No abstract

As the interconnected power grid becomes increasingly complicated and external environment is changeable, the difficulty of power grid dispatching is increased. Thus, the forecasting track of power grid operation state based on risk assessment can be used to predict the operation trend of power grid state and provide a reference for real-time operation state and guide power grid dispatching. Based on overall power grid risk index system, the operation trend for a period in the future is predicted to diagnose the state of power grid by using fuzzy inference, fuzzy clustering and AHP. In this paper, Ningxia Power Grid is simulated as an example to describe the forecasting track of its operation state in the next 100 minutes. The results present that the track can be used to analyze the causes, which increase or decrease the risk degree of overall power grid, then major leading factors are specifically analyzed. And the track can be also used as guidance for dispatching operators to take measures. Furthermore, the track is proved to be reasonable.

Aiming at the disadvantages of converting traditional transient stability margin into power system control measures, this paper proposes a new transient stability margin characterization method based on critical cutset transient stability available capacity (TATC). Compared with traditional transient stability margin based on fault clearance time or transient energy function, TATC can directly reflects power system transient stability margin form the view of power which is more conducive for power system planning and operation personnel to grasp system transient stability state, at the same time, is also advantageous for prevention measures and emergency control measures to be developed directly according TATC. Simulation results based on IEEE50 machine 145 bus system show that the proposed TATC can effectively characterize power system transient stability margin.

The idea of conservatism has led to significant progress in offline reinforcement learning (RL) where an agent learns from pre-collected datasets. However, it is still an open question to resolve offline RL in the more practical multi-agent setting as many real-world scenarios involve interaction among multiple agents. Given the recent success of transferring online RL algorithms to the multi-agent setting, one may expect that offline RL algorithms will also transfer to multi-agent settings directly. Surprisingly, when conservatism-based algorithms are applied to the multi-agent setting, the performance degrades significantly with an increasing number of agents. Towards mitigating the degradation, we identify that a key issue that the landscape of the value function can be non-concave and policy gradient improvements are prone to local optima. Multiple agents exacerbate the problem since the suboptimal policy by any agent could lead to uncoordinated global failure. Following this intuition, we propose a simple yet effective method, Offline Multi-Agent RL with Actor Rectification (OMAR), to tackle this critical challenge via an effective combination of first-order policy gradient and zeroth-order optimization methods for the actor to better optimize the conservative value function. Despite the simplicity, OMAR significantly outperforms strong baselines with state-of-the-art performance in multi-agent continuous control benchmarks.

No abstract