隨著社會的發展，用戶對供電可靠性要求不斷提高，電氣設備長時間重負荷或超負荷運行，以及負荷的高電壓瞬間變化等都會導致高壓設備元件接觸部位發生過熱異常故障等狀況，為了預防電器設備的故障情形，通常會運用手持式熱像儀定期巡檢，或裝設定置式熱像儀進行全時監控，依據熱影像所得溫度來判斷該電氣設備是否具有異常過熱、故障現象，但在分析的當下往往無法發現即早期的異常或故障時已經有較嚴重的損害，以至於來不及做出對策，因而發生非預期性之故障事件。

為此，本論文將以案場收集的熱影像等相關數據進行研究，提出一套溫度預測模型，運用歷史資料進行模型的訓練，將完成的模型帶入即時數據，透過模型計算輸出未來時段的溫度，達成預測效果，再搭配設備溫度異常檢測判斷，達到預先診斷之作用，並將研究結果進行整合，透過網頁的前後端串接，把數據資訊與操作界面化，研製成為設備溫度預測系統，為現場監控、維護人員提供更多的早期指標資訊判斷設備是否異常、故障，爭取更長的維護時間，提高電氣系統可靠性。
c

With the development of society, users' requirements for power supply reliability continue to increase. Long-term heavy load or overload operation of electrical equipment and instantaneous changes in high voltage of the load will cause overheating and abnormal failures in the contact parts of high-voltage equipment components. For the failure of electrical equipment, a handheld thermal imager is usually used for regular inspection, or a set thermal imager is installed for full-time monitoring. According to the temperature obtained by the thermal image, the electrical equipment is judged to be abnormally overheated or faulty. At the moment of analysis, it is often impossible to find that the early abnormality or failure has already caused more serious. it is too late to make a countermeasure, so an unexpected failure event occurs.
To this end, this paper will research the thermal image and other related data collected on the scene propose a set of temperature prediction models. Use historical data for model training, bring the completed model into real-time data, and output the future period time through the model calculation Temperature. To achieve the prediction effect, combined with equipment temperature abnormal detection and judgment. To achieve the role of pre-diagnosis, and integrate the research results, develop an equipment temperature prediction system, through the web page to show data information and the operation interface. it provides on-site monitoring and maintenance personnel with early indicator information to determine whether the equipment is abnormal or faulty, strives for longer maintenance time, and improves the reliability of the electrical system.

[ 1 ]
L. Zhu-Mao, L. Qing, J. Tao, L. Yong-Xin, H. Yu and B. Yang, "Research on Thermal Fault Detection Technology of Power Equipment based on Infrared Image Analysis," 2018 IEEE 3rd Advanced Information Technology, Electronic and Automation Control Conference (IAEAC), Chongqing, 2018, pp. 2567-2571.
[ 2 ]
A. Dragomir, M. Adam, M. Andruçcâ, A. Munteanu and E. Boghiu, "Considerations regarding infrared thermal stresses monitoring of electrical equipment," 2017 International Conference on Electromechanical and Power Systems (SIELMEN), Iasi, 2017, pp. 100-103.
[ 3 ]
A. Dragomir, M. Adam, M. Andruşcă, A. Munteanu, "Long Term Thermal Stresses of a Withdrawable Electrical Contact", The 6-th International Conference on Modern Power Systems - MPS Mai 18 - 21, 2015.
[ 4 ]
A. Dragomir, M. Adam, M. Andruşcă, M. Molodeschi and R. Pantelimon, "About thermal stresses monitoring and diagnosis of electrical equipment," 2014 International Conference and Exposition on Electrical and Power Engineering (EPE), Iasi, 2014, pp. 289-294.
[ 5 ]
N. Xing, J. Zhang, R. Yu and J. Yao, "An Early Warning Analysis Technique of Electric Equipment Temperature Based on Association Rule Mining," 2018 IEEE International Conference on Energy Internet (ICEI), Beijing, 2018, pp. 258-263.
[ 6 ]
X. Kong, H. Zang, "Temperature Forecast for Substation Equipment Based on Optimized General Regression Neural Network", Electric Power, vol. 49, no. 07, pp. 54-59, 2016.
[ 7 ]
Q. Li, M. Musavi and D. Chamberlain, "Overhead conductor thermal rating using neural networks," 2011 IEEE International Conference on Smart Measurements of Future Grids (SMFG) Proceedings, Bologna, 2011, pp. 139-142.
[ 8 ]
J. Guo, Z. Zhang, W. Gao, H. Hu, D. Wang and Y. Mao, "Overheating Risk Warning Model based on Thermal Circuit Model and Load Forecasting for Distribution Transformers," 2019 IEEE Sustainable Power and Energy Conference (iSPEC), Beijing, China, 2019, pp. 2891-2895.
[ 9 ]
Michael A. Nielsen, Neural Networks and Deep Learning, Determination Press, (2015)
[ 10 ]
Deep RAN: A Scalable Data-driven platform to Detect Anomalies in Live Cellular Network Using Recurrent Convolutional Neural Network