| Nowadays,people’s demand for energy is increasing,and the existing fossil energy on earth is limited and in short supply.Therefore,the research and application of new energy can effectively solve the shortage of traditional fossil energy question.In the power industry,the installed capacity of solar photovoltaic power generation worldwide is increasing year by year due to its low cost,high return,sustainable characteristics.However,problems also follow.Since the subsidies enjoyed by photovoltaic power generation mainly depend on their own power generation,in order to obtain high subsidies,unscrupulous users use certain technical means to make the electricity meter of photovoltaic on-grid meters inflated,so as to defraud subsidies.If it is widely used,it will cause the loss of national fiscal revenue,and it will also bring potential safety hazards to the grid-connected operation of photovoltaic power plants,which will greatly affect the healthy development of the photovoltaic power generation industry.In order to ensure the healthy development of the photovoltaic industry and monitor the illegal fraudulent subsidy behavior of photovoltaics,this paper proposes two Photovoltaic stealing identification methods.Firstly,the photovoltaic data is preprocessed,and the meteorological factors that affect the photovoltaic output are analyzed.Based on this,feature construction and similar day clustering are carried out.The main methods of electricity stealing are introduced,and electricity stealing samples are constructed according to their characteristics.Secondly,in view of the operability and complexity of the electricity stealing identification method,this paper firstly proposes a photovoltaic power theft identification method based on the mean impact value-heuristic forward search.By using BP neural network and MIV calculation,and based on the relevant data in a certain region in May 2018,the calculation example is simulated to obtain the MIV of photovoltaic users,and then according to the heuristic forward search screening results and detection thresholds to determine whether the photovoltaic users steal electricity or not,the identification results confirm the effectiveness of the method.On this basis,the simulation and result judgment of the relevant data in January and August were further carried out,and finally the universality of the proposed identification method was verified.In view of the problems that the method is not time-sensitive,flexible and has artificial errors,a photovoltaic power theft identification method based on long short-term memory neural network quantile regression model(QRLSTM)for photovoltaic output interval prediction is proposed.The comparison of the neural network quantile regression(QRNN)prediction method shows the good prediction performance of the QRLSTM method;then the prediction intervals under different confidence levels in three similar days are combined with different electricity stealing judgment indicators according to the time scale(short-term,medium and long-term),A three-layer photovoltaic power theft screening framework is constructed and the degree of user power theft is qualitatively considered.Based on this,the same users as before are used to conduct electricity stealing analysis,and according to the values of various judgment indicators,users are judged for photovoltaic electricity stealing.Finally,by comparing the recognition results with the MIV-heuristic forward search method,the superiority of the recognition method proposed in this chapter is verified.The thesis adopts a three-layer screening and analysis method for photovoltaic users to establish a relatively complete judgment process and carry out photovoltaic anti-stealing work well.At the same time,the photovoltaic output interval prediction takes into account the output randomness and fluctuation range,and quantitatively describes the uncertainty in the prediction;it can provide a more comprehensive decision-making basis for the power system dispatching department. |
PDF Full Text Request