Research On Power Output Forecasting Of Photovoltaic Generation System In Microgrid

As the capacity of photovoltaic power generation system micro network is increasing,in alleviating energy crisis and reduce the pollution of the environment at the same time, the characteristics of stochastic volatility and discontinuity of the photovoltaic output also the safety of power grid and the quality of economic growth has brought serious challenges.To accurately forecast of photovoltaic power, power supply departments help timely adjust power supply plan, coordination of routine power and photovoltaic power generation as a whole, improve the ability to the safe and stable operation.Therefore, to explore the influence factors of photovoltaic power and accuracy, the connection between the proper prediction model is established, and selection and to predict the high similarity of samples training, can improve the photovoltaic output prediction accuracy.To this end, this paper mainly do the following work:First, in actual operation of photovoltaic power generation system as the research object, according to the monitoring system to record the history of the output and the corresponding meteorological information sampling, in order to analyze the various factors influencing the photovoltaic output. Factor is calculated using Pearson correlation coefficient method and correlation degree between the photovoltaic(pv) power, which can be considered various factors on the output capacity of photovoltaic(pv). Combined with photovoltaic output values and the corresponding weather information and determine the type of weather index of each sampling date.Selection and pv output related degree bigger factors of meteorological feature vector, the weather type index consistent data collection, using the method of weighted Euclidean distance, selection and every 1 h to predict daily weather properties close to sample, the resulting sample can reflect to forecast each time the weather situation, training samples with high similarity forecast model prediction accuracy can be improved.Secondly, based on the analysis of the existing prediction algorithm, on the basis of characteristics and characteristics of the training sample data was proposed based on improved particle swarm optimization and least squares support vector machine(SVM) photovoltaic output short-term forecasting method.In view of the traditional least squares support vector pause to take parameters depend on the experience in trial efficiency low faults, introduces the particle swarm optimization algorithm, particle swarm optimization algorithm based on particle swarm of parallel search feature to search for the optimal objective function value, thus giving parameters optimal values.In order to overcome the premature convergence phenomenon in a standard particle swarm optimization algorithm, an improved measures are put forward, according to the particle swarm information selecting initial population and premature convergence, the species were distributed evenly over the whole solution space, when calculating the late particles easily plunged into local optimum, update the particle’s position, jump out of local optimal area lead particles, particles in the population, to enhance the global optimization ability.Finally, selecting different weather type index in different season 4 sets of data for this article proposed pv output short-term prediction model for testing and assessment results show that the proposed model prediction accuracy is higher, to the dispatching department to make reasonable scheduling plan to provide certain reference basis.