Interruptible load optimization method based on deep reinforcement learning

Abstract

The invention relates to an interruptible load optimization method based on deep reinforcement learning, and belongs to the field of power distribution network demand response. The method comprises the following steps: acquiring an observation state of a system at a moment t through an intelligent electric meter and a measurement device, and taking the observation state as an observation sample; training a neural network based on a dueling depth Q network (DDQN) by using the acquired observation sample, periodically clearing the observation sample to ensure that the neural network learns the latest observation state, and acquiring the trained neural network through a certain number of calculation iterations; reading the data in the measurement device to acquire the real-time state of a power distribution network; sending to the trained neural network; and screening interruptible load points by using a node voltage within the allowable range after the interruptible load action as a constraint condition and user satisfaction as an index. According to the invention, the power consumption habit of a user can be automatically recognized, a set of interruptible load points meeting the operation condition and having the high user satisfaction are screened out, and the method has good application prospects.

Description

technical field [0001] The invention relates to the field of distribution network demand side management, and more specifically, relates to an interruptible load optimization method based on deep reinforcement learning. Background technique [0002] Distributed energy resource (DER) output has random fluctuations due to the influence of weather factors. Its extensive access to the distribution network leads to a decline in the power quality of the distribution network. Demand-side response is a flexible and fast user-side The response method is that the users of the distribution network directly adjust their own load demand and power consumption mode. It is completely a voluntary participation behavior of the user, which aims to quickly improve demand elasticity and smooth the load curve. effective measures. [0003] The Chinese invention patent application with the publication number CN 108767883 A discloses a response processing method on the demand side, including: obtai...

AI Technical Summary

Problems solved by technology

[0003] The Chinese invention patent application with the publication number CN 108767883 A discloses a response processing method on the demand side, including: obtaining historical data of power consumption of users, the historical data including historical data of power consumption of fixed power consumption equipment and controllable power consumption The historical data of power consumption of the equipment; the statistics of the historical data are used to identify the abnormal points of the load; the statistics of the historical data of the removed abnormal points of the load are carried out to obtain the average power consumption data in the billing cycle; the controllable power consumption equipment The relationship between power and time is the independent variable, the relationship between the charging and discharging power of the energy storage device and time is the independent variable, the total energy consumption of the preset working cycle of the controllable power consumption device is the constraint, and the capacity of the energy storage device is As a constraint condition, the relationship between the power and time of the controllable power consumption device and the charge and discharge power of the energy storage device and Optimize the relationship between time; process the controllable power consumption device and energy storage device according to the optimization result, and optimize the relationship between the power and time of the controllable power consumption device and the relationship between the charging and discharging power and time of the energy storage device at the same time , which further increases the role of existing equipment in stabilizing power consumption, reduces costs and reduces fluctuations in the power grid, but can only perform statistical analysis on historical data, cannot automatically identify users' electricity consumption habits, and cannot be optimally controlled strategy, cannot guide electricity users to use electricity rationally

Images