Data-driven state estimation of integrated electric-gas energy system

As the energy consumption in the world continues to increase, the integrated energy systems combined with multiple types of energy are gradually developing to enhance the utilization efficiency of each type of energy. However, some issues bring challenges to the state estimation of the coupled electric-gas integrated energy system. The issues include that measurement data of the integrated energy system has low redundancy, there exists a large measurement error of the integrated energy system, and the measurement devices of the electric and gas network are not standardized in terms of sampling time units. Considering that the data-driven model has high portability and the ability to distill and summarize information, the data-driven state estimation model of the electric-gas coupled integrated energy system is established in this paper. Bayesian learning is used to obtain the probabilistic statistical features of the measurement data. Super Latin sampling is applied to generate the complete measurement data. The rationality of the generated data is checked by the energy flow analysis of the integrated energy system to obtain the training sample set for the deep learning network. A hybrid deep learning network coupled with the convolutional neural network and long and short-term memory is proposed, and the root mean square error is utilized to train the hybrid deep learning network, which effectively improves the error accuracy of the state estimation of the electric-gas coupled integrated energy system. Compared with the classical model-driven method of state estimation, the arithmetic simulation verifies the effectiveness of the proposed method.