CN-121996392-A - Intelligent water conservancy centralized control center system based on intelligent on-duty body cluster and scheduling method

CN121996392ACN 121996392 ACN121996392 ACN 121996392ACN-121996392-A

Abstract

The invention discloses an intelligent water conservancy centralized control center system based on an intelligent on-duty body cluster and a scheduling method, belonging to the technical field of intelligent water conservancy and industrial Internet, wherein the system comprises a multi-source data fusion and state sensing module, a multi-agent cooperative scheduling engine, a digital twin simulation deduction module, a task decomposition and cooperative execution module and the intelligent on-duty body cluster, the intelligent on-duty body deployed in each water conservancy facility collects operation data in a non-invasive visual perception and analog peripheral mode and executes control operation, the multi-agent cooperative scheduling engine performs cross-facility joint optimization based on global state data, and the digital twin module performs simulation verification on a scheduling scheme and then issues and executes the simulation verification.

Inventors

WANG JIANJUN
XU CHAOHUI
CHEN XIFENG
ZHU XIAOXIAO
WANG JINGWEN
LIN LUYING
WANG JIFEN
ZHANG SHAOYU

Assignees

水利部农村电气化研究所
杭州思绿数字科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260409

Claims (10)

1. Intelligent water conservancy centralized control center system based on intelligent on-duty body cluster, its characterized in that adopts high in the clouds centralized control platform layer, the three-layer architecture of side intelligent on-duty body layer and safe communication layer, includes: the multi-source data fusion and state sensing module is deployed at the cloud end and is used for converging multi-source heterogeneous data from the intelligent on-duty body cluster, performing space-time alignment and feature fusion and generating a global state data set; The multi-agent cooperative scheduling engine is deployed at the cloud, each water conservancy facility is modeled as an agent with sensing, decision and execution capabilities, a centralized training decentralized execution architecture is adopted, a global value evaluation network of the multi-agent cooperative scheduling engine adopts QMIX network structures to carry out nonlinear mixing on the local value of each agent, and the multi-agent cooperative scheduling engine is realized through mixed network weight non-negative constraint Monotonic constraint of (2); The digital twin simulation deduction module is deployed at the cloud end and is used for carrying out simulation verification in a digital twin body consistent with an actual water conservancy system before the dispatching scheme is issued and executed, the simulation time scale is more than 100 times of the actual time, and the dispatching scheme is allowed to be issued and executed only after the simulation verification passes through the safety constraint; The task decomposition and collaboration execution module is deployed at the cloud end and is used for analyzing the simulated and verified scheduling scheme into a task execution graph with a dependency relationship and issuing the task execution graph; The intelligent on-duty body cluster consists of edge intelligent terminals arranged beside each hydraulic facility control room, is connected with the bypass of the existing facility control system in a non-invasive access mode, does not reform an original control loop, is responsible for local data acquisition and control operation execution, and is locally and safely operated according to a preset program when a cloud network is interrupted; and the safety communication layer is connected with the cloud end and the side end by adopting a water conservancy private network VPN encryption tunnel.
2. The system of claim 1, wherein the multi-source data fusion and status awareness module employs a multi-head attention mechanism to achieve deep fusion of different types of operational data, the number of attention heads ranging from 4 to 8.
3. The system of claim 1, wherein the global rewards function of the multi-agent collaborative scheduling engine comprises four sub-terms of power generation benefit, flood control safety, ecological flow and collaborative efficiency, and each sub-term weight is adaptively adjusted according to current hydrographic weather conditions through a dynamic attention mechanism.
4. The system of claim 1, wherein the hydraulic simulation engine of the digital twin simulation deduction module performs a numerical solution to the riverway unsteady flow based on the san-valance equation group and performs a discrete using PREISSMANN four-point implicit differential format.
5. The system of claim 1, wherein the digital twin simulation deduction module further comprises a scenario generator for generating multiple meteorological hydrologic scenarios for pressure testing based on a historical extreme conditions database and monte carlo sampling method.
6. The system of claim 1, further comprising an intelligent cockpit module for constructing a three-dimensional visualization scene using WebGL technology, providing an interpretable presentation of scheduling decisions.
7. The intelligent water conservancy scheduling method based on the system of any one of claims 1 to 6 is characterized by comprising the steps of S1, data acquisition and fusion, S2, collaborative scheduling decision, S3, simulation verification, wherein the scheduling scheme generated in the step S2 carries out simulation deduction in a digital twin body, and the intelligent water conservancy scheduling method enters the step S4 only under the condition that safety constraint is met, S4, task distribution, and S5, local execution and feedback.
8. The method according to claim 7, wherein in the step S2, the first step is The individual agents are at the moment Is a local observation of (2) The system consists of four parts, namely a real-time operation parameter vector of the facility, a recent historical state sequence, weather forecast data and associated states of upstream and downstream facilities.
9. The method according to claim 7, wherein in the step S3, for each scheduling scheme, a monte carlo simulation evaluation is performed under 200 scenes, and when the simulation index under any scene exceeds the safety threshold, the information of the offending scene is recorded and fed back to the step S2 to change the safety constraint.
10. The method of claim 7, wherein in step S5, when the cloud network is interrupted, the intelligent on-duty body at the side can still operate safely according to the predetermined program.

Description

Intelligent water conservancy centralized control center system based on intelligent on-duty body cluster and scheduling method Technical Field The invention relates to the technical field of intelligent water conservancy and industrial Internet, in particular to an intelligent water conservancy centralized control center system based on an intelligent on-duty body cluster and a scheduling method. Background With the deep advancement of water conservancy informatization construction, the unified monitoring and cooperative scheduling of river basin-level water resources become a core target of water conservancy modernization. In the water conservancy and hydropower industry, the push of unattended and watershed integrated scheduling is a development trend, and the traditional centralized control center construction mode faces serious challenges. The invention CN120338210A discloses a reservoir dispatching method based on deep learning self-adaptive dynamic network and reinforcement learning, which belongs to the technical field of reservoir dispatching and comprises the steps of acquiring real-time water level data, meteorological data, regional cloud layer and radar image data of earth surface characteristics of a reservoir, carrying out data fusion on the three data, inputting the fused data into a preset self-adaptive dynamic transducer network model to obtain predicted output, wherein the predicted output comprises a reservoir water level sequence, a warehouse-in flow sequence and a lower flow sequence, constructing a reinforcement learning state space, inputting the reinforcement learning state space into a preset reinforcement learning network, wherein the action space of the reinforcement learning network comprises flood discharge, power generation flow and ecological flow, the optimization target of the reinforcement learning network is a maximal accumulated discount rewarding, the rewarding function is a weighting function of flood prevention rewarding items, power generation items and ecological rewarding items, and the reinforcement learning network is an optimal combination scheme for finally converging and outputting the flood discharge quantity, the power generation flow and the ecological flow through continuous iteration optimization. However, the above prior art has the following disadvantages. Firstly, the method is only optimized for the dispatching of a single reservoir, a centralized deep learning and reinforcement learning architecture is adopted, cross-facility cooperative dispatching can not be carried out on a plurality of heterogeneous water conservancy facilities, the requirement of basin-level water resource integrated management is difficult to meet, and when various facilities such as a hydropower station, a water pump station, a sluice gate and the like are required to be comprehensively operated in a combined mode, the method is difficult to effectively play a role. Secondly, the method relies on direct data access to the existing reservoir automation system, requires the bottom layer equipment to have a unified communication protocol such as IEC61850 and a standardized data interface, and needs large-scale automation system transformation to realize data acquisition and remote control for various hydraulic facilities with different construction ages, complicated equipment brands and numerous legacy systems, so that the transformation cost is high, the implementation period is long, and the risks and the resistance are huge. Thirdly, although the reinforcement learning decision process of the method adopts a self-adaptive dynamic transducer network to predict, the link of virtual simulation verification of a scheduling scheme in a digital twin environment is lacking, and the scheduling scheme directly issues and executes the scheduling scheme, so that a certain safety risk exists, and especially under extreme meteorological conditions such as typhoon season flood control and pre-leakage scenes, unexpected results can be generated, and the safety guarantee of executing after simulation can not be realized. Fourth, the method adopts a centralized deep learning architecture, all data acquisition, prediction calculation and decision optimization are completed in the cloud, the requirement on network real-time performance is high, the safe operation of facilities cannot be guaranteed under the condition of communication interruption or network delay, and the edge autonomous execution capability is lacked. Therefore, there is a need for an intelligent water conservancy centralized control center system and method that can quickly integrate various heterogeneous water conservancy facilities in a low-cost, non-invasive manner, support cross-facility collaborative optimization scheduling, and have simulation verification and marginal autonomous execution capabilities. Disclosure of Invention The invention aims to overcome the defects of the prior art and provides an intelligent water c