Adaptive Robust Optimal Operation Considering Load And Photovoltaic Output Interval Prediction In Microgrids

In order to satisfy the ever-increasing demand for energy consumption and realize the use of global low-carbon energy,solidly promoting the development and utilization of renewable energy has become the focus of global energy development.Microgrid(MG)has become an important way to absorb distributed renewable energy.To reduce the negative impact of the intermittency and volatil-ity of load and renewable energy on the stable economic operation in MGs,it is necessary to research its uncertainty.Interval prediction is often used to characterize the uncertainty of load and renewable energy generation power.Robust scheduling with strong operability can obtain the optimal scheduling strategy of the system based on the prediction intervals without obtaining the probability distribution of uncertain factors.This has become an important measure for system optimal scheduling consider-ing uncertainty.Therefore,based on MGs,this thesis pays attention to the uncertain factors of MG load and distributed photovoltaic power generation,researches the interval prediction method and robust dispatch modeling method considering the uncertainty,and explores the robust optimal opera-tion scheme,which improves the level of renewable energy consumption and load supply capacity of MGs.The detailed research contents are as follows:1)Load interval prediction model based on linear mixed integer programming.In order to adaptively obtain high-quality prediction intervals,based on the evaluation index of the prediction intervals,an interval prediction optimization model that minimizes the average width of the interval is proposed.That is,minimize the average width of the interval on the basis of ensuring the interval coverage probability.Integer variables are introduced to represent the coverage of the prediction intervals to consider the constraint of the coverage probability of the prediction intervals,and a mixed-integer programming model based on linear regression is established.The quantile regression model based on linear regression is used to obtain the predictive sub-interval? the integer variable value of the sample in the sub-interval is pre-defined to speed up the model solving process.Numerical experiments based on actual load data compared other methods and verified the superiority of the proposed method in predictive performance.2)Combined interval prediction model of photovoltaic power based on the biased convex cost function and auto-encoder.Aiming at the deficiencies of the interval prediction method based on the upper and lower bound estimation framework,a new type of convex biased loss function is proposed by employing the characteristics of the sigmoid function.The extreme learning machine(ELM)is selected as the base prediction engine of the upper and lower limits,and the input weight matrix of the ELM network is initialized by auto-encoder technique.Based on the prediction engine of the upper and lower bounds,a two-layer optimization framework is constructed to optimize the prediction interval model.The upper problem employs the grid search method to select the optimal hyperparameters of the biased loss function and determines the optimal combination of base prediction engines.The lower problem utilizes convex optimization to train the corresponding base prediction engines one by one under a given set of hyperparameters.Finally,based on actual photovoltaic data,the prediction performance of the proposed model is verified by season.3)Two-stage robust optimization model of MGs based on adaptive uncertain budget.First,based on the obtained prediction intervals,an interval correction method considering historical corre-lation is proposed to correct the unreasonable forecast in the photovoltaic power prediction interval to improve the sharpness of the interval.Then,based on the comprehensive interval width to estimate the volatility of the forecasting target,an adaptive budget value method considering the correlation of the prediction is proposed to determine the appropriate uncertainty budget value,thereby construct-ing a polyhedral uncertainty set.Finally,considering the connection of micro gas turbines and energy storage,the direct load control mode of demand-side management method is introduced,with the goal of minimizing the economic cost of comprehensive operation,a MG two-stage adaptive robust opti-mal scheduling model is established.The column constraint generation algorithm is used to solve the day-ahead robust scheduling decision of MGs.Specific calculation examples illustrate the rationality and superiority of the proposed model in MG optimal scheduling.