Reinforcement learning-based micro-grid optimal scheduling method

Abstract

The invention relates to a reinforcement learning-based micro-grid optimal scheduling method, and the method comprises the following steps of obtaining the prediction information of a micro-grid wind-light element and the boundary information of a power supply, and constructing amicro-grid optimal scheduling reinforcement learning model based on reinforcement learning; carrying out self-learning on the micro-grid optimization scheduling reinforcement learning model based on reinforcement learning, and accumulating the scheduling knowledge learned in the self-learning process to obtain a micro-grid with prior scheduling knowledge; mining and utilizing the learned scheduling knowledge through transfer learning, and building a similarity calculation model for re-utilizing the scheduling knowledge; and performing fine adjustment learning in a new micro-grid optimal scheduling task by using the prior scheduling knowledge to obtain an optimal scheduling strategy of the new task. According to the method, the reinforcement learning and the transfer learning are introduced into the micro-grid optimization scheduling, the reinforcement learning has the strong self-learning and memory ability, and the experience knowledge learned in an optimization process can be stored in neural network parameters.

Description

technical field [0001] The invention relates to the technical field of smart grids, in particular to a method for optimal scheduling of micro-grids based on reinforcement learning. Background technique [0002] With the rapid growth of society's electricity demand, the traditional power grid is facing the imbalance between supply and demand caused by the depletion of fossil resources, as well as ecological and environmental problems such as global warming and ozone layer destruction caused by the use of traditional energy. As a green and sustainable energy, renewable energy can reduce environmental pollution in the process of production and consumption, and alleviate energy shortage and ecological environment problems to a certain extent. Therefore, countries around the world have begun to focus on distributed power generation technologies with renewable energy as the core. However, due to climate, environmental and other factors, the power supply quality and reliability of ...

AI Technical Summary

Problems solved by technology

For example, Newton's method, linear programming, quadratic programming, interior point method, etc., these methods have advantages in solution speed and convergence reliability, but when the problem is complex nonlinear, the objective function and constraints are discontinuous, the optimization The performance cannot be guaranteed, it is easy to fall into local optimum or even fail, and the application flexibility is poor

2) Intelligent algorithms, such as mixed integer linear programming method, dynamic programming method, genetic algorithm, particle swarm optimization algorithm, ant colony algorithm, etc. Compared with classical mathematical methods, heuristic algorithms rely less on mathematical models and are easy to deal with Linear problems have high application flexibility and are widely used in different optimization problems of power systems, but the parameter setting of the heuristic algorithm is highly random, and the results are greatly affected by it

[0004] Through in-depth analysis, it is found that these two types of algorithms have a common problem. The solution to the optimization task is isolated, and the learned experience and knowledge have not been effectively accumulated during the learning process. Therefore, when facing new tasks, they have to To start learning from the beginning, in simple terms, is to not have the ability to accumulate and utilize optimization information

Images