Power grid real-time adaptive decision-making method based on deep reinforcement learning

Abstract

The invention relates to a power grid real-time adaptive decision-making method based on deep reinforcement learning. The method comprises the following steps: step 1, modeling a novel power system unit adaptive scheduling problem as a Markov decision-making process; 2, researching the basic principle of an SAC algorithm, and solving a strategy enabling the accumulated reward value of the MDP model in the step 1 to be maximum; step 3, designing a neural network pre-training scheme based on behavior cloning in the IL, simulating expert experience, optimizing an original action space, proposing an IL-SAC algorithm, training a corresponding power grid optimization scheduling agent based on the IL-SAC algorithm and 105 real power grid scene data, and performing optimization scheduling on the power grid. During testing, the intelligent agent can output a real-time decision scheme in response to different power grid scene data, and intelligent regulation and control of a novel power grid system are realized. The power grid dispatching strategy can be output in real time.

Description

technical field [0001] The invention belongs to the technical field of deep reinforcement learning, and relates to a real-time self-adaptive decision-making method for a power grid, in particular to a real-time self-adaptive decision-making method for a power grid based on deep reinforcement learning. Background technique [0002] With the continuous development of the social economy and the continuous construction of industrial modernization, the demand for energy continues to grow, and energy problems have gradually emerged. To meet the needs of social development, my country's new energy industry has developed very rapidly. Behind the rapid development of new energy, there are It is the problem of overproduction of new energy. In the development of new energy such as hydropower and wind power, these problems are more serious and prominent. Building a smart grid operation mode that maximizes the consumption of new energy is becoming a complex task beyond human expertise. T...

AI Technical Summary

Problems solved by technology

[0005] (1) The existing technology does not fully consider the robust operation of the new power grid when there is a sudden failure under the random fluctuation of new energy and the consumption of new energy when the proportion of new energy is high;

[0006] (2) Existing technologies cannot solve the problem of explosive action space in the decision-making process of adaptive unit scheduling based on deep reinforcement learning for new power grids and the vulnerability of power grid systems in the process of exploring optimization strategies

[0007] After searching, no documents of the prior art identical or similar to the present invention were found

Images