基于MWSA的热力系统单参数时序预测方法

Sequential Prediction Method of Single Parameter for Thermal System Based on MWSA

热力系统的状态参数变化可以实时反映系统的运行状态,针对热力系统参数运行数据预测手段匮乏的现状,基于4种算法提出一种单参数预测方法并简称MWSA,对当前设备状态参数进行分解降噪、趋势提取和时序预测,并将预测结果作为下一步运行管理策略和装备维修的参考,对系统的长期安全稳定运行具有重要意义.首先,利用中值回归经验模态分解(MREMD)方法将监测得到的运行状态参数分解为若干个本征模态函数(IMF)和残余分量.然后,对不符合筛选条件的分量进行小波阈值降噪(WTD),并将去噪后的分量与原本符合筛选条件的分量重组成新的IMF分量.最后,利用基于奇异值分解(SVD)和优化参数排列熵(PE)的K-means聚类算法,对重组后的IMF分量进行分类,取熵值较低的一类分量重构为趋势项并采用整合滑动平均自回归模型(ARIMA)进行预测.经实际案例验证,该方法能够有效克服原始参数时序中高频噪声的干扰,与不采用降噪处理的同类方法相比,该方法预测的准确度更高.

The changes in the status parameters of the thermal system reflect the operating status of the system in real time.The forecast results of the trend extraction and time series prediction of the current equipment status parameters can be used as a reference for the next operation management strategy and equipment maintenance,which can be used for the long-term system safe and stable operation.In this paper,a method which is described as MWSA,based on the midpoint and regression based empirical mode decomposition(MREMD),the wavelet threshold denoising(WTD)and midpoint and techniques,the singular value decomposition(SVD)and optimized parameter permutation entropy(PE),and an auto regressive integrated moving average model(ARIMA),is applied to the single-parameter time series prediction of thermal systems.First,the MREMD is used to decompose the monitored operating state parameters into a number of intrinsic mode functions(IMF)and residual components.Next,the components that do not meet the screening conditions are subjected to wavelet thresholding.The denoised components and the components that originally meet the filtering conditions are recomposed into new IMF components.Finally,the K-means clustering algorithm based on SVD and PE is used to classify the recomposed IMF components,the component with a lower entropy value is selected and reconstructed into a trend item,and ARIMA is used to predict.An actual case verifies that this method can effectively overcome the interference of high-frequency noise in the original parameter timing,and the prediction accuracy is higher than that of similar methods without noise reduction treatment.