,the nonconvex minimum on and off time constraint is equivalently transformed into a linear constraint.Thus,the non-convex UC problem could be equivalently transformed into a convex quadratic programming problem where the objective function is quadratic and the constraint condition is linear.The globally optimal solutions are obtained for the first time by commercial optimization solver GUROBI,which may serve as a comparative reference for relevant scholars.In addition,by combining the proposed convex quadratic programming model with the data-driven robust optimization method,a data-driven robust convex quadratic programming model for UC problem is proposed,which takes into account the uncertainty of wind and photovoltaic power.3)A multi-source coordinated optimal dispatching model considering the dispatchability of nuclear power units is proposed.Combined with the actual operation experience and the simulation experiment results of nuclear power units,the schedulability of nuclear power units,namely the feasibility of nuclear power participating in peak load regulation,is analyzed.According to the operating characteristics of nuclear power units,its safe adjustable range and safety constraints of peak-shaving operation are proposed,and the peak-shaving operation model of nuclear power is established.In view of the particularity that nuclear power units cannot be frequently adjusted and some safety limits,a two-stage dispatching method combining pre-dispatch and re-dispatch is adopted to obtain the low-power operation time and optimal peak regulation depth of nuclear power units,respectively.Robust optimization is used to deal with the uncertainty of wind and photovoltaic power,which can always ensure the feasibility of scheduling decisions.Thus,the short-term optimal scheduling model of multi-source coordination of wind,photovoltaic and nuclear power,hydro and thermal power is constructed.4)In this paper,the dynamic coordination of hydro and thermal power,wind,photovoltaic and nuclear power is studied,and a multi-time scale robust scheduling model for multi-source power systems is proposed.According to the characteristic that the accuracy of wind power prediction increases with the decrease of the time scale,the corresponding well-matched scheduling model is established on day-ahead,intraday and real-time time scales,respectively.In other words,at the day-ahead scheduling stage,the robust kernel density estimation is used to mine the big data information of wind and photovoltaic power.Then,according to the obtained probability density information,the data-driven uncertainty set is constructed through the quantile function.Next,a multi-source joint optimization data-driven robust scheduling model with the target of minimizing the total operating cost of thermal units,peak regulation cost of nuclear power plants and water spillage loss of hydropower stations is constructed.A rolling optimization model with minimum water spillage and minimum thermal power adjustment cost is established at the intraday rolling scheduling stage.At the stage of real-time scheduling,a real-time scheduling model with minimum thermal power adjustment is constructed.In addition,in view of the shortcomings of the traditional water spillage strategy,such as not reflecting the principle of fair dispatch,ignoring the interests of each hydropower station and not conducive to creating social benefits,and combined with the actual demand of each hydropower station hoping to maximize its own interests,a category of water spillage method that can make the water spillage amount or abandoned hydropower amount distributed fairly in each hydropower station is proposed This method includes eight water spillage strategies,and there are eight alternative minimum objective functions for water spillage,which correspond to the loss of water spillage or abandoned hydropower loss under different water spillage strategies.The proposed method overcomes the disadvantages of the traditional water spillage strategy,and effectively takes into account the interests of various hydropower stations and the overall optimization of the whole basin.