Stochastic Dynamic Power Research Articles

Stochastic Dynamic Power Dispatch With Human Knowledge Transfer Using Graph-GAN Assisted Inverse Reinforcement Learning

Stochastic Dynamic Power Dispatch With Human Knowledge Transfer Using Graph-GAN Assisted Inverse Reinforcement Learning

Risk-averse stochastic dynamic power dispatch based on deep reinforcement learning with risk-oriented Graph-Gan sampling

The increasing penetration of renewable energy sources (RES) brings volatile stochasticity, which significantly challenge the optimal dispatch of power systems. This paper aims at developing a cost-effective and robust policy for stochastic dynamic optimization of power systems, which improves the economy as well as avoiding the risk of high costs in some critical scenarios with small probability. However, it is hard for existing risk-neutral methods to incorporate risk measure since most samples are normal. For this regard, a novel risk-averse policy learning approach based on deep reinforcement learning with risk-oriented sampling is proposed. Firstly, a generative adversarial network (GAN) with graph convolutional neural network (GCN) is proposed to learn from historical data and achieve risk-oriented sampling. Specifically, system state is modelled as graph data and GCN is employed to capture the underlying correlation of the uncertainty corresponding to the system topology. Risk knowledge is the embedded to encourage more critical scenarios are sampled while aligning with historical data distributions. Secondly, a modified deep reinforcement learning (DRL) with risk-measure under soft actor critic framework is proposed to learn the optimal dispatch policy from sampling data. Compared with the traditional deep reinforcement learning which is risk-neutral, the proposed method is more robust and adaptable to uncertainties. Comparative simulations verify the effectiveness of the proposed method.

Stochastic DVS-based dynamic power management for soft real-time systems

This paper introduces a stochastic dynamic power management policy for soft real-time systems. Such a system comprises a single processor with the capability of dynamic voltage scaling (DVS). The policy uses DVS to consume less power in the processor while satisfying some performance constraints. The idea is based on a Markovian model of the system, which presents an analytical technique for tuning the system parameters and evaluating the effectiveness of the policy. Real-time jobs arrive according to a Poisson process and have exponentially distributed service times and relative deadlines. The power management policy is designed to reduce the long-run power consumption of the system while guaranteeing an upper bound on the fraction of jobs missing their deadlines. It is also shown that the method can be applied to non-Markovian systems using a hybrid analytical-simulation method. Moreover, the method can improve the tradeoff between power and performance when jitters are introduced in the traffic intensity or on the average service time. To satisfy the desired performance constraints of the system, a number of population-dependent speeds are required. Such speeds are extracted from the model of an ideal system with continuous speeds and no speed switching overhead. Since the speed levels in real processors are not continuous, each required speed is approximated in a probabilistic manner by switching between its two immediate neighbors. The time and energy overheads of the switchings are also taken into account. The analytical and simulation results illustrate that the long-run power consumption of the resulting system is close to that of the ideal system, while the performance constraints are completely satisfied.