CN-121981336-A - Purification system load distribution prediction method, system, equipment and medium

CN121981336ACN 121981336 ACN121981336 ACN 121981336ACN-121981336-A

Abstract

The application relates to the technical field of water management. The method comprises the steps of extracting residual characteristics from acquired flowmeter time sequence data, loading the residual characteristics to a static topological graph structure to obtain a residual space-time diagram, fusing a physical influence matrix and the residual space-time diagram through a double-channel attention mechanism to generate a dynamic adjacency matrix, carrying out graph convolution operation based on the dynamic adjacency matrix to generate node embedded vectors, decoding the node embedded vectors to generate an overcurrent pressure predicted value and a load deviation degree, generating a dynamic threshold value based on historical load data and working condition data acquired in real time, and evaluating the load deviation degree according to the dynamic threshold value to obtain an overflow risk early warning signal so as to achieve the technical effects of improving the modeling accuracy of the fluid transmission characteristics of a pipe network, enhancing the collaborative analysis capability of working conditions and flow data and optimizing the overflow risk early warning accuracy under the dynamic working conditions.

Inventors

ZHAO MEI
ZHANG DONG
ZHAO SEN
SUN BOWEI
LI CHUANHONG

Assignees

交通运输部公路科学研究所

Dates

Publication Date: 20260505
Application Date: 20260225

Claims (10)

1. A method for predicting load distribution in a decontamination system, the method comprising: Constructing a topology model based on the acquired pipe network geometric connection relation and preset pipeline physical properties to obtain a static topology graph structure; Carrying out hydrodynamic simplified calculation on the working condition data acquired in real time to generate a physical influence matrix; extracting residual characteristics from the acquired flowmeter time sequence data, and loading the residual characteristics to the static topological graph structure to obtain a residual time-space diagram; The physical influence matrix and the residual space-time diagram are fused through a double-channel attention mechanism to generate a dynamic adjacency matrix; The node embedded vector is decoded to generate an overflow pressure predicted value and a load deviation degree, a dynamic threshold is generated based on historical load data and working condition data acquired in real time, and the load deviation degree is evaluated according to the dynamic threshold to obtain an overflow risk early warning signal.
2. The purge system load distribution prediction method according to claim 1, wherein the extracting residual features from the acquired flow meter time series data comprises: Carrying out hydraulic balance calculation on the node connection relation in the static topological graph structure and the pressure distribution in the working condition data acquired in real time through a trained pipe network hydraulic transmission model to obtain a theoretical flow reference sequence; Calculating the real-time deviation between the actual measurement flow time sequence data and the theoretical flow reference sequence, and generating an original residual signal; Carrying out multi-scale fluctuation analysis on the original residual signal, and extracting transient fluctuation amplitude, duration and frequency domain energy distribution characteristics; And fusing the transient fluctuation amplitude, the duration and the frequency domain energy distribution characteristic to obtain the residual characteristic.
3. The method according to claim 1, wherein the evaluating the load deviation according to the dynamic threshold value, to obtain an overflow risk early warning signal, comprises: Performing multi-level interval matching on the load deviation degree and the dynamic threshold value to generate a risk level mark; combining preset topological weights of key nodes of the pipe network, and carrying out space weighting correction on the risk level marks to obtain enhanced risk marks; And carrying out overflow event pattern matching based on the enhanced risk mark, and generating the overflow risk early warning signal.
4. The method for predicting load distribution of a purification system according to claim 3, wherein the performing spatial weighted correction on the risk level label by combining a preset topology weight of a key node of a pipe network to obtain an enhanced risk label comprises: calculating the space influence factor of each node according to the topology weight of the key nodes of the pipe network; carrying out nonlinear weighted fusion on the risk level marks and the space influence factors of the corresponding nodes to obtain weighted results; And adjusting the weighted result based on the preset pipe network region functional sensitivity, and generating the enhanced risk marker, wherein the expression of the enhanced risk marker is as follows: Wherein, the Representing nodes Is a risk-enhancing marker of (c), Representing nodes Is a risk level indicator of (1), Representing nodes Is used for the spatial influencing factor of (a), Representing the non-linear mapping function, Representing nodes Is used for the functional sensitivity coefficient of the (c), The normalization function is represented as a function of the normalization, Representing the target node index.
5. A method of decontamination system load distribution prediction as claimed in claim 3, wherein said performing overflow event pattern matching based on said enhanced risk signature to generate said overflow risk early warning signal comprises: Matching the enhanced risk mark with a preset overflow event feature pattern library to obtain a pattern similarity matrix; identifying a key abnormal propagation path from the pattern similarity matrix to generate a risk conduction pattern; and triggering an early warning decision according to the risk conduction mode, and generating the overflow risk early warning signal.
6. The method of claim 1, wherein decoding the node embedded vector to generate an over-current pressure prediction value and a load bias comprises: Analyzing the pressure related characteristics of the node embedded vectors to generate the overcurrent pressure predicted value of each purifying node; Based on the node connection relation of the static topological graph structure, the overflow pressure predicted values of adjacent nodes are aggregated to obtain regional pressure gradient distribution, and the expression of the regional pressure gradient distribution is as follows: Wherein, the The pressure gradient distribution of the region is shown, Representing nodes Is used for the over-current pressure prediction value of the (a), Representing nodes Is used to determine the neighbor set of a neighbor, Representing the whole set of the nodes of the pipe network, Representing neighbor nodes Is used for the over-current pressure prediction value of the (a), Representing the index of the neighbor node, Representing a target node index; and calculating a load distribution deviation coefficient according to the regional pressure gradient distribution and the preset rated load capacity of the node, and generating the load deviation degree.
7. The purification system load distribution prediction method according to claim 1, wherein the constructing a topology model based on the acquired pipe network geometric connection relationship and the preset pipe physical attribute to obtain a static topology structure comprises: Constructing an undirected graph topology framework based on the acquired pipe network node coordinates and the pipeline connection relation; injecting a preset pipeline physical attribute set into the undirected graph topology framework to obtain an attribute enhancement topology; And marking functional node types in the attribute enhancement topology, and generating the static topological graph structure, wherein the functional node types comprise purifying nodes, shunting nodes and buffering nodes.
8. A decontamination system load distribution prediction system, the system comprising: The static topology modeling module is used for constructing a topology model based on the acquired pipe network geometric connection relation and preset pipeline physical properties to obtain a static topology graph structure; The working condition dynamics analysis module is used for carrying out fluid dynamics simplified calculation on the working condition data acquired in real time to generate a physical influence matrix; the residual space-time construction module is used for extracting residual characteristics from the acquired flowmeter time sequence data, and loading the residual characteristics to the static topological graph structure to obtain a residual space-time graph; The dynamic graph rolling module is used for fusing the physical influence matrix and the residual space-time diagram through a double-channel attention mechanism to generate a dynamic adjacent matrix; The risk early warning decision module is used for decoding the node embedded vector to generate an overflow pressure predicted value and a load deviation degree, generating a dynamic threshold value based on historical load data and working condition data acquired in real time, and evaluating the load deviation degree according to the dynamic threshold value to obtain an overflow risk early warning signal.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of a purification system load distribution prediction method according to any one of claims 1 to 7.
10. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program when executed by a processor implements the steps of a purification system load distribution prediction method according to any one of claims 1 to 7.

Description

Purification system load distribution prediction method, system, equipment and medium Technical Field The invention relates to the technical field of water management, in particular to a purification system load distribution prediction method, a purification system load distribution prediction system, purification system load distribution prediction equipment and purification system load distribution prediction media. Background With the rapid development of intelligent water technology, the purification system plays an increasingly critical role in urban water resource management, sewage treatment and environmental protection. The load distribution prediction is used as a core technology for safe operation of the purification system, and by dynamically analyzing key parameters such as pipeline flow, pressure and the like, overflow accidents can be effectively prevented, water quality safety is guaranteed, and resource utilization is optimized. In the traditional technology, static modeling is carried out by relying on a simplified pipe network physical model, dynamic coupling effects of geometric connection and pipeline attributes are ignored, real fluid transmission characteristics cannot be accurately reflected, working conditions and flow data are processed in a fracturing mode, cooperative analysis of the two is lacking, overflow risk precursors are difficult to capture in real time, in addition, risk assessment is carried out through fixed threshold values and manual experience, dynamic working conditions cannot be adapted, and key risk nodes are difficult to position. Disclosure of Invention Based on the above, it is necessary to provide a method, a system, a device and a medium for predicting load distribution of a purification system, so as to achieve the technical effects of improving the modeling accuracy of the fluid transmission characteristics of a pipe network, enhancing the collaborative analysis capability of working conditions and flow data, and optimizing the overflow risk early warning accuracy under dynamic working conditions. In a first aspect, the present application provides a purification system load distribution prediction method, the method comprising: Constructing a topology model based on the acquired pipe network geometric connection relation and preset pipeline physical properties to obtain a static topology graph structure; Carrying out hydrodynamic simplified calculation on the working condition data acquired in real time to generate a physical influence matrix; extracting residual characteristics from the acquired flowmeter time sequence data, and loading the residual characteristics to a static topological graph structure to obtain a residual time-space graph; The physical influence matrix and the residual space-time diagram are fused through a double-channel attention mechanism to generate a dynamic adjacency matrix; the method comprises the steps of decoding node embedded vectors, generating an overcurrent pressure predicted value and a load deviation degree, generating a dynamic threshold value based on historical load data and working condition data acquired in real time, and evaluating the load deviation degree according to the dynamic threshold value to obtain an overflow risk early warning signal. In one embodiment, extracting residual features from acquired flow meter timing data includes: Carrying out hydraulic balance calculation on node connection relations in a static topological graph structure and pressure distribution in working condition data acquired in real time through a trained pipe network hydraulic transmission model to obtain a theoretical flow reference sequence; Calculating the real-time deviation between the actual measurement flow time sequence data and the theoretical flow reference sequence, and generating an original residual error signal; carrying out multi-scale fluctuation analysis on the original residual signal, and extracting transient fluctuation amplitude, duration and frequency domain energy distribution characteristics; and fusing transient fluctuation amplitude, duration and frequency domain energy distribution characteristics to obtain residual characteristics. In an embodiment, evaluating the load deviation according to the dynamic threshold to obtain the overflow risk early warning signal includes: carrying out multi-level interval matching on the load deviation degree and the dynamic threshold value to generate a risk level mark; Combining preset topological weights of key nodes of the pipe network, and carrying out space weighting correction on the risk grade marks to obtain enhanced risk marks; and carrying out overflow event pattern matching based on the enhanced risk mark, and generating an overflow risk early warning signal. In an embodiment, combining preset topological weights of key nodes of a pipe network, performing spatial weighted correction on risk level marks to obtain enhanced risk marks, including: calculating th