基于MIC-LSTM与CKDE的风电机组机舱温度区间预测

Wind Turbine Nacelle Temperature Interval Prediction Based on MIC-LSTM and CKDE

风电安全技术的发展在新能源生产安全中具有重要意义。风力发电机组机舱温度预测可提前发现机舱温度的异常变化,为监测和控制系统提供温度预警信号,从而保障内部设备安全稳定运行。提出基于最大信息系数(MIC)的变量筛选方法,选取与机舱温度相关性较高的变量作为输入变量,然后基于长短时记忆(LSTM)网络建立了多变量机舱温度单点预测模型,通过与其它3类预测模型的性能对比表明了所提方法精度更高;基于LSTM网络模型的预测结果及其误差数据集,采用条件核密度估计(CKDE)法建立了不同置信度下机舱温度预测值的波动区间,依据具体实例验证了不确定性区间预测模型的有效性和可靠性。

The development of wind power safety technology is of great significance in the safety of renewable energy production. The wind turbine nacelle temperature prediction can find abnormal changes in advance,provide temperature warning signals for the monitoring and control system,and thus ensure the safe and stable operation of internal equipment. A variable selecting method based on maximum information coefficient(MIC)is proposed,and variables with high correlation with nacelle temperature are extracted as input variables,and then single point prediction model for multivariable nacelle temperature based on long short-term memory(LSTM)network shows that the proposed method has higher accuracy by comparing the performance with other three types of prediction models.Conditional kernel density estimation(CKDE) method establishes the fluctuation interval of the predicted value of the nacelle temperature under different confidence degrees based on the prediction results of the LSTM network and the generated error data set,and the validity and reliability of uncertainty interval prediction model are verified according to specific examples.