Search

In the presence of increasing uncertainties brought by intermittent renewable energy sources, security and risk management continue to be the most critical concern in modern power system operation. Risk-aware, real-time security constrained economic dispatch (RT-SCED) provides an efficient solution towards promptly, economically and robustly responding to the changes in the power system operating state. Despite different model-based methods have been developed to handle uncertainties, significant computation burden arise to incorporate $N$-1 contingency constraints with a higher temporal resolution in RT-SCED. Driven by similar computational challenges, risk evaluation is often overlooked in the current application of deep reinforcement learning (DRL) based data-driven methods in RT-SCED. This article proposes a DRL-based risk-aware RT-SCED methodological framework by incorporating a novel data-driven risk evaluation model to foster efficient agent-environment interactions. The real-time dispatch policies are constructed with an improved twin delayed deep deterministic policy gradient method. The policy network features a residual network architecture and incorporates an active power allocation mechanism to integrate empirical dispatch knowledge, preventing early termination and fostering more efficient learning behavior. Case studies validate the superior performance of the proposed method in risk-aware RT-SCED on cost efficiency, uncertainty adaptability and computational efficiency, through benchmarking against model-based and data-driven baseline methods.

A growing integration of electric vehicles (EV) and fast charging stations (FCS) has paved the way for the emergence of transportation and power distribution networks. However, previous work on the coordination of EV in the transportation system mainly focuses on static traffic assignment (STA) models which are unable to characterize the temporal evolution of EV flows on traffic links. Meanwhile, limited prior effort in dynamic traffic assignment (DTA) employs a static FCS model which is unable to describe the temporal evolution of charging and discharging (C&D) behavior of EV flows, while respecting the state-of-charge (SoC) related operating constraints, hindering full exploitation of EV flexibility potentials. To bridge the knowledge gap, this paper proposes a novel DTA model to optimize the spatial-temporal distribution of EV flows on roads and at FCSs, taking into account their flexible C&D options and SoC-related constraints, through the solution of only a single linear program. Case studies validate the effectiveness of the proposed DTA model by benchmarking its performance against traditional STA models, and corroborate the core benefits from the proposed spatial-temporal coordination of EV C&D demand in the power distribution network in terms of peak demand reduction, improved RES absorption, reduced carbon emission as well as more efficient network congestion management.

The rapidly increasing penetration of power electronic equipment has complicated the dynamic characteristics of modern power systems, which has promoted the development of simulation analysis methods for online dynamic security analysis (DSA) systems. Accurate models lay the foundation for online DSA, while the high complexity and low observability of power electronic equipment models hinder the online implementation of conventional model identification approaches. In this paper, we propose an online parameter identification (PI) framework based on deep reinforcement learning (DRL), suitable for power electronic equipment models. The DRL-based online PI formulation is proposed by integrating a model-based online PI and an improved DRL-based PI approach. On this basis, the framework for DRL-based online PI is constructed through tailored design of interfaces and training mechanisms. The DRL agent is able to learn the PI policy offline by constructing a training scenario database and the trained policy can be deployed to realize accurate online identification. Case studies showcase the application of the proposed method on a high-voltage direct current (HVDC) system. Simulation results demonstrate that the proposed method exhibits superior computational advantage, while achieving a similar level of accuracy compared to conventional methods.