Dual-Uncertainty-Aware State Estimation of Electric-Gas Integrated Energy System via CNN-LSTM and Physical Constraint Data Alignment

As a complex system that integrates multiple forms of energy, there are multiple time scale issues with the measurement data between different systems in the Electric-Gas Integrated Energy System (EGIES). Meanwhile, to accurately monitor and predict the current state of the system, state estimation (SE) is becoming increasingly important for EGIES. However, during the SE process, the measurement uncertainty caused by the measurement data and the cognitive uncertainty caused by the model have not been widely studied. This paper establishes a novel robust SE method using MC-Dropout enhanced CNN-LSTM model considering uncertainty. Firstly, a data alignment method combining the internal physical constraint of natural gas pipelines with spline interpolation is developed for effectively addressing the issues of consistency and synchronization of measurement data at different time scales. Secondly, by integrating CNN-LSTM networks with MC-Dropout technique while accounting for the effects of both cognitive uncertainty and measurement uncertainty on SE, this approach significantly enhances the accuracy and robustness of SE. Finally, by constructing probability density functions, interval SE results were obtained at different confidence levels, which intuitively expressed the uncertainty of SE and evaluated the accuracy of point estimation. Numerical experiments were conducted on IEEE 33 node and 14 node gas systems, and the results indicated that the proposed method achieves substantial improvements in SE accuracy for EGIES by accounting for cognitive and measurement uncertainties.