Research On Wind Power Prediction And Fluctuation Suppression Control Method Based On PSO-Kalman And Adaptive Wavelet Packet Decomposition

As a green and clean energy,wind power has great potential for development.To achieve the goal of carbon neutrality,wind power is the main force.However,due to the randomness,intermittence and instability of wind power,the wind power fluctuates greatly in a short period of time.The large-scale integration of wind power into the power grid will inevitably bring a huge impact on the power system.Because of this,there is a phenomenon of wind abandonment in wind farms,resulting in idle and waste of resources.In order to improve the utilization of wind power and reduce the impact of wind power fluctuations on the power system,this thesis proposes a wind power prediction and fluctuation suppression control method.Firstly,the wind power is predicted,and then a certain capacity of energy storage is configured to compensate the prediction error.Finally,the predicted power is judged according to the grid-connected standard.If the predicted power fluctuation is small,the grid-connected standard can be directly connected.If the predicted power fluctuation is large,the wavelet packet decomposition is used to stabilize the wind power fluctuation until the predicted power meets the grid-connected standard.The following is the specific research content :(1)In order to improve the prediction accuracy of wind power point,the particle swarm optimization algorithm(PSO)is used to improve the Kalman filter prediction algorithm(Kalman).The selection of process noise covariance Q and over-observed noise covariance R easily affects the prediction accuracy of Kalman filter algorithm,resulting in large prediction error.Based on the Kalman filter prediction algorithm,this thesis fixes the observation noise covariance R,uses the Kalman filter prediction mean square error as the fitness function,uses the particle swarm optimization algorithm to optimize the process noise covariance Q,finds a process noise covariance that minimizes the prediction error of the Kalman filter algorithm,and establishes a wind power point prediction model PSO-Kalman based on the optimal process noise covariance.(2)In order to obtain the optimal and most reliable prediction interval,after obtaining the point prediction result,the nonparametric kernel density estimation and optimization method are used to predict the interval.Based on the prediction error,the non-parametric kernel density estimation is used to obtain the error probability density function,and the probability density function is integrated to obtain the cumulative distribution function.A certain degree of confidence is set,and the constraints satisfying the confidence are obtained on the cumulative distribution function.The traditional interval prediction method uses symmetrical quantiles on the cumulative distribution function to obtain the error fluctuation interval and then obtain the prediction interval,but the error fluctuation interval corresponding to the symmetrical quantile is not the shortest error fluctuation interval.Under the constraint conditions set in this thesis,the optimization method is used to solve the shortest error fluctuation interval on the error probability density function,so as to obtain the optimal and most reliable prediction interval.(3)By configuring a certain capacity of energy storage to compensate for the prediction error,the wind power prediction power can be equal to the actual wind power,and the wind farm can also provide the optimal prediction interval to the grid,making the wind power controllable and schedulable.This thesis studies the energy storage configuration of wind farms from two aspects,tracking the prediction interval,compensating the difference between the actual power outside the prediction interval and the upper and lower limits of the prediction interval,and realizing the output prediction interval of wind farms to the grid;track the predicted power,compensate the difference between the actual power and the predicted power,and realize the wind farm to output the predicted power to the grid.(4)In order to reduce the wind power fluctuation and make the wind power meet the grid-connected standard,the adaptive wavelet packet decomposition is used to stabilize the wind power fluctuation.According to the grid-connected standard,the number of decomposition layers is adaptively selected to avoid too few decomposition layers,resulting in the power after decomposition not meeting the grid-connected standard.It also avoids too many decomposition layers,which requires larger capacity energy storage devices and increases energy storage costs.Finally,the hybrid energy storage device is used to stabilize the decomposed sub-high frequency component and high frequency.The battery stabilizes the sub-high frequency component,and the super capacitor stabilizes the high frequency component.In order to prolong the service life of energy storage and prevent overcharge and overdischarge of energy storage,Q-learning is used to control the SOC of energy storage device,so that the SOC of energy storage is maintained in the range of 0.2-0.8.While stabilizing wind power fluctuations,the service life of energy storage is prolonged and the cost of energy storage is reduced.Finally,the effectiveness of the wind power prediction and fluctuation smoothing control method proposed in this thesis is verified by experimental simulation analysis.The method proposed in this thesis has higher prediction accuracy,smaller prediction error,and more reliable prediction interval obtained under this method.Therefore,the energy storage configured is more scientific and reasonable.The wind power processed by the method in this thesis reduces the fluctuation,and the wind power curve is smoother,which reduces the impact of wind power on the power system to a certain extent.Under the control of the method in this thesis,there is no overcharge and over discharge phenomenon,and the SOC of the energy storage is maintained in a safe and stable range,which effectively prolongs the service life of the energy storage.The method proposed in this thesis can help power system planners better understand the future situation of wind power generation,so as to reasonably formulate power supply plans and avoid the shortage or excess of energy supply.It can make wind power controllable and schedulable,reduce the fluctuation of wind power,improve the wind power consumption rate,and improve the economic benefits and operational stability of wind farms.