Research On Generating Capacity Prediction Of The Photovoltaic System

Solar energy receives more and more attention because of its cleanliness,non-polluting and renewable, and photovoltaic power generation has become the hotspotin the world’s renewable energy sector. However, photovoltaic power possessescharacteristics of uncertainty and intermittency, the large-scale photovoltaic gridincreases the difficulty of grid scheduling, which also affects the security, stability andeconomic operation of power systems. Accurate prediction of photovoltaic power ispremise of effectively reducing the adverse effects for power systems with large-scalephotovoltaic, which also has important guiding significance on grid scheduling program,conventional energy planning, photovoltaic power generation planning and so on.In this paper, taking the photovoltaic power generation as a research object, byanalyzing the impact factors of photovoltaic power, research on generating capacityprediction of the photovoltaic power is carried out. Firstly, the effects of season types,weather types and meteorological factors such as radiation intensity, ambient temperature,and cloudiness on photovoltaic power are analyzed, then the input variables of predictionmodel are determined, and the method that making use of similarity theory to determinetraining samples is proposed. Secondly, by analyzing the advantages and disadvantagesof the traditional BP algorithm, the improved BP algorithm based on momentum,steepness factor and adaptive learning rate is proposed, and the appropriate predictionmodel of photovoltaic power is established. Then, against the deficiency of improved BPalgorithm and PSO algorithm, chaos search and adaptive mutation are introduced intoPSO in order to improve the algorithm’s global convergence probability and speed, thenprediction model based on chaos search and AMPSO-BPNN is established, and a methodthat using the power of similar days to correct predictive value is proposed. Finally,instance validation of the prediction models based on actual weather stations and PVpower plants is done, and by analyzing the correlation between PV power output andload, the research direction of photovoltaic power station operation is further clarified.The software of photovoltaic system power forecasting is developed in MicrosoftVisual C++6.0environment, the prediction effect of different photovoltaic powerforecasting models is verified, and the forecast results show that the proposed model andalgorithm has higher forecasting accuracy and convergence rate, and the correctionmethod based on power of similar days is feasible.