The existing prediction methods ignore the backward information in the time series, resulting in insufficient feature extraction of gas turbine oil temperature series. In this study, a BiTCN-BiGRU-Attention combined prediction model based on black-winged iris optimization algorithm is proposed. The collected gas turbine data is normalized, and the feature importance analysis is performed to screen out the features that have the most influence on the prediction target. The BKA algorithm is used to optimize the BiTCN-BiGRU-Attention network structure to obtain the optimal parameters and realize the prediction of gas turbine lubricating oil temperature. The effectiveness of the proposed method is verified on a gas turbine data set. Compared with the traditional machine learning algorithm and deep learning algorithm, the proposed method can accelerate the convergence speed, enhance the processing ability of time series data, and improve the accuracy and anti-noise interference ability of the prediction model. The accuracy rate is 97.59%, which provides a practical reference for the development and selection of gas turbine oil temperature prediction model.