Research On Uncertainties Modeling Based On Wind Power Forecasting Error

Wind energy is deemed as the most competitive renewable energy. As the core of renewable energy generation, wind power has obtained a rapid development. Nevertheless, wind power is the uncertainty of energy. The large-scale integration of wind farms with power grid would make it harder to protect the safe and stable operation of power system caused by stochastic and intermittent wind, it also may complicate the problem of wind power consumption and traditional optimization scheduling. What would become a hot research area in wind power integration and optimization scheduling is how to describe the uncertainties of wind power and apply to complex power system with stochastic coupling. In this paper, study on modeling of wind power forecasting error and the application problem of unit commitment of power system with wind power plants.1. Characteristics of wind resources about wind speed and wind directionin actual wind farms were deeply analyzed. Then wind power forecasting model based on NWP and modified BP neural network was studied.1h,3h and 24h-ahead of wind power forecasting was carried on. Simulation results showed that this model’s forecasting precision better than RBF, PSOBP and basic BP neural network. But as the same as the existing forecasting models, there were still error in forecasting results. Therefore study on the uncertainties of forecasting error is necessary.2. The uncertainty of wind power forecasting error based on statistical model was studied. Firstly, on the basis of the large wind power data and distribution scatterplots of 4h,6h, and 24h-ahead of forecasting, the forecasting error distribution model of Gaussian, Cauchy, Laplace and a kind of non-Gaussian model named versatile distribution was analyzed. Among them, versatile model had a comparatively higher fitting accuracy. To improve the applicability of that model, forcasting power segment was optimized in order to divide forcasting power segment automatically following the size of historical data. According to the qualitative and quantitative simulation comparison test, the improved versatile model was able to adapt to the real time changes of the forecasting error model parameters under different wind farms and different timescale of forecasting, which improved the fitting efficiency of the wind power forecasting error.3. The fuzzy model which was modeled for describing the uncertainty of wind power forecasting error was studied on the basis of credibility theory. Membership function of forecasting error was modeled by Cauchy distribution function which could reflect the distribution characteristics of actual forecasting error and credibility measure is obtained by membership function, thus fuzzy model of forecasting error based on credibility measure was established.4. Statistical and fuzzy model were applied to unit commitment scheduling. The probabilistic wind power forecasting results were concluded in spinning reserve for application of statistical model. Spinning reserve constraints with fuzzy variable were processed as crisp equivalents for application of fuzzy model to finish and conduct introduction of uncertainty information. The problem of unit commitment scheduling was calculated by discrete particle swarm optimization (BPSO) algorithm and small-world particle swarm optimization (SW-PSO) algorithm. Finally the example demonstrated that effectiveness of unit commitment model was used to describe uncertainty of wind power forecasting error, and expected that could provide for the decision maker the use value reference for optimal scheduling of wind farm integrated power system.