Research On Control Strategy And Capacity Configuration For Hybrid Energy Storage System Based On Wind Power Forecasting

Wind power is one of the fastest-growing forms among new energy power generation technologies. And since wind power has the advantage of mature technology and high development value, it is playing an important role in national energy strategy. Randomness, intermittent and volatility are inherent characteristics of wind power, which can bring about many adverse impacts on the normal operation of the power system. Therefore, it is necessary to conduct in-depth research on measures that can improve the controllability of wind power. Make prediction for wind power and smoothing of its fluctuation using storage equipments are two kinds of effective measures to cope with the uncertainty of wind power. In this paper, the aforementioned measures are combined, which aims to provide reference for operation of the wind farm.Firstly, the basic theory of SVM is expounded, and wind power forecasting is made for a wind farm of North China region. Comparing the forecasting results with that of BP neural network, it shows that SVM has a better performance on prediction accuracy and robustness, verifying the existing conclusion. Then a classification forecasting algorithm for wind power forecasting is proposed aiming at the saturation characteristic of SVM in large sample environment. The key point of this classification forecasting algorithm is converting large sample set to small sample set by making classification, train out the classification model and after that multiple regression models are built for making prediction. The effectiveness of the proposed method is verified by the numerical example.A new method for determining smoothing target of wind power is put forward. Firstly, use wavelet tools to decompose historical data, then use aforementioned SVM classification forecasting algorithm to make prediction for the low frequency component, and take the prediction results as the smoothing target of wind power. Example shows that, compared with carrying out direct wavelet for the very day, the smoothing target determined by this way is smoother, and the volatility of the corresponding energy storage component is greater.A double-layer control strategy for hybrid energy storage system (HESS) is proposed. The first layer strategy is based on the SOC five-partition model of energy storage system. Monitor the SOC level real-timely and conduct real-time correction to the charge and discharge power according to different principles to maintain SOC in a reasonable range. The second layer strategy is used to make power adjustment when SOC between different types of energy storage devices is imbalanced by exchanging energy, which can achieve the full use of energy storage devices. And the energy exchanging is realizing by introducing power cell. Example proves that the double-layer control strategy has nice performance both on maintaining the SOC level and reducing the capacity of energy storage devices.An optimization model for determining the capacity of HESS is presented. Economical optimum model is built, and in this model, the objective function is the minimum comprehensive cost, which includes acquisition cost, penalty cost and cost for abandoned wind; the double-layer control strategy is the control gist. Besides, allocation factor for convex power is defined to make the best compromise between abandoned wind power and compulsive power that injecting to the power grid. Finally, particle swarm optimization (pso) algorithm is adopted to determine the capacity of HESS.