| In response to the national "double carbon" goal,in recent years,the proportion of new energy connected to the power system has increased,and the proportion of power provided by traditional units has decreased.However,the reactive power and voltage regulation demand of the receiving power grid is still provided by the traditional power supply,resulting in the rapid decline of the reactive power and voltage regulation capacity of the receiving power grid.The main contradiction of the safe and stable operation of the power grid has changed from the problem of power angle stability to the problem of voltage stability.The above changes in the dual characteristics of source load have brought significant changes and potential safety hazards to the operation mode,system characteristics and reactive power and voltage control of power grid.At the same time,the volatility,intermittence and randomness of new energy equipment have a certain impact on the security and stability of power system.The traditional power flow calculation frequency may not meet the regulation speed of power system in the future.How to quickly calculate the residual voltage stability margin of each node of power system,and coordinate the regional reactive power resources for voltage control on this basis is one of the main research hotspots at present.Firstly,this paper introduces the new energy modeling method in traditional power system simulation and the significance of the application of artificial intelligence in power system.Combined with the research status and the development trend of power system in the future,this paper expounds the possible impact of a large number of new energy connected to the power grid,then describes the advantages of the application of artificial intelligence in power system,and determines the feasibility of voltage stability margin calculation through data-driven deep learning.Combined with the traditional standard probability distribution model of new energy applied in the traditional power system simulation,the measured data of new energy in Huai’an area,the research object of this subject,are preprocessed,the data are improved by means of multiple data cleaning and data interpolation,and a comprehensive probability model is improved on the basis of the traditional model.The fit between the comprehensive probability model and the real output data is verified by simulation,While retaining the data characteristics of new energy output,it can meet the needs of in-depth learning and a large amount of data training.According to the operation section of power system,the temporal and spatial characteristics of new energy output are studied.Based on the traditional copula and combined with the comprehensive probability model,a copula model based on photovoltaic and wind power is improved.Through Monte Carlo sampling,combined with the joint probability distribution of new energy equipment in Huai’an area,collect section data to generate deep learning training data set.Thus,the section data containing high-dimensional nonlinear mapping relationship is generated,which provides the data feature basis for the training of deep learning neural network.Aiming at the problem that the traditional data-driven deep learning neural network has poor training effect due to the lack of data features.Firstly,the optimal power flow is combined with the new energy section data to calculate the power flow,so as to generate a large number of power flow section data.Using hybrid neural network,the new energy coupling matrix is transformed into eigenvector through convolution neural network and incorporated into the training data set for data enhancement.Then,combined with the actual calculation example in Huai’an area,it is analyzed.Through the example results,the multi node voltage stability margin index of section is analyzed,which further determines the correctness and effectiveness of the research method in this paper. |
PDF Full Text Request