The excessive temperature of STATCOM thyristor valves will lead to their cracking and even breakdown. Therefore, timely and accurate prediction of the temperature is essential to improve the safety and reliability of STATCOM thyristor valves operation. Based on the least square twin support vector regression (LSTSVR) and seven input values such as STATCOM water temperature, return water temperature, water supply flow, thermal conductivity of heat dissipating material of IGBT module, STATCOM output voltage, STATCOM output current, and collector current of gate valve set, a monitoring system is constructed to predict the temperature of STATCOM thyristor valves. The comparison with the field measured data shows that the LSTSVR model achieves the accurate prediction of STATCOM thyristor valves temperature, and the prediction accuracy of the model is better than the least square support vector regression (LSSVR) model. The effectiveness and reliability of the method are proved by practical cases.