Research On Primary Frequency Modulation Optimization Of Wind Farms With Large Scale Wind Turbines

In recent years,clean energy has developed rapidly.As a kind of clean energy,the installed capacity and power generation of wind turbines are increasing year by year.Large-scale wind turbines connected to the grid pose a challenge to the stability of grid frequency.Therefore,it brings higher technical requirements to the primary frequency modulation performance of wind turbines,that is,the wind farm adjusts the total output power of the wind farm in time according to the frequency deviation,and distributes the power reference to each wind turbine in a certain strategy.However,the wind farm has a wide area,the distribution of wind turbines is dispersed,and the output characteristics of each wind turbines are quite different.In order to solve the problem of optimal power allocation,the two-step coordinated optimal power scheduling allocation method for wind farms is proposed,and a multi-objective optimal power scheduling allocation model is established,which takes the times of start-up and shut-down and the mean of the power limited index as the objective functions.The first step of the two-step coordinated optimization scheduling method is the cluster analysis of the wind turbines.The number of wind turbines is large.If all of them participate in the calculation of power allocation,the computational complexity is high.In order to reduce the computational complexity,the paper uses clustering algorithm to cluster wind turbines based on operational data.According to the classification results,some wind turbines participate in subsequent scheduling calculations.The self-adjusting chaotic particle swarm optimization algorithm(SACPSO)is proposed.The test function proves that the SACPSO algorithm has better performance than particle swarm optimization(PSO).self-adjusting particle swarm optimization(SAPSO)and chaotic particle swarm optimization(CPSO).In order to improve the performance of clustering algorithm,PSO algorithm and SACPSO algorithm are used to optimize the fuzzy C-means clustering algorithm(FCM)to get the particle swarm optimization fuzzy C-means algorithm(PSOFCM)and the self-adjusting chaotic particle swarm optimization fuzzy C-means algorithm(SACPSOFCM).It is proved that the SACPSOFCM algorithm has better clustering performance than the PSOFCM algorithn through the Iris data set.Based on the actual data of wind turbines,PSOFCM algorithm and SACPSOFCM algorithm are used to classify the wind turbines into two types:the benchmarking wind turbines that don't participate in the subsequent scheduling calculation and the wind turbines to be dispatched that participate in the subsequent scheduling calculation.It is proved that the SACPSOFCM algorithm can achieve the correct classification of wind turbines.The second step is to perform power optimization scheduling calculation for the wind turbines to be dispatched.In order to solve the multi-objective scheduling problem of wind turbines,this paper proposes a double-layer particle swarm optimization algorithm(DLPSO)based on binary particle swarm optimization algorithm(BPSO)and SACPSO algorithm and chaotic second-generation non-dominated sorting genetic algorithm(CNSGA-?),a multi-objective power scheduling allocation model is established,which takes the times of start-up and shut-down and the mean of the power limited index as the objective functions.Based on the clustering results obtained by the first step,the wind turbines to be dispatched are substituted into the multi-objective scheduling model to participate in the power scheduling calculation.The results show that compared with the DLPSO algorithm,the CNSGA-? algorithm is more suitable for multi-objective scheduling problems and has better optimization performance.Finally,the two-step coordinated power optimization method based on SACPSOFCM and CNSGA-? algorithm is compared with the traditional power weighted average allocation method,which proves that the two-step coordinated power optimization method has better multi-objective scheduling performance.