高壓模鑄式變壓器為許多企業供電的重要角色，然而在長期的使用下因環境的雜質、本身先天瑕疵或人員的操作不當等劣勢所累積，最後造成它的絕緣劣化，雖然在傳統的定期性維修保養(Predetermined Maintenance)有一定降低故障發生的機率，但定期性維修保養可能對設備進行超量的維修保養，造成人力物力資源的大量浪費，且無法應對突發的狀況，當突發狀況發生，將會造成無法彌補的損失，甚至是人員的工安意外。
為此，本論文將針對高壓模鑄式變壓器從定期性維修保養升級至狀態性維修保養(Condition-based Maintenance)，其中監測診斷設備狀態最為直接、最為透明的方式就是測量其溫度，從溫度的變化進而得知設備狀態的變化，透過安裝定置式熱像儀獲取設備區域性溫度，也透過可得知的電氣訊號輔助診斷，做長期的資料收集，進而對收集到的資料做數據分析。收集大量數據後，再進一步做人工智能(Artificial Intelligence)故障診斷，先採用卷積神經網路判斷是否故障，及故障類型，最後使用影像分割神經網路，將故障位置標示出來，結果指出，建置熱影像監控系統後，在事故發生前，能有效地提前給予告警，通知人員提早地做設備維修保養，大大地降低工安意外的風險以及提高電力的可靠度。

High-voltage cast resin transformers play an important role in power supply for many enterprises. However, under long-term usage, high-voltage equipment has been affected with environment, inherent defects, or improper operation of personnel, which ultimately leads to insulation aging. Although there is an effective reduction in the probability of failure with the traditional predetermined maintenance, it may maintain the equipment excessively, resulting in a large waste of human and material resources. It is also unable to cope with unexpected situations. When an unexpected situation occurs, it will cause serious losses, and even the accident of personnel safety.
To this end, this thesis will upgrade from predetermined maintenance to condition-based maintenance against high-voltage cast resin transformers. The most direct and transparent way to monitor diagnose the state of equipment is to measure its temperature. You can know the state from the change of temperature. Obtain the areal temperature by installing fixed thermal cameras and obtain the available electrical signal to assist in diagnosis for long-term data collection. And then, we analyse collected data. After having enough data, we can further do fault diagnosis by artificial intelligence. First, Decide whether there is a failure or not, and classify the type of failure. Finally, mark the fault location by segmentation network. The results indicate that establishment of thermal image monitoring system can effectively give alarms in advance before the accident. It can notify personnel to maintain the equipment early, and it will greatly reduce the risk of industrial safety accidents and improve the reliability of power supply.

[1] Z. Dong and L. Chen, "Image registration in PCB fault detection based on infrared thermal imaging," 2019 Chinese Control Conference (CCC), Guangzhou, China, 2019, pp. 4819-4823.
[2] T. Dutta, J. Sil and P. Chottopadhyay, "Condition monitoring of electrical equipment using thermal image processing," 2016 IEEE First International Conference on Control, Measurement and Instrumentation (CMI), Kolkata, 2016, pp. 311-315.
[3] O. Janssens, M. Loccufier and S. Van Hoecke, "Thermal imaging and vibration-based multisensor fault detection for rotating machinery," in IEEE Transactions on Industrial Informatics, vol. 15, no. 1, pp. 434-444, Jan. 2019.
[4] FLIR, "FLIR camera adjustments boson," FLIR Application Note, September 2018.
[5] A. Paul, P. Biswas, P. Bhattacharya and S. P. Maity, "Power-log based plateau limit histogram equalization algorithm for infrared image enhancement," 2017 Tenth International Conference on Contemporary Computing (IC3), Noida, 2017, pp. 1-5.
[6] L. Sun, J. Wang and S. Li, "A new infrared image enhancement algorithm," 2017 29th Chinese Control And Decision Conference (CCDC), Chongqing, 2017, pp. 421-424.
[7] S. Kumar, L. Hussain, S. Banarjee and M. Reza, "Energy load forecasting using deep learning approach-LSTM and GRU in spark cluster," 2018 Fifth International Conference on Emerging Applications of Information Technology (EAIT), Kolkata, 2018, pp. 1-4.
 
[8] Z. Hui and H. Fuzhen, "An intelligent fault diagnosis method for electrical equipment using infrared images," 2015 34th Chinese Control Conference (CCC), Hangzhou, 2015, pp. 6372-6376.
[9] G. Mirzaei, M. M. Jamali, P. V. Gorsevski, J. Frizado and V. P. Bingman, "Regularized logistic regression based classification for infrared images," 2014 48th Asilomar Conference on Signals, Systems and Computers, Pacific Grove, CA, 2014, pp. 487-490.
[10] B. R. Mete and T. Ensari, "Flower classification with deep CNN and machine learning algorithms," 2019 3rd International Symposium on Multidisciplinary Studies and Innovative Technologies (ISMSIT), Ankara, Turkey, 2019, pp. 1-5.
[11] A. T. Vo, H. S. Tran and T. H. Le, "Advertisement image classification using convolutional neural network," 2017 9th International Conference on Knowledge and Systems Engineering (KSE), Hue, 2017, pp. 197-202.
[12] S. Kido, Y. Hirano and N. Hashimoto, "Detection and classification of lung abnormalities by use of convolutional neural network (CNN) and regions with CNN features (R-CNN)," 2018 International Workshop on Advanced Image Technology (IWAIT), Chiang Mai, 2018, pp. 1-4.
[13] T. J. Ramírez-Rozo, J. C. García-Álvarez and C. G. Castellanos-Domínguez, "Infrared thermal image segmentation using expectation-maximization-based clustering," 2012 XVII Symposium of Image, Signal Processing, and Artificial Vision (STSIVA), Antioquia, 2012, pp. 223-226.