Optimal method for regional power grid operation point dispatching based on deep q-network

Abstract

The invention provides a deep Q-network-based scheduling optimization method for regional power grid operating points, including the steps of: determining the scheduling architecture of the regional power grid operating point dispatching center; modeling the dynamic decision-making process of regional power grid operating point scheduling as a corresponding Markov Decision-making process model, which includes state, action, cost and optimization objective function; use deep Q network to solve the Markov dynamic decision-making process model; regional power grid operation points include at least the output of conventional thermal power units and wind power in the next scheduling period Unit output and flexible load reduction; the dispatch of regional grid operating points at least includes the dynamic determination of the grid operating point in the next scheduling period based on ultra-short-term forecast information of wind power and load power and current period operating point information. The invention can effectively deal with the randomness of new energy output and load demand, fully exploit the potential of source-load interactive scheduling, maintain the dynamic balance of regional power grid power, and improve the operating efficiency of the power grid system.

Description

technical field [0001] This patent relates to the technical field of power grid intelligent dispatching, in particular to a method for optimizing the dispatching of regional power grid operating points based on a deep Q network. Background technique [0002] With the rapid development of large-scale wind power grid-connected and flexible loads, source-load bilateral uncertainties have an increasingly serious impact on power grid dispatching operations. Source-load interaction is an important part of the source-grid-load interaction framework system, which can Promote the rational use of power generation and consumption resources, and improve the comprehensive benefits of system operation safety, economy, and cleanliness. [0003] In the existing technology, there is a scheme that considers interruptible load and incentive load in the day-ahead scheduling model, which reflects the potential peak-shaving benefit of load-side resources, such as Yang Nan et al [Yang Nan, Wang Bo...

AI Technical Summary

Problems solved by technology

These schemes deal with the randomness of resources on the power generation side through flexible load scheduling, and have certain reference significance for dynamic scheduling of power grid operating points. However, traditional methods for dealing with uncertain grid scheduling problems mainly focus on robust optimization methods, uncertain Planning, etc., when solving the problem of dynamic scheduling of regional power grid operating points, as the scale of the problem increases, the traditional algorithm will have problems such as large amount of calculation, "curse of dimensionality", and difficulty in obtaining the global optimal solution. big challenge

Images