Day-ahead and look-ahead scheduling of gas-power systems through a decentralized data-driven method

The utilization of gas-fired units to support the development of wind power generation intensifies the interactions between gas and power systems. However, challenges exist in the scheduling of these systems. The day-ahead scheduling framework restricts the scheduling horizon of gas-power systems to one day and cannot manage the final energy level of storage systems. In addition, limited access to the information of each system due to data privacy makes centralized scheduling of these systems impossible in practice. This paper presents a decentralized data-driven model for day-ahead and look-ahead scheduling of the gas-power system integrated with storage systems. The proposed model extends the scheduling horizon from one day to two consecutive days with the aim of i) minimizing the operation costs of gas-power systems and ii) optimizing the energy level of storage systems at the last hour of the first day. The decentralized structure is designed based on the alternating optimization procedure to preserve the data privacy of each system. Moreover, data-driven Wasserstein-based distributionally robust optimization is used to cover the uncertainty of wind. The competency of the introduced decentralized data-driven model is analyzed on the 6-node/6-bus gas-power and the 10-node/24-bus gas-power networks. Results demonstrate that if the look-ahead scheduling framework is considered along with day-ahead scheduling, the total day-ahead operation cost of gas-power systems decreases by 1.48 k$ for the small-scale system and 2.77 k$ for the large-scale system.