This paper develops a segmented real-time dispatch model for power-gas integrated systems(PGISs), where power-to-gas(P2G) devices and traditional automatic generation control units are cooperated to manage wind power ...This paper develops a segmented real-time dispatch model for power-gas integrated systems(PGISs), where power-to-gas(P2G) devices and traditional automatic generation control units are cooperated to manage wind power uncertainty. To improve the economics of the real-time dispatch in regard to the current high operation cost of P2Gs, the wind power uncertainty set is divided into several segments, and a segmented linear decision rule is developed, which assigns adjustment tasks differently when wind power uncertainty falls into different segments. Thus, the P2G operation with high costs can be reduced in real-time adjustment. Besides, a novel segmented stochastic robust optimization is proposed to improve the efficiency and robustness of PGIS dispatch under wind power uncertainty, which minimizes the expected cost under the empirical wind power distribution and builds up the security constraints based on the robust optimization. The expected cost is formulated using a Nataf conversion-based multi-point estimate method, and the optimal number of estimate points is determined through sensitivity analysis. Furthermore, a difference-ofconvex optimization with a partial relaxation rule is developed to solve the non-convex dispatch problem in a sequential optimization framework. Numerical simulations in two testing cases validate the effectiveness of the proposed model and solving method.展开更多
We consider qualitatively robust predictive mappings of stochastic environmental models, where protection against outlier data is incorporated. We utilize digital representations of the models and deploy stochastic bi...We consider qualitatively robust predictive mappings of stochastic environmental models, where protection against outlier data is incorporated. We utilize digital representations of the models and deploy stochastic binary neural networks that are pre-trained to produce such mappings. The pre-training is implemented by a back propagating supervised learning algorithm which converges almost surely to the probabilities induced by the environment, under general ergodicity conditions.展开更多
基金supported by the National Natural Science Foundation of China (No. 51907025)Fundamental Research Funds for the Central Universities。
文摘This paper develops a segmented real-time dispatch model for power-gas integrated systems(PGISs), where power-to-gas(P2G) devices and traditional automatic generation control units are cooperated to manage wind power uncertainty. To improve the economics of the real-time dispatch in regard to the current high operation cost of P2Gs, the wind power uncertainty set is divided into several segments, and a segmented linear decision rule is developed, which assigns adjustment tasks differently when wind power uncertainty falls into different segments. Thus, the P2G operation with high costs can be reduced in real-time adjustment. Besides, a novel segmented stochastic robust optimization is proposed to improve the efficiency and robustness of PGIS dispatch under wind power uncertainty, which minimizes the expected cost under the empirical wind power distribution and builds up the security constraints based on the robust optimization. The expected cost is formulated using a Nataf conversion-based multi-point estimate method, and the optimal number of estimate points is determined through sensitivity analysis. Furthermore, a difference-ofconvex optimization with a partial relaxation rule is developed to solve the non-convex dispatch problem in a sequential optimization framework. Numerical simulations in two testing cases validate the effectiveness of the proposed model and solving method.
文摘We consider qualitatively robust predictive mappings of stochastic environmental models, where protection against outlier data is incorporated. We utilize digital representations of the models and deploy stochastic binary neural networks that are pre-trained to produce such mappings. The pre-training is implemented by a back propagating supervised learning algorithm which converges almost surely to the probabilities induced by the environment, under general ergodicity conditions.
