Power Forecasting Approach And System Of Grid-connected PV Plant

Due to the intermittent fluctuations characteristic of photovoltaic (PV) power generation, the grid-connection of large scale PV plants will bring great challenges and difficulties to power system dispatching and management. How to consume these renewable energies by the greatest degree under the security and stability constraints become one of the research focuses in the area of new energy power system. Power forecasting is one of the key technologies to solve this problem, the research on power forecasting approach and system of grid-connected PV plant has very important academic and applied value.There are significant differences between the map relations fitted through PV power forecast model under different weather statuses, and it is very difficult to forecast the PV power accurately in variational weather conditions only use one single model. Moreover, the corresponding historical data distributions of different weather statuses are imbalance, so the applicability of the single forecast model trained by these imbalance data could not be guaranteed. That means the prediction accuracy could not meet the requirements. For this reason, based on the study of the variation relation between surface and extraterrestrial solar irradiance, the feature parameters reflecting weather status characteristic are extracted from solar irradiance data sequence, pattern recognition model for weather statuses based on support vector machine is constructed, the type label of those historical data missing weather type information could be identified, the integrity and availability of historical data are guaranteed. Classification step-wise forecast approach is put forward after the comparison of different power forecast realization modes. The overall block diagram and specific technology roadmap guiding the realization work of the proposed classification PV power forecast approach are also illustrated in this paper.As the basis of PV power forecasting, the accuracy of solar irradiance forecasting is most important for power forecasting. Although artificial neural network (ANN) based solar irradiance forecast models have good performance, there are still some deficiencies exist in the current ANN forecast models, such as high input dimensions, complex model structures, and its output values also hasn’t been modified reasonably. For this reason, the ANN based solar irradiance forecast models are improved from three aspects including take full use of available data, minimize information redundancy and control the input dimensions. According to the setup parameters of variable scale modification, weight coefficient and reference value generated from historical data and similarity metrics within weather type, time periodicity and neighboring similarity based two dimensions variable scale modification method for solar irradiance forecast values is presented. Simulation results show that the precision of solar irradiance prediction has been improved visibly by the above measures.As the key step of PV power forecasting, output characteristic model of PV power generation has great influence to the accuracy of power forecasting. The implement processes of current characteristic model of PV power generation are very complicated, the parameter optimization is quite difficult, and its applicability is also not very satisfactory. For this reason, the direct effect and indirect effect through other factors to PV power of each meteorological influence factor are analyzed, the running state space is constructed by the selected main meteorological influence factors. The relevant data model reflecting the output characteristic of PV power generation is established using actual operating data. Finally, the power mapping forecast method based on running state weighted distance defined according to the relevance between PV power and influence factors is presented to obtain the forecast value from the relevant data model.At last, considering the requirements of power grid dispatching and power plant optimal operation, referring to the relevant standards and technical specifications of PV power forecasting, a PV power forecast system based on the results of this study is developed and put into application. The actual operation achieved good results and verifies the validity of the approaches and models presented in the paper.