煤矿井下环境特殊，各种安全隐患严重地威胁着工作人员，其中设备故障是隐患之一。煤矿井下设备运行状况下，当存在故障隐患时，经常会伴随着温度的上升，当设备温度过高时有可能会引发火灾和爆炸等事故，对这类设备运行过程中的温度进行监测记录非常必要。针对检测到的设备温度数据进行分析研究，可以发现其内在规律，进而预测温度未来的变化趋势，实现对设备的安全预警，从而有效地避免因设备故障带来的损失。 　　　论文以基于无线传感器网络的煤矿井下设备温度监测系统为基础，重点对设备温度时间序列进行分析建模和预测，并对预测结果进行预警分析。通过对几种不同的温度预测方法进行研究并对其特点进行分析比较，得出在不同情况下，应根据数据样本的特点采用不同的预测方法。在大样本情况下，可以采用人工神经网络的方法进行预测，结合经验模态分解和相空间重构，以期得到较准确的预测结果；在小样本数据情况下，优先采用ARIMA进行预测模型的建立，对时间序列进行准确的短期预测。为了对设备温度安全警度进行预警，本文利用DS证据理论的时域融合，采用了基于经验模态分解、BP神经网络和DS证据理论的预警方法，实验结果验证了该方法的有效性。 　　　设计了设备温度预警系统软件，预警系统的主要功能包括设备温度曲线的绘制、温度预测及预警等。利用Visual C++进行设备温度预警系统的界面设计，通过采用Visual C++与Matlab混合编程来实现软件的设计，该软件能实现预期功能。

Coal mine environment is special and a variety of security risks threaten the staff seriously. Equipment failure is one of the risks. When there is some hidden fault during the running of coal mine equipment, equipment temperature usually rises. It is possible to cause fires and explosions when the equipment temperature is too high. So it is very necessary to monitor and record the equipment temperature during operation. Through analyzing the equipment temperatures detected the inherent laws can be found, which can help to predict the future trend of the temperature change and realize equipment safety warning. Thus losses caused by equipment failure can be avoided effectively. 　　　In the paper, the basis is the coal mine equipment temperature monitoring system based on wireless sensor networks, and the focus is the equipment temperature time series modeling, forecasting and the prewarning analysis of the forecasting results. Through the research of several different temperature forecasting methods and the comparison of their characteristics, the conclusion that under different circumstances different forecasting methods should be adopted according to the characteristics of the data sample is made. In the case of large samples, the artificial neural network method can be used to predict, combined with empirical mode decomposition and phase space reconstruction to obtain more accurate predictions; in the case of small samples, ARIMA should be used preferentially to establish forecasting model to make short-term prediction accurately for time series. In order to make early warning for equipment temperature safety degree, time domain integration of DS evidence theory is adopted and the warning method based on empirical mode decomposition, BP neural network and DS evidence theory is used. The results show that the prewarning method is effective. 　　　Equipment temperature warning system software is designed and the main functions of the prewarning system are plotting temperature curve, making temperature prediction and warning, etc. Visual C++ is used to design the equipment temperature warning system interface. By hybrid programming with Visual C++ and Matlab, the prewarning software design is finished and the software can realize expected functions.