基于概率神经网络的集控式风电场风险评估

Risk assessment of centralized control wind farm based on probabilistic neural network

目前对风电机组的风险评估方法多以对某关键部件的风险评估情况来分析风电机组的整体运行情况,由于各关键部件具有较强的耦合作用,需要综合考虑各部件对风电机组的影响情况。为了更好地对风电机组进行维护检修,文中以风电机组的历史运行数据为基础,对其进行风险状态评估,通过风电场集控中心及其数据模型的建设,对风电机组的运行数据进行采集,提取出反映风电机组关键部件运行状态的特征量,使用发电机组理论风速功率曲线与实际输出功率的缺额来描述其运行状态。以关键部件的特征量为输入,风电机组的风险程度为输出,建立概率神经网络模型,通过实例仿真可以看到模型的预测分类效果较好,应用该方法对风电机组的风险状态能进行较好的评估,并为运维检修提供参考。

Currently, the risk assessment methods of wind turbines are mostly based on a key component to analyze the overall operation of the wind turbine. Due to the strong coupling effect of the key components, the influence of each component on the wind turbine must be considered. In order to maintain and repair the wind turbine better, this paper takes the historical data of the wind turbine as the basis, and conducts the risk assessment. The data model is established to collect the operating data of the wind turbine through the construction of the centralized control center of the wind farm, The characteristic quantity reflecting the running state of the key components of the wind turbine is extracted and the theoretical wind speed power curve and the actual output power vacancy of the generator set are used to describe its running state. The probabilistic neural network model is set up with the input of the characteristic quantity of the key components, and the risk degree of the wind turbine as the output. The prediction classification effect of the model can be seen through the example simulation. This method can be used to evaluate the risk state of the wind turbine well and provide reference for the operation and maintenance.