Reservoir group scheduling decision behavior mining method and reservoir scheduling automatic control device

Abstract

The invention provides a reservoir group scheduling decision behavior mining method and a reservoir scheduling automatic control device, and the method comprises the following steps: 1, determining a research scene, collecting basic data, historical scheduling data and reservoir region station meteorological data of a reservoir group system, and determining an impact factor and a decision variable of a reservoir group scheduling behavior; 2, determining a deep learning algorithm for mining reservoir group scheduling decision behaviors; 3, examining the accuracy of reservoir group scheduling data; 4, constructing a reservoir group scheduling decision behavior mining model coupling the reservoir basic principle and the deep learning model; and step 5, calibrating hyper-parameters of the model based on training set samples, updating network parameters of the model based on model loss function back propagation, determining optimal hyper-parameters of the model according to simulation precision of the model in a test set, finally establishing a mapping relation between influence factors of reservoir group scheduling behaviors and decision variables, and realizing mining of reservoir group scheduling decision behaviors.

Description

technical field [0001] The invention belongs to the technical field of reservoir dispatching, and in particular relates to a reservoir group dispatching decision-making behavior excavation method and a reservoir dispatching automatic control device. [0002] technical background [0003] Reservoirs are widely used flood storage and dry-water projects. According to the "2018 National Water Conservancy Development Statistical Bulletin", a total of 98,822 reservoirs have been built in my country. With the transformation of watershed dispatching management and the development of science and technology, the "sky-air-ground integrated observation technology" has recorded a large amount of historical hydrological information and water regime information in the actual dispatching process, causing the data information in the field of reservoir dispatching to explode. increase. At present, how to mine the existing massive data information to provide suggestions for reservoir dispatchi...

AI Technical Summary

Problems solved by technology

[0006] (1) There are many influencing factors in the decision-making process of reservoir dispatching. For the traditional shallow learning model, when the input dimension increases, the calculation difficulty increases. It is necessary to use the factor screening method to determine the input characteristics. It is difficult to consider multiple influencing factors. It is also difficult to adapt to various decision-making scenarios of reservoir scheduling when fully mining data information;

[0007] (2) Using the deep learning model to mine reservoir dispatching behavior does not require manual factor screening, and can fully absorb historical dispatching data and hydrometeorological information, but it only establishes the mapping relationship between influencing factors and decision variables based on available data information, and cannot be combined The water balance in the process of reservoir dispatching and the objective constraints of dispatching decisions are difficult to comply with the principles of reservoir dispatching, and it is impossible to obtain accurate and effective data for reservoir dispatching

Images