Study On Multi-energy Complementarity And Multi-Scale Optimal Scheduling Of Hydro-wind-Photovoltaic In The Middle And Lower Reaches Of Yalong River Basin

The development of clean energy bases and the construction of inter-regional UHV DC transmission channels are important measures for the implementation of new power systems in China,and are also important ways to achieve the goal of "double carbon".However,with the randomness,volatility and intermittent characteristics of renewable energy installed capacity and power penetration rate gradually increase,the safe and stable operation of the power grid has been challenged.Apply system engineering and multi-energy complementary technologies to establish a collaborative optimization system between multi-energy sources and between sources,networks,and loads to form a scheduling mode of flexible sharing and bundled delivery,which has become a key means to increase the consumption of new energy.This paper focuses on the inter-regional power transmission scheduling problem of clean energy bases under the uncertainty of wind and photovoltaics,and takes the cascade hydropower,wind power,and photovoltaic systems in the middle and lower reaches of the Yalong River as the research object to carry out multi-scale stochastic optimal scheduling research.The main research results are as follows:(1)Aiming at quantifying the uncertainty of wind and solar output,the forecast box,multivariate normal distribution and Copula function are used to describe the conditional dependence between the new energy forecast deviation and the forecast sequence,and the temporal and spatial correlation of the historical sequence of wind power output.A new energy output uncertainty scenario generation technology based on inverse transformation sampling,covariance stochastic process and Copula conditional probability is proposed.And through Kmeans clustering to reduce the scenario,the typical wind and solar output uncertainty combination scenarios is generated.The results show that the generated typical scenarios can take into account both temporal correlation and spatial correlation,and can be used as input information for multi-energy complementary stochastic optimal scheduling model.(2)Focusing on the short-term and real-time scheduling of cross-regional power transmission in the Yalong River clean energy base with uncertain wind and solar output,a dayahead multi-grid peak-shaving optimal scheduling model based on expected scenarios and an intraday rolling optimal scheduling model based on chance constraints are established.Then,a Mixed integer linear programming(MILP)transformation scheme for the above two stochastic optimization models is proposed.The results show that wind and solar power curtailment rate in typical day of the non-flood season is 0.56%,which occurs in the stage of limited flexibility of hydropower reduction in load valley.The renewable energy power curtailment rate in typical of the flood season is 16.68%,which occurs during load valleys and ramps.After intraday rolling optimization,the consumption rate of new energy can be greatly improved.However,with the increase of confidence level in typical day of the non-flood season,energy bases will gradually increase their power purchases from large power grids to make up for the lack of flexibility during peak load periods.The research shows that the duration of high photovoltaic power generation and the adjustable water volume of hydropower are the key factors affecting the daily dispatch and operation of the cascade hydro-wind-solar system in the middle and lower reaches of the Yalong River.(3)Focusing on the mid-and long-term scheduling of hydro-wind-solar systems,a shortterm compensation benefit quantification method for hydropower,wind power and photovoltaics based on the principle of power balance is proposed for the typical cross-regional power transmission process.Then,a mid-and long-term dispatch model of hydro-wind-solar energy coupled with short-term compensation benefits is constructed.It is difficult to solve the model,so a parallel dynamic programming algorithm based on message passing interface(MPI)is proposed.The results show that the medium-and long-term dispatching model built can take into account complementary benefits and power generation benefits,and can provide a reliable boundary for water volume control and discharge for the economic operation of the short-term hydro-wind-solar system.Under the environment configuration of the supercomputing platform with 192 cores,the time-consuming of the parallel dynamic programming algorithm is 1.43 h,which can meet the timeliness requirements for medium and long-term optimal scheduling.(4)For the extraction of medium-and long-term scheduling rules,a long series of implicit stochastic optimization is carried out through parallel dynamic programming,and a "knowledge base" for medium-and long-term scheduling decisions is generated.On this basis,a scheduling rule extraction model based on machine learning is established to realize the fitting of the nonlinear mapping relationship from sequence to sequence.Compared the simulation effect of multi-layer perceptron(MLP)and convolutional neural network-long short-term memory network(CNN-LSTM),and verified the applicability and superiority of CNN-LSTM.