Flexible emergency pressure relief control method and system for water conveyance structures

Abstract

The application discloses a kind of flexible emergency pressure relief control method and system of water conveyance structure, it is related to the safety technical field of hydraulic engineering.The method comprises: obtaining the multimodal monitoring signal of water conveyance structure and extracting the abnormal feature criterion therein;Based on physical causal constraints, multi-source data fusion is carried out on the abnormal feature criterion to obtain a judgment conclusion containing abnormal types and confidence levels;Based on the conclusion, time-domain state prediction is performed on the first type of nodes with monitoring equipment, and spatial topology deduction is performed on the second type of nodes without monitoring equipment to obtain a network risk assessment result;Integrate the judgment conclusion and risk assessment result, control the flexible emergency pressure relief device to execute pressure relief action, and implement linkage control on the downstream pipe section with hydraulic correlation.The application can improve the technical problems of multi-source signal fusion conflict and insufficient topology deduction accuracy under complex working conditions, and realize network early warning and flexible intervention.

Description

Technical Field

[0001] This invention relates to the field of water conservancy engineering and automation control technology, and in particular to a flexible emergency pressure relief control method and system for water conveyance structures. Background Technology

[0002] Long-distance, large-diameter water pipelines and pressurized tunnels are the physical lifelines of modern water resource allocation projects. In hydrodynamic operating environments, transient changes in fluid states and hydraulic turbulence can cause impact damage to the pipeline network structure. Achieving early identification of abnormal fluid fluctuations and dynamic prediction of the entire network's risk evolution, and implementing interventions within a defined time window, helps ensure the physical integrity and long-term continuous and stable operation of water conservancy facilities.

[0003] Currently, safety monitoring and control of pressurized water transmission systems typically rely on fixed-location hardware devices such as pressure transmitters and flow meters. In conventional system processing, various sensor signals are often directly input into the data processing center in parallel, and abnormal operational status is determined through fixed threshold over-limit alarm rules or black-box neural network models trained based on historical samples. In the physical intervention stage, pipeline disaster prevention systems can rely on fixed safety relief valves at a single site for pressure relief.

[0004] In summary, existing methods still have shortcomings in terms of the ability to fuse multi-source monitoring data to resist interference and maintain physical consistency under transient hydrodynamic conditions, as well as in the joint simulation of the spatiotemporal global situation of pipeline networks across nodes and the ability to smoothly intervene in end-point physical execution. Therefore, it is necessary to study a safety control scheme for water transmission pipeline networks. Summary of the Invention

[0005] The purpose of this invention is to provide a flexible emergency pressure relief control method and system for water conveyance structures, in order to solve the aforementioned problems in the prior art.

[0006] The technical solution, in its first aspect, is a flexible emergency pressure relief control method for water conveyance structures, comprising:

[0007] Acquire multimodal monitoring signals from water conveyance structures;

[0008] Extracting abnormal feature criteria from multimodal monitoring signals;

[0009] Based on physical causal constraints, multi-source data fusion is performed on the anomaly feature criteria to obtain anomaly judgment conclusions that include anomaly type and confidence level.

[0010] Based on the anomaly determination, time-domain state prediction is performed on the first type of nodes with monitoring equipment, and spatial topology deduction is performed on the second type of nodes without monitoring equipment to obtain the network-wide risk assessment results.

[0011] Based on the anomaly determination conclusion and the network-wide risk assessment results, the flexible emergency pressure relief device is controlled to perform pressure relief actions, and coordinated control is implemented on downstream pipe sections that have a hydraulic relationship with the anomaly node.

[0012] In conjunction with the first aspect, abnormal feature criteria are extracted from multimodal monitoring signals, including:

[0013] The sliding time window mechanism is used to detect the pressure change trend in multimodal monitoring signals and extract the pressure change criterion.

[0014] Frequency domain transformation is performed on the multimodal monitoring signal, and the energy proportion of the characteristic frequency band is extracted to obtain the water hammer spectrum criterion;

[0015] The traffic prediction residual is obtained by using a pre-built state tracking model, and traffic anomaly criteria are extracted.

[0016] In conjunction with the first aspect, in the process of fusing multi-source data based on physical causal constraints to obtain anomaly determination conclusions that include anomaly type and confidence level, a two-layer processing architecture is specifically adopted, including:

[0017] The abnormal feature criteria are input into the preset front-end physical gate layer. By verifying the causal consistency between different abnormal feature criteria, the evidence source weight label corresponding to each abnormal feature criterion is output.

[0018] The abnormal feature criteria and the corresponding evidence source weight labels are input into the post-evidence theory fusion layer to perform multi-source evidence synthesis calculation and output the abnormality judgment conclusion.

[0019] In conjunction with the first aspect, based on the anomaly determination conclusion, time-domain state prediction is performed on the first type of nodes equipped with monitoring devices, specifically including:

[0020] Obtain the pre-configured hydraulic parameters of the pipeline network topology;

[0021] A discrete state-space prediction model is constructed based on the hydraulic parameters of the pipeline network topology. The discrete state-space prediction model includes a state transition matrix and an input control matrix.

[0022] The initial state vector of the discrete state-space prediction model is updated based on multimodal monitoring signals;

[0023] By using the state transition matrix and the input control matrix, a multi-step iterative deduction is performed on the initial state vector to output the multi-step predicted state distribution of the first type of node in the future time period.

[0024] Combining the first aspect, a spatial topology simulation was performed on the second type of nodes without monitoring equipment to obtain the overall network risk assessment results, specifically including:

[0025] Obtain the predicted state of the first type of node, which is the output of time-domain state prediction for the first type of node with monitoring equipment;

[0026] Based on the topological connection relationship and physical hydraulic impedance characteristics of the pipeline network nodes, a graph Laplace matrix is ​​constructed.

[0027] The predicted state of the first type of node is propagated to the second type of node along the pipeline network topology using the graph Laplace matrix. The flow rate of the pipeline segment is calculated by combining the pressure and flow coupling relationship of the pipeline segment, so as to generate the risk assessment results of the whole network.

[0028] In conjunction with the first aspect, the process of extracting abnormal feature criteria from multimodal monitoring signals also includes:

[0029] Identify the current operating mode of the water supply network;

[0030] Based on the operating condition mode, the sensitivity parameters of the criteria for each dimension of multimodal monitoring signals are adjusted collaboratively.

[0031] The feature extraction window update points of each multimodal monitoring signal are aligned with a preset unified control cycle, so that the judgment indicators of different dimensions are output synchronously as abnormal feature criteria within the same time window.

[0032] In conjunction with the first aspect, controlling the flexible emergency pressure relief device to perform pressure relief actions includes:

[0033] When the risk assessment results of the entire network indicate that there is a risk of transient overpressure impact in the pipeline network, the flexible buffer medium cavity inside the flexible emergency pressure relief device is used to absorb the energy of the transient pressure wave through reversible elastic deformation.

[0034] Based on the anomaly type and confidence level in the anomaly determination conclusion, the adjustable pressure relief channel of the flexible emergency pressure relief device is controlled to perform graded actions or continuously adjustable pressure relief opening operations.

[0035] Based on the results of the overall network risk assessment, upper limit constraints and anti-shake control are implemented on the opening change rate and action holding time of the adjustable pressure relief channel.

[0036] Based on the first aspect, extract the criteria for traffic anomaly detection, including:

[0037] Analyze the temporal variation patterns of traffic forecast residuals across multiple sliding windows;

[0038] If the time-series change pattern exhibits a step-like abrupt change exceeding the step threshold, then output an anomaly criterion characterizing water load fluctuations.

[0039] If the time-series change pattern shows a continuous offset that accumulates beyond the offset threshold, and is accompanied by abnormal temperature signals in the multimodal monitoring signals of a local pipe section, then an abnormality criterion characterizing leakage will be output.

[0040] Secondly, a flexible emergency pressure relief control system for a water conveyance structure, used to implement the method as described in any one of the first aspects, specifically includes:

[0041] The signal acquisition module is used to acquire multimodal monitoring signals from water conveyance structures;

[0042] The feature extraction module is used to extract abnormal feature criteria from multimodal monitoring signals;

[0043] The data fusion module is used to perform multi-source data fusion on anomaly feature criteria based on physical causal constraints to obtain anomaly judgment conclusions that include anomaly type and confidence level.

[0044] The risk assessment module is used to perform time-domain state prediction for the first type of nodes with monitoring equipment and spatial topology deduction for the second type of nodes without monitoring equipment based on the anomaly determination conclusion, so as to obtain the risk assessment results of the entire network.

[0045] The linkage control module is used to control the flexible emergency pressure relief device to perform pressure relief actions based on the anomaly judgment conclusion and the risk assessment results of the entire network, and to implement linkage control on downstream pipe sections that have hydraulic correlation with the abnormal node.

[0046] Thirdly, a processing apparatus comprising the system described in the second aspect.

[0047] Beneficial effects: This invention solves the technical problems of multi-source signal fusion conflicts and insufficient topology inference accuracy under complex operating conditions, and realizes network-wide early warning and flexible intervention. This will be described in detail in specific embodiments. Attached Figure Description

[0048] Figure 1 A flowchart of a flexible emergency pressure relief control method for water conveyance structures provided in this application embodiment.

[0049] Figure 2 This is a flowchart illustrating an example of extracting abnormal feature criteria from multimodal monitoring signals, provided as an embodiment of this application.

[0050] Figure 3 This is a flowchart illustrating another example of extracting abnormal feature criteria from multimodal monitoring signals, provided as an embodiment of this application.

[0051] Figure 4 A flowchart illustrating the pressure relief action of the flexible emergency pressure relief device provided in this application embodiment.

[0052] Figure 5 This is a flowchart illustrating the extraction of traffic anomaly criteria provided in an embodiment of this application.

[0053] Figure 6This is a schematic diagram of the overall structure of the flexible emergency pressure relief control system for water conveyance structures provided in this application embodiment.

[0054] Figure 7 This is a flowchart illustrating the multimodal anomaly detection and multi-source data fusion process provided in an embodiment of this application.

[0055] Among them, 1-pressure sensor, 2-flow sensor, 3-temperature sensor, 4-acoustic sensor, and 5-downstream branch valve. Detailed Implementation

[0056] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0057] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a predetermined order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in sequences other than those illustrated or described herein. Furthermore, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0058] To address the aforementioned issues, the applicant conducted in-depth searches and analyses, and discovered:

[0059] Under actual operating conditions, due to the high-frequency nonlinear characteristics of fluid transient phenomena, the signals collected by fixed-point equipment are easily affected by noise from the physical environment or local water cavitation, resulting in time sequence misalignment and numerical characteristic contradictions among multidimensional sensing signals.

[0060] Meanwhile, single black-box prediction models, which rely on data-driven approaches, are difficult to accurately describe the physical evolution and attenuation transmission process of hydraulic transients in spatial topological pipe networks.

[0061] In addition, traditional mechanical intervention methods are prone to technical problems such as delayed action response or water column separation and secondary oscillation in the pipeline network due to excessive release when facing high-frequency and high-pressure instantaneous shock waves.

[0062] To solve these problems, combined with Figures 1 to 7 The present invention will be specifically described through the following embodiments.

[0063] This embodiment provides a flexible emergency pressure relief control method for water conveyance structures, which includes the following steps:

[0064] Step 101: Acquire multimodal monitoring signals of water conveyance structures;

[0065] In this step, multimodal monitoring signals are various physical attribute data collected by sensor arrays deployed at key locations in water pipelines, pressurized tunnels, or related hydraulic structures. Specifically, they may include pressure and flow signals reflecting the fluid dynamics state, as well as acoustic and temperature signals reflecting the structural operating environment.

[0066] In view of this, the system can perceive the operating status of the water conveyance system in real time from multiple perspectives such as transient pressure shocks, flow fluctuations, energy spectrum anomalies, and local thermal changes, providing data support for anomaly detection.

[0067] Step 102: Extract abnormal feature criteria from multimodal monitoring signals;

[0068] This step can also be achieved using the following method:

[0069] The sliding time window mechanism is used to detect the pressure change trend in multimodal monitoring signals and extract the pressure change criterion.

[0070] Frequency domain transformation is performed on the multimodal monitoring signal, and the energy proportion of the characteristic frequency band is extracted to obtain the water hammer spectrum criterion;

[0071] A pre-built state tracking model is used to estimate the state of the flow signal in the multimodal monitoring signal, obtain the flow prediction residual, and extract the flow anomaly criteria.

[0072] This model represents a different level of application compared to the discrete state-space model discussed later.

[0073] One is used for flow residual calculation in the feature extraction stage, and the other is used for multi-step prediction in the risk assessment stage.

[0074] In this step, the extraction process refers to identifying technical indicators that deviate from the normal operating benchmark by performing time-domain analysis, frequency-domain analysis, or state-space transformation analysis on the acquired raw signal.

[0075] The extracted abnormal feature criteria can be used as a quantitative basis for characterizing whether the system has potential water hammer effects, abnormal valve operation, leakage, or accidental start-up and shutdown conditions.

[0076] For example, for pressure signals, the extracted abnormal feature criteria can be expressed as the rate of pressure change per unit time; while for acoustic signals, it can be expressed as the energy distribution within a predetermined characteristic frequency band.

[0077] Step 103: Based on physical causal constraints, multi-source data fusion is performed on the anomaly feature criteria to obtain an anomaly judgment conclusion that includes anomaly type and confidence level.

[0078] In this step, physical causal constraints refer to using the physical laws of fluid movement in the water supply network to determine the temporal and causal relationships between different monitoring signals.

[0079] Accordingly, when pressure sensor 1 detects a sudden change in pressure, the corresponding acoustic sensor 4 should capture the sound wave oscillation of the characteristic frequency within a predetermined physical propagation time. Through logical cross-verification, the risk of false alarms due to single-point sensor failure or localized environmental noise can be reduced.

[0080] Based on this, the generated anomaly determination conclusion clarifies the nature of the current anomaly, such as determining it as a water hammer event or valve malfunction, and provides the confidence level of the determination result.

[0081] Step 104: Based on the anomaly determination conclusion, perform time-domain state prediction on the first type of nodes with monitoring equipment, and perform spatial topology deduction on the second type of nodes without monitoring equipment to obtain the network-wide risk assessment results.

[0082] In other words, in response to the anomaly determination conclusion, and in combination with the multimodal monitoring signal, time-domain state prediction is performed on the first type of node with monitoring equipment to obtain the predicted state of the first type of node;

[0083] Based on the predicted state of the first type of nodes, a spatial topology simulation is performed on the second type of nodes without monitoring equipment to obtain the risk assessment results of the entire network.

[0084] In this step, the first type of node refers to physical locations with deployed hardware sensors and real-time observation capabilities, such as pump station outlets or key valve chambers; while the second type of node refers to nodes in the water transmission network that, although there are no direct monitoring methods, have a hydraulic coupling relationship with the first type of node through the pipeline topology.

[0085] Furthermore, temporal state prediction uses real-time state data and physical prediction models to extrapolate the changing trends of known measurement points in the future; while spatial topology extrapolation uses the principle of graph theory diffusion to extend the information of known nodes to unknown nodes in spatial dimensions.

[0086] It should be understood that the combined risk assessment results of the two can quantitatively reflect the risk of pressure exceeding limits at each node within the entire pipeline network.

[0087] Step 105: Based on the anomaly determination conclusion and the network-wide risk assessment results, control the flexible emergency pressure relief device to perform pressure relief actions, and implement linkage control on downstream pipe sections that have a hydraulic relationship with the anomaly node indicated by the anomaly determination conclusion.

[0088] In this step, a flexible emergency pressure relief device, which is an automated pressure relief component with a buffer energy-absorbing structure and continuously adjustable opening characteristics, is used to control the device to perform pressure relief actions. This reduces transient pressure peaks through controlled fluid release, thereby protecting the pipeline structure from damage.

[0089] At the same time, coordinated control, which is based on the hydraulic correlation of the pipeline network topology, coordinates the adjustment of branch valves or the distribution of flow in the affected downstream pipe sections to prevent the further spread and amplification of abnormal pressure waves in the entire pipeline network system and improve the overall safety of the system operation.

[0090] Based on the above embodiments, the abnormal feature criteria in the multimodal monitoring signals are extracted, including the following steps:

[0091] Step 201: Identify the current operating mode of the pipeline network where the water conveyance structure is located;

[0092] In this embodiment, the operating mode is the global fluid dynamics performance type of the system based on scheduling instructions and historical operating states.

[0093] Accordingly, the operating conditions can be divided into the following modes:

[0094] Stable water supply mode corresponds to the period when the pipeline network is in constant flow transmission, the valve opening is fixed and there is no pump station start-up or shutdown operation;

[0095] The high-risk window mode corresponds to the transition phase before and after pre-planned pump station start-up or shutdown operations, as well as the transition phase of large-scale pipeline valve opening adjustment.

[0096] The conflict suppression mode is a special state triggered when local signals oscillate violently and cannot be reconciled due to sensor noise interference in the preceding period.

[0097] Therefore, the current disturbance background of the pipeline network can be obtained, thereby avoiding false alarms or missed alarms caused by using a single fixed threshold.

[0098] Furthermore, the triggering mechanism for identifying operating conditions can rely on the timestamp of the scheduling instructions issued by the control center for feedforward matching, or it can be determined online based on the low-frequency trend changes in traffic status.

[0099] Step 202: Based on the operating condition mode, coordinately adjust the criterion sensitivity parameters for each dimension of the multimodal monitoring signal;

[0100] In this embodiment, the criterion sensitivity parameter refers to the numerical threshold used to trigger anomaly detection in each dimension, specifically including:

[0101] Pressure mutation adjustment coefficient used to constrain pressure signals;

[0102] Water hammer sensitivity coefficient used to constrain acoustic signals;

[0103] Flow residual determination coefficient used to constrain flow state.

[0104] Furthermore, based on the identified operating conditions, collaborative constraints are applied to ensure consistent parameter execution directions.

[0105] For example, in a stable water supply mode, the pressure surge adjustment coefficient and water hammer sensitivity coefficient are simultaneously increased to filter out conventional hydraulic fluctuation noise.

[0106] For example, in the high-risk window mode, since the pipeline network is prone to water hammer effect, the system synchronously lowers the parameters to reduce the threshold for anomaly judgment, making the system more sensitive to initial pressure changes and characteristic spectrum energy, thereby achieving pre-intervention.

[0107] For example, in conflict suppression mode, the continuity time requirement for acoustic peak stability can be increased, reducing false triggering within a single window.

[0108] Another example is that the adjustment of the criterion sensitivity parameter can also be achieved through closed-loop iteration based on historical false alarm statistics, and its update calculation relationship is as follows:

[0109] Я _new =Я _old +ΔЯ×E _count ;

[0110] Among them, Я _new For the updated pressure mutation adjustment coefficient, Я _old E is the initial pressure surge adjustment coefficient used in the current control cycle, ΔЯ is the preset iteration step size constant, and E is the initial pressure surge adjustment coefficient used in the current control cycle. _count This is a statistical value of the number of system false alarm overflows recorded within a unit time window.

[0111] In some implementations, in addition to using online iterative formulas, a static lookup table containing multiple sets of discrete parameter configuration items can be pre-constructed. When a working mode switching command is received, the corresponding parameter combination can be directly extracted from the lookup table and the update can be issued.

[0112] Step 203: Align the feature extraction window update points of each multimodal monitoring signal with a preset unified control cycle, and perform anomaly detection on each dimension signal based on the adjusted criterion sensitivity parameters, so that the judgment indicators of different dimensions are synchronously output as anomaly feature criteria within the same time window.

[0113] In some embodiments, the unified control period is a global time reference scale used to enforce the computational rhythm of different processing modules. Because the first-order difference computation response of pressure signals in the time domain is relatively fast, the conversion of acoustic signals from the time domain to the frequency domain requires accumulating samples of a predetermined length and performing a Fourier transform, while the Kalman state estimation of flow signals has an iterative delay, the processing time for the three signals in physical space differs.

[0114] Correspondingly, a unified control cycle can be set to force each module to cache its intermediate calculation results until the module with the longest processing time completes its calculation.

[0115] For example, the end point of the sliding time window, the truncation point of the Fourier transform, and the residual output point of the Kalman filter are all aligned to the boundary points of the control cycle.

[0116] This mechanism ensures that the pressure assessment indicators, spectrum assessment indicators, and traffic assessment indicators input to the data fusion module belong to the same physical time slice, thereby suppressing the fusion time misalignment problem caused by communication delay or asynchronous sampling.

[0117] Optionally, if there is an inherent physical transmission lag in the field actuator or communication link, the lag time constant can be directly appended to the alignment calculation logic of the unified control cycle to further suppress asynchronous errors at the hardware level.

[0118] In conjunction with the above embodiments, the anomaly feature criteria are input into a preset pre-physical gating layer. By verifying the causal consistency between different anomaly feature criteria, the evidence source weight labels corresponding to each anomaly feature criterion are output. Specifically, this may include the following steps:

[0119] Step 301: Obtain the trigger sequence relationship of different abnormal feature criteria.

[0120] The trigger sequence refers to the order in which the pressure mutation criterion, water hammer spectrum criterion, and flow anomaly criterion exceed their respective judgment thresholds on the time axis.

[0121] Because the pressure wave propagation speed of the fluid medium in the water conveyance system is greater than the actual fluid movement speed, there is a time difference in the response of various physical quantities when a transient disturbance occurs.

[0122] Based on this, by reading the timestamps when each abnormal feature criterion is generated, and establishing the state transition sequence of the multimodal signal in the current unified control cycle and its preceding historical cycles, a temporal basis can be provided for the causal verification in the following text.

[0123] Step 302: Based on the prior physical sequence of the water conveyance system, the trigger sequence is compared, and for abnormal feature criteria that do not meet the causal preconditions, the corresponding evidence source weight label is marked as a desensitized observation state.

[0124] Among them, the a priori physical sequence represents the co-occurrence relationship of physical phenomena determined by the continuity equation and momentum equation in fluid mechanics.

[0125] For example, a water hammer event can manifest as a sharp fluctuation in pipeline pressure and exhibit a predetermined high-frequency resonance in the acoustic signal.

[0126] If the current comparison results show that the water hammer spectrum criterion is triggered alone, and no pressure change occurs within the preceding time window, then the triggering sequence violates the causal preconditions of fluid mechanics.

[0127] It should be understood that, for such isolated triggering or time-reversed abnormal feature criteria, the corresponding evidence source weight label can be set to a desensitized observation state.

[0128] Step 303: For abnormal feature criteria that meet the causal preconditions, mark the corresponding evidence source weight label as either priority confirmation state or normal fusion state.

[0129] For example, when the order of triggering conforms to the deductive order of the physical mechanism, the system can assign higher computational weights to the corresponding feature criteria.

[0130] Correspondingly, when the pressure mutation criterion has been triggered and the water hammer spectrum criterion has exceeded the judgment threshold, this phenomenon conforms to the physical precondition logic of pressure first and then spectrum. The update operation marks the evidence source weight label corresponding to the water hammer spectrum criterion as a priority confirmation state.

[0131] For other synchronously triggered signals that do not have a clear physical exclusion relationship, their evidence source weight labels are marked as normal fusion state.

[0132] This step allows for the interception of spurious signals caused by high-frequency noise from a single sensor or localized sporadic disturbances before probability calculations are performed.

[0133] In some embodiments, extracting traffic anomaly criteria includes:

[0134] Analyze the temporal variation patterns of traffic forecast residuals across multiple sliding windows;

[0135] If the time-series change pattern exhibits a step-like abrupt change exceeding the step threshold, then output an anomaly criterion characterizing water load fluctuations.

[0136] If the time-series change pattern shows a slow and continuous shift that accumulates beyond the offset threshold, and is accompanied by abnormal temperature signals in the multimodal monitoring signals of a local pipe section, then an anomaly criterion characterizing a suspected leak will be output.

[0137] In some scenarios, anomaly refers to deviation from a preset benchmark, and anomaly criteria are used to characterize leakage features.

[0138] If the time-series variation pattern does not meet any of the above conditions, then output the criteria representing the normal state of the flow.

[0139] Furthermore, for causal verification of the traffic flow signal, the temporal variation patterns of the traffic flow prediction residuals within multiple sliding windows are analyzed and differentiated.

[0140] As an example, if the time-series change pattern is characterized by a step change in the magnitude of the flow prediction residual exceeding a preset step threshold at the boundary of a single sliding window, it is determined to correspond to a sudden switch in water load, such as the instantaneous start-up and shutdown of a large downstream user, and an abnormal criterion representing the fluctuation of water load is output.

[0141] The step threshold can be determined by taking the mean value plus a preset standard deviation of the residual amplitude corresponding to the normal load and switching event in the historical operation record of the pipeline network. It can be adaptively adjusted according to the actual pipeline network operation data.

[0142] In another example, if the time-series change pattern is characterized by a slow and continuous offset that accumulates beyond the offset threshold and is accompanied by an asymmetrical distribution of local pipe segment temperature signals, it is determined that it does not meet the physical characteristics of water hammer fluctuations. An abnormal criterion representing suspected leakage is output, and its influence weight in transient water hammer determination is limited.

[0143] Another example is that the offset threshold can be determined based on the cumulative statistical deviation baseline of the flow prediction residual under normal operating conditions of the pipeline network, and can be set according to the leakage sensitivity requirements of the actual pipeline section.

[0144] Furthermore, the abnormal feature criteria and the corresponding evidence source weight labels are input into the post-evidence theory fusion layer to perform multi-source evidence synthesis calculation and output the abnormality judgment conclusion.

[0145] In some embodiments, the post-evidence theory fusion layer performs multi-source evidence synthesis computation, including:

[0146] Furthermore, the numerical statistics of the anomaly feature criteria are extracted, and the numerical statistics of the anomaly feature criteria are mapped to the initial confidence level using a nonlinear distribution function.

[0147] Here, the statistics correspond to the dimensionless calculation results output from each feature extraction stage, such as the prediction residual after normalization or the energy proportion value of a predetermined frequency band. To convert the statistics of the open interval into quantifiable confidence levels conforming to a probabilistic framework, a distribution function is used for calculation. The transformation relationship can be described as follows:

[0148] φ=1 / (1+exp(-1×λ×(s-τ)));

[0149] Where φ is the initial confidence level, exp is the exponential function with the natural constant as the base, λ is the mapping steepness parameter, s is the statistic of the anomaly feature criterion, and τ is the preset judgment center threshold.

[0150] By configuring the value of the mapping steepness parameter, the numerical sensitivity range of the feature signal when it transitions from a normal state to an abnormal state can be controlled, so that the initial confidence of the final output is strictly distributed within the closed interval of 0-1.

[0151] Furthermore, based on the initial confidence level, basic probability assignment values ​​are assigned to the preset set of anomaly type hypotheses.

[0152] The preset set of anomaly types includes a set of specific fault classifications that may occur in the water supply network, specifically including subsets of water hammer events, subsets of valve malfunction events, subsets of leakage events, and subsets of normal states.

[0153] Considering the physical perception limitations of a single-dimensional sensor, the computing unit can divide and distribute the calculated initial confidence level into subsets of each hypothesis set according to a preset ratio based on the physical sensitivity characteristics of each sensor.

[0154] For example, pressure features have a strong ability to detect both water hammer events and valve malfunctions. Therefore, their initial confidence levels will be primarily assigned to a joint subset covering both hypotheses, thus forming the basic probability allocation value of the predetermined evidence sources within the current decision period.

[0155] Furthermore, for the abnormal feature criteria of the corresponding evidence source weight label being marked as the desensitized observation state, a preset discount factor is introduced to proportionally reduce the corresponding basic probability allocation value, and the reduced and stripped probability quality is transferred to the set of uncertain terms.

[0156] For evidence sources in a desensitized observation state, their data input is not erased. Instead, their proportion in multi-source fusion computing is reduced through dimensionality reduction. The specific formula for this proportional reduction is as follows:

[0157] m _δ (S _i )=δ×m _origin (S _i);

[0158] Where, m _δ (S _i ) represents the basic probability allocation value after proportional reduction, δ represents the preset discount factor, and m _origin (S _i S represents the basic probability assignment value of the original extraction. _i This is a subset of specific physical faults within the set of anomaly type hypotheses.

[0159] At the same time, in order to ensure that the sum of the probability mass conservation within the evidence theory framework is 1, the reduced and stripped values ​​must be transferred, and the transfer operation logic is as follows:

[0160] m _δ (Θ) = 1 - δ × φ;

[0161] Where, m _δ (Θ) represents the total probability mass of the set of uncertainties after discounting, δ is the preset discount factor, φ is the initial confidence level of the desensitized evidence source, and Θ represents the complete set of uncertainties in the identification framework representing the unknown state.

[0162] In this embodiment, the discount factor can be set to 0.4.

[0163] Furthermore, the evidence synthesis rule is applied to each of the processed basic probability allocation values, and the result with the highest probability after synthesis is extracted as the anomaly determination conclusion.

[0164] In this embodiment, the system inputs all valid evidence source data, including those verified by the gating layer and with discount correction, into the standard evidence theory synthesis module, and performs orthogonal multiplication and normalization processing.

[0165] For example, in the water hammer determination scenario, if the statistical quantity of the pressure feature criterion far exceeds the determination center threshold, and the initial confidence obtained after mapping by the sigmoid activation function is close to the full value, while the acoustic spectrum criterion is confirmed as a normal fusion state in the causal verification, and the flow evidence source is marked as a desensitized observation state because its triggering sequence violates the hydraulic causal pre-conditioning logic, then a discount factor is introduced to the flow evidence source to reduce the probability quality, and the reduced part is transferred to the set of uncertain terms.

[0166] After the evidence synthesis rule is calculated, the combined direction of the two high-weighted evidences, pressure and acoustics, makes the water hammer event hypothesis have the highest synthesis probability. Based on this, the anomaly judgment conclusion is output as water hammer event with a confidence level of high credibility.

[0167] In this embodiment, a pre-gating constraint mechanism based on fluid dynamics causality is introduced. By comparing the signal triggering sequence before evidence synthesis, dynamic discounting and quality transfer are performed on desensitized evidence that violates physical common sense. This can suppress computational distortion caused by high-conflict information sources and improve the system's anti-disturbance identification accuracy from a purely data-driven level to a physically and logically self-consistent level.

[0168] In some embodiments, the specific steps of performing time-domain state prediction on the first type of node with monitoring equipment based on the anomaly determination conclusion are further described, specifically including:

[0169] Step 401: Obtain the pre-configured hydraulic parameters of the pipeline network topology;

[0170] Among them, the hydraulic parameters of the pipeline network topology, which are the basic static data describing the physical properties and geometric dimensions of the water conveyance system, can be obtained by reading a pre-built equipment attribute database. Specifically, they include the actual length, inner diameter cross-sectional area, absolute roughness, and elastic modulus of each pipe segment, as well as the standard density of the fluid and the pressure wave propagation velocity in the predetermined pipe material.

[0171] It should be understood that this parameter forms the data basis for transforming fluid dynamics partial differential equations into discrete state-space models that can be computed in real time.

[0172] Step 402: Construct a discrete state-space prediction model based on the hydraulic parameters of the pipeline network topology. The discrete state-space prediction model includes a state transition matrix and an input control matrix.

[0173] In other embodiments, a discrete state-space prediction model is constructed based on the hydraulic parameters of the pipeline network topology. The discrete state-space prediction model includes a state transition matrix, an input control matrix, and an initial state vector.

[0174] Among them, the discrete state-space prediction model is used to extrapolate the dynamic trajectory of flow and pressure changes over time for the first type of node. This model embeds the physical mechanism into the matrix structure.

[0175] Among them, the state transition matrix represents the spontaneous evolution law and spatial coupling relationship of the internal pressure and flow state of the system under the condition of no external intervention;

[0176] The input control matrix represents the forced excitation effect of external actions, such as adjusting valve opening or changing pump station speed, on the system state.

[0177] Furthermore, a discrete state-space prediction model is constructed based on the hydraulic parameters of the pipeline network topology, including:

[0178] Optionally, based on the hydraulic parameters of the pipeline network topology, the pipeline flow inertial parameters, linearized friction parameters, and nodal elastic stiffness parameters characterizing the changes in fluid state are calculated respectively.

[0179] In this embodiment, the lumped parameter method can be used to reduce the dimensionality of continuous fluid distribution parameters. The specific calculation process can be achieved using the following formula:

[0180] α _e =A _pe / (ρ×L _e );

[0181] Where, α _e Corresponding to the inertial parameters of the pipeline flow, A _pe The cross-sectional area of ​​the corresponding pipe section, ρ corresponds to the fluid density, and L _e Corresponding pipe segment length.

[0182] β _e =(f _e ×abs(Q _e )) / (D _e ×A _pe );

[0183] Where, β _e Corresponding linearized friction parameter, f _e Corresponding to the Darcy friction coefficient, abs() corresponds to the absolute value operation, Q _e Corresponding to the initial steady-state flow rate, D _e Corresponding to the inner diameter of the pipe section, A _pe The cross-sectional area of ​​the corresponding pipe section.

[0184] κ _j =(ρ×(a _j ) 2 ) / V _j ;

[0185] Among them, κ _j The corresponding nodal elastic stiffness parameter, ρ corresponds to the fluid density, and a _j The pressure wave propagation velocity at this node is V. _j The equivalent volume of the control body corresponding to this node.

[0186] Optionally, the unit eigenvalue can be configured as a pressure self-transfer block module in the state transition matrix;

[0187] For example, the state transition matrix is ​​divided into several functionally distinct sub-block regions. The pressure self-transfer block module corresponds to the region in the upper left corner of the matrix, representing the inertial retention capability of the pressure state of each node.

[0188] For example, the computing unit constructs a diagonal matrix with the same dimension as the number of nodes of the first type, and configures all the elements on the diagonal to 1 as the input data of the block module. This reflects that within a shorter control cycle, the pressure baseline value of the previous moment will be fully transmitted to the current moment.

[0189] Optionally, in conjunction with a preset control cycle, the node elastic stiffness parameters are mapped to the flow-to-pressure block module in the state transition matrix, and the pipeline flow inertia parameters are mapped to the pressure-driven flow block module in the state transition matrix.

[0190] Among them, the preset control cycle refers to the time step parameter of the discretization calculation;

[0191] The flow-to-pressure block module is used to quantify the magnitude of pipeline pressure rise and fall caused by the net inflow or outflow of a node over time.

[0192] The pressure-driven flow segmentation module is used to quantify the contribution of the pressure difference between the two ends of the pipe section as the driving force to the flow change caused by the acceleration of the fluid inside the pipe section under time integration.

[0193] For example, the elastic stiffness parameter of the unit calculation node is multiplied by the preset control period, and combined with the preset direction sign coefficient, and then filled into the flow-to-pressure block module in the upper right corner of the matrix.

[0194] Simultaneously, the inertial parameters of the pipeline flow obtained by the unit are multiplied with the preset control cycle, and the product is configured in the pressure-driven flow segmentation module in the lower left corner of the matrix.

[0195] Optionally, the flow self-attenuation block module in the state transition matrix is ​​configured based on the linearized friction parameters;

[0196] Among them, the flow self-decay block module corresponds to the lower right corner of the state transition matrix and is used to describe the dissipation process of kinetic energy of the fluid under the action of pipe wall friction resistance.

[0197] The specific operation is to calculate the difference between 1 and the product of the linearized friction parameter and the preset control period, and then use this difference as a diagonal element to form a local diagonal matrix and configure it into the block module.

[0198] It should be understood that with this configuration, the model can accurately simulate the natural attenuation characteristics of fluid fluctuations during propagation.

[0199] Optionally, assembling each modular unit generates a state transition matrix with hydraulic physical constraints;

[0200] The calculation program concatenates the initialized pressure self-transfer module, flow-to-pressure module, pressure-driven flow module, and flow self-attenuation module according to the order of variables in the state vector.

[0201] Based on this, the combined large matrix obtained by splicing is the state transition matrix. All elements inside this matrix are derived from parameters with clear physical dimensions, which are used to make subsequent calculations conform to the basic laws of the one-dimensional water hammer physical equation.

[0202] For example, the input control matrix can be constructed in the following ways:

[0203] Based on the excitation response relationship between the actuators (including valves, pumping stations, etc.) in the pipeline network and the flow rate and node pressure of adjacent pipe sections, the control gain of each actuator is mapped to the matrix column corresponding to the dimension of the state vector.

[0204] For example, the matrix can be constructed based on the standard modeling method of linear state-space model in control system theory, combined with the physical installation location and hydraulic characteristic parameters of each actuator.

[0205] In other embodiments, an input control matrix that characterizes the excitation effect of external control actions on the system state can be constructed based on the cross-sectional area of ​​pipe segments in the hydraulic parameters of the pipeline network topology and the preset control period.

[0206] In some other embodiments, the feature line method can also be introduced as an offline evaluation tool to periodically verify and numerically correct the accuracy of the transition matrix using historical datasets.

[0207] Step 403: Update the initial state vector of the discrete state-space prediction model based on the multimodal monitoring signal;

[0208] In this embodiment, the initial state vector corresponds to the initial boundary conditions for multi-step deduction.

[0209] For example, the system acquires the actual pressure and flow data parsed from the sensor at the current moment, and calculates the Kalman gain using the standard Kalman filter update criterion.

[0210] Next, using the calculated Kalman gain and actual observation data, the prior state estimate output at the previous time step is numerically corrected to obtain a high-precision posterior state estimate at the current time step, and this estimate is then updated and overlaid on the initial state vector.

[0211] Step 404: Use the state transition matrix and the input control matrix to perform multi-step iterative deduction on the initial state vector, and output the multi-step predicted state distribution of the first type of node in the future time period.

[0212] In other embodiments, the state transition matrix, the input control matrix, and the preset boundary control variables can be used to perform multi-step iterative deduction on the updated initial state vector to output the multi-step predicted state distribution of the first type of node in the future time period.

[0213] Among them, multi-step iterative deduction belongs to a time-based cyclic operation process.

[0214] For example, the currently updated initial state vector is extracted, multiplied by the state transition matrix, and then accumulated with the product of the input control matrix and the preset boundary control variables to obtain the predicted state for the first time step in the future.

[0215] Next, the predicted state is used as a new input variable, and the matrix multiplication and accumulation operations are repeated. The preset deduction step threshold is then used to obtain a series of state vector sequences at discrete time points.

[0216] It should be understood that this sequence constitutes a multi-step predictive state distribution, which can provide data support for identifying whether the pipeline network will experience pressure over-limit risks in the future.

[0217] In some embodiments, an optional implementation method for performing spatial topology simulation on second-type nodes without monitoring equipment to obtain the network-wide risk assessment results is described, namely:

[0218] Step 501: Obtain the predicted state of the first type of node produced by performing time-domain state prediction on the first type of node with monitoring equipment.

[0219] In other words, obtain the predicted state produced by performing time-domain state prediction on the first type of node.

[0220] In some embodiments, the predicted state of the first type of node, i.e. the multi-step prediction data output by the preceding processing module in the time dimension, specifically includes the pressure prediction sequence and flow prediction sequence of the first type of node in multiple future control cycles.

[0221] Since physical disturbances in a pipeline system must propagate through a medium, the predicted state of this first type of node can be used to constitute the physical excitation source and dynamic boundary conditions for spatial topology deduction.

[0222] Based on this, prediction data with high confidence is extracted into a memory buffer to drive the state evolution calculation of areas without monitoring equipment.

[0223] Step 502: Construct a graph Laplace matrix based on the topological connection relationship and physical hydraulic impedance characteristics of the pipeline network nodes.

[0224] In this embodiment, the water supply network can be abstracted as an undirected weighted graph, where nodes correspond to branch points or ends of the network, and edges correspond to actual physical pipe segments. The computational unit defines the edge weights of the graph using the physical hydraulic impedance characteristics of each pipe segment, and the computational relationship is as follows:

[0225] w_ij =a _ij / l _ij ;

[0226] Among them, w _ij Let a be the edge weight of the pipe segment between node i and node j. _ij Let l be the pressure wave velocity within this pipe section. _ij This is the physical length of the pipe section.

[0227] Based on this, the difference between the diagonal matrix and the adjacency matrix is ​​calculated to obtain the graph Laplace matrix. This graph Laplace matrix characterizes the spatial connectivity of the pipe network and the time delay characteristics of pressure wave propagation in different pipe segments, and its off-diagonal elements reflect the hydraulic coupling strength between adjacent nodes.

[0228] Step 503: Use the graph Laplace matrix to propagate the predicted state of the first type of nodes along the pipeline topology to the second type of nodes, and combine the pressure-flow coupling relationship of the pipeline segment to calculate the flow rate of the pipeline segment, so as to summarize and generate the risk assessment results of the entire network.

[0229] In this embodiment, after obtaining the estimated pressure of all second-type nodes, the flow rate change trend of the pipe section without flowmeters can be inferred by using the equivalent relationship of the momentum equation in fluid mechanics.

[0230] Next, the calculation unit extracts the pressure difference between adjacent nodes, subtracts the friction loss estimated based on the empirical formula for flow velocity, and divides it by the comprehensive hydraulic impedance of the pipe section to obtain the estimated flow rate of the pipe section.

[0231] Based on this, the predicted state of the first type of node and the inferred state of the second type of node are spatiotemporally assembled to generate a pressure and traffic data field covering all nodes, which serves as the basic input for the network-wide risk assessment results.

[0232] The summary and generation of the overall network risk assessment results includes:

[0233] Optionally, for each node in the entire network, the probability of overpressure exceeding the safe upper limit and the probability of negative pressure falling below the safe lower limit are calculated based on the estimated state.

[0234] Specifically, the system can pre-set upper limits for pipe wall pressure resistance and lower limits for cavitation prevention at nodes with different pipe materials and elevations. The calculation unit further substitutes the mean and variance of the node-estimated pressure sequence into the Gaussian probability density function to calculate the integral area where the instantaneous pressure value exceeds the upper limit and the integral area where it falls below the lower limit in the future time period, thereby outputting quantified overpressure risk probability and negative pressure risk probability indicators.

[0235] Optionally, based on the physical structure bearing capacity preference of the water transmission network, differentiated assessment weights are assigned to the overpressure risk probability and the negative pressure risk probability, and the comprehensive risk index of each node in the entire network is calculated by combining the truncation function.

[0236] In this embodiment, an asymmetric risk measurement logic may be used to calculate the comprehensive risk index, namely:

[0237] R _total =f _trunc (ω _1 ×P _over +ω _2 ×P _under );

[0238] Among them, R _total Based on comprehensive risk indicators, the data will be directly sent to the linkage control module as the basis for decision-making regarding the activation of flexible pressure relief and valve adjustment. _over P represents the probability of overpressure risk. _under ω represents the probability of negative pressure risk. _1 As the first assessment weight assigned to overpressure risk, ω _2 To assign a second assessment weight to the negative pressure risk, f _trunc This is a preset numerical truncation function used to rigidly constrain the weighted summation value within a closed interval of 0-1.

[0239] In some scenarios, for pipe sections with significant terrain elevation differences, the value of the second evaluation weight can be dynamically increased to enhance the alarm sensitivity to water column separation phenomena.

[0240] Furthermore, the hydraulic lumped parameters are precisely mapped to a block matrix in the discrete state space, and spatial topology derivation is performed by combining the graph Laplace diffusion operator and matrix exponentiation.

[0241] It should be understood that this architecture, under the premise that the calculation process strictly follows the laws of one-dimensional hydraulics, achieves millisecond-level low-latency simulation of the global pressure distribution of the pipeline network.

[0242] According to one aspect of this application, the predicted state of a first-type node is diffused along the network topology to a second-type node using a graph Laplacian matrix, including:

[0243] Furthermore, the pre-configured propagation attenuation coefficient is obtained.

[0244] Before entering the spatial diffusion calculation, the online simulation system retrieves the propagation attenuation coefficients generated offline from the parameter configuration library.

[0245] The coefficient above serves as a dimensionless control parameter, used to constrain the diffusion domain of the graph Laplace matrix in the mathematical model, preventing the predicted energy from being amplified indefinitely in the undirected graph in a non-physical manner.

[0246] Furthermore, the diffusion operator matrix is ​​constructed by combining the propagation attenuation coefficient with the Graph Laplace matrix.

[0247] In this embodiment, the difference operation is performed on the product of the identity matrix and the propagation attenuation coefficient with the graph Laplace matrix. The resulting difference matrix is ​​the diffusion operator matrix, which is equivalent to the state transition operator of the discretized one-dimensional water hammer equation in the spatial domain. It can describe the evolution law of the single-step transition of the node state to its first-order neighbor node.

[0248] Furthermore, using the predicted state of the first type of node as the initial boundary for deduction, the deduced state of the second type of node is calculated by using the matrix exponentiation operation of the diffusion operator matrix corresponding to the topological distance of the target node.

[0249] In this embodiment, the inference calculation can be performed using graph algebra operations to achieve dimensionality reduction analysis. That is, the computing unit determines the topological distance constant between the second type of node to be tested and the first type of node, and uses this topological distance constant as an exponent to perform matrix exponentiation on the diffusion operator matrix.

[0250] Furthermore, the system directly applies the higher-order operators obtained from the exponentiation operation to the predicted state matrix of the first type of nodes, thereby calculating the stress projection state of the second type of nodes at the corresponding prediction time. This operation folds the multi-step spatial transfer into a single matrix multiplication, which helps reduce the online computational load.

[0251] According to one aspect of this application, the pre-configured propagation attenuation coefficient can be pre-built through an offline calibration process, which can be implemented using the following steps 1-3:

[0252] Step 1 involves obtaining known disturbance events from historical pipeline operation records, which is the initial extraction step in the offline parameter calibration process.

[0253] Among them, known disturbance events correspond to real hydraulic transient processes with complete upstream and downstream signal records stored in the historical operation database, such as planned pump shutdown operations or large-span adjustment operations of main valves.

[0254] Next, the system retrieves the time series of multimodal monitoring signals at the time of the event by searching historical logs, and uses it as a reference sample for parameter inversion.

[0255] Step 2: Extract the actual pressure attenuation magnitude and propagation topology distance of known disturbance events between upstream and downstream monitoring nodes.

[0256] Accordingly, extreme value analysis is performed on the extracted known disturbance event time series to calculate the ratio of the pressure peak values ​​at the upstream excitation point and the downstream response point within the same disturbance period, which is used as the actual pressure attenuation amplitude.

[0257] Simultaneously, the graph structure model of the pipeline network is analyzed, and the shortest path hop count between the two monitoring nodes is calculated, which is used as the propagation topology distance.

[0258] In other embodiments, eigenvalue decomposition is further performed on the graph Laplacian matrix to extract effective eigenvalues ​​in the corresponding propagation direction.

[0259] Step 3: Based on the actual pressure attenuation amplitude, propagation topological distance, and effective eigenvalues ​​of the graph Laplace matrix, the propagation attenuation coefficient is calculated analytically in reverse.

[0260] In this embodiment, the model parameters can be back-calibrated using measured attenuation data, i.e.:

[0261] γ _p =(1-η 1 / h_0) / λ _eff ;

[0262] Where, γ _p Here, η is the calculated propagation attenuation coefficient, h is the measured actual pressure attenuation amplitude, and η is the propagation attenuation coefficient. _0 To propagate topological distance, λ _eff The effective eigenvalues ​​of the graph Laplacian matrix corresponding to the propagation direction.

[0263] Therefore, attenuation parameters that are adapted to the actual physical state of the current pipeline network can be obtained and stored in the system's parameter configuration library.

[0264] In one possible implementation, controlling the flexible emergency pressure relief device to perform a pressure relief action includes:

[0265] Step 601: When the risk assessment results of the entire network indicate that there is a risk of transient overpressure impact in the pipeline network, the flexible buffer medium cavity inside the flexible emergency pressure relief device is used to absorb the energy of the transient pressure wave through reversible elastic deformation.

[0266] Among them, the flexible emergency pressure relief device corresponds to the physical execution hardware installed at the pipeline node;

[0267] The flexible buffer medium cavity can be a closed capsule structure made of EPDM rubber, fluororubber or fiber-reinforced composite elastic material, and its interior is filled with a compressed gas medium with a preset initial pressure. The gas medium can be nitrogen or dry air.

[0268] When the high-frequency transient overpressure beam reaches this node, because the fluid is incompressible while the gas is compressible, the fluid inside the tube squeezes the capsule, causing it to undergo reversible elastic deformation with volume contraction.

[0269] Based on this, the deformation process does negative work on the intruding fluid volume, converting the kinetic energy of the transient water hammer into the potential energy of gas compression, thereby achieving a delay-free reduction of the pressure wave peak within the physical dead time before the mechanical valve opens.

[0270] Step 602: Based on the anomaly type and confidence level in the anomaly determination conclusion, control the adjustable pressure relief channel of the flexible emergency pressure relief device to perform graded actions or continuously adjustable pressure relief opening operations.

[0271] Among them, the adjustable pressure relief channel, namely the pilot-operated control valve driven by a servo motor or hydraulic push rod, can have its flow area adjusted linearly or discretely according to an external electrical signal.

[0272] Furthermore, the control unit receives the anomaly determination conclusion output by the preceding fusion module and extracts the anomaly type and confidence level value as input variables for the control command.

[0273] For example, when the anomaly type is confirmed as a water hammer event and the confidence level reaches a predetermined high threshold level, the control unit outputs a continuous large-opening analog control signal to drive the pressure relief channel to quickly release the fluid volume in the pipe to reduce the overall base pressure.

[0274] When the confidence level is at a predetermined intermediate threshold, the control unit issues discrete, graded action commands to control the valve to maintain a partial opening position, preventing excessive fluid release and secondary negative pressure in the pipeline network under uncertain information conditions.

[0275] In some scenarios, a high threshold level can correspond to a confidence level of not less than 0.85;

[0276] The corresponding threshold level corresponds to a confidence level within a closed interval of 0.6-0.85;

[0277] The threshold division interval can be flexibly adjusted according to the actual working conditions of the project.

[0278] Step 603: Based on the results of the overall network risk assessment, implement upper limit constraints and anti-shake control on the opening change rate and action holding time of the adjustable pressure relief channel;

[0279] Among them, the opening change rate corresponds to the change in the valve flow cross-sectional area per unit time;

[0280] Action hold time refers to the physical time span during which the valve remains stationary after reaching the target opening degree.

[0281] Accordingly, based on the pressure prediction sequence in the overall network risk assessment results, an upper limit is set for the opening change rate. The calculation unit subtracts the calculated target opening from the current opening. If the absolute value of this difference divided by the preset execution time results in a change rate that exceeds the upper limit constraint, the system forcibly extends the execution time parameter to reduce the mechanical action rate.

[0282] In other embodiments, the anti-shake control mechanism requires that after the valve executes any opening adjustment command, it must wait for a duration greater than or equal to the action holding time, during which the control unit blocks reverse adjustment commands to the valve.

[0283] In other embodiments, the action hold time is set to be greater than 1.5 times the round-trip cycle of the pressure wave in the longest physical branch of the pipeline, thereby isolating the resonant feedback between the local control command and the system's inherent fluid frequency.

[0284] Among these measures, coordinated control will be implemented on downstream pipe sections that have a hydraulic relationship with the abnormal nodes, including:

[0285] Optionally, the linkage priority for downstream pipe sections is determined based on the confidence level of the anomaly determination conclusion;

[0286] The downstream pipe section corresponds to the branch of the physical pipe located behind the flow direction of the abnormal trigger node in the spatial topology, and is determined to be within the influence domain of the pressure wave through the aforementioned graph Laplace matrix deduction.

[0287] Linkage priority is a scheduling parameter that characterizes the combined magnitude of intervention response speed and action amplitude.

[0288] Accordingly, the control unit extracts the final confidence value of the anomaly determination conclusion and maps it to a preset priority lookup table; the scheduling decision of the pipeline network is transformed from a single local node pressure relief to a network collaborative scheduling that includes spatial depth defense.

[0289] Optionally, when the confidence level is within a preset boundary range, flow restriction or pre-venting pressure reduction actions for downstream pipe sections are performed first.

[0290] In this embodiment, within a preset boundary range, it is determined that there is a physical disturbance in the pipeline network, but the specific type or destructive force of the abnormal event has not yet been confirmed.

[0291] Accordingly, the control system issues low-disturbance action commands to downstream pumping stations or flow control valves within the hydraulically correlated range. The flow restriction operation is achieved by reducing the preset proportion of the downstream branch flow velocity limit, while the pre-depressurization action involves opening the downstream air vent and sludge discharge valve to a slightly open state.

[0292] It should be understood that this operation is intended to deplete excess kinetic energy in the pipeline system in advance at the cost of minimal disruption to regular water supply, thereby reducing the pressure on the pipeline foundation when a potential main water hammer wave arrives.

[0293] Optionally, when the confidence level is higher than a preset threshold and the risk level increases, the linkage scope for downstream pipe sections is expanded step by step and flow redistribution is performed.

[0294] For example, the preset threshold can be configured to be 0.85.

[0295] When the confidence level exceeds the preset threshold and the comprehensive risk index output by the network-wide risk assessment results shows a monotonically increasing trend, it can be determined that a certain high-risk transient shock has occurred.

[0296] Furthermore, the control unit dynamically increases the propagation attenuation coefficient in the graph Laplace diffusion model, thereby increasing the number of high-risk impact nodes calculated in the deduction, and thus expanding the coverage of the linkage control command.

[0297] Furthermore, for downstream branches included in the new linkage scope, a forced flow redistribution is implemented by simultaneously shutting down the inlet valves of high-risk branches and opening the distribution valves of the safety bypass pipelines.

[0298] In this application, the reversible elastic deformation of the internal medium cavity is used to passively absorb transient wave peak energy, while simultaneously coupling an anti-shaking constraint based on the opening change rate based on the risk confidence interval and a hierarchical control strategy. This mechanism smooths out the secondary excitation of the pipeline system caused by the physical valve opening and closing process, upgrading single-point mechanical protection to a flexible active collaborative intervention network covering the upstream and downstream hydraulic topology.

[0299] According to one aspect of this application, the system of the present invention may be optionally installed in a water pipeline or a pressurized tunnel, comprising:

[0300] Pressure sensor 1, flow sensor 2, acoustic sensor 4 and temperature sensor 3 are installed on the pipeline to perform multimodal monitoring of the operating status of the water conveyance system.

[0301] A data acquisition and communication module that communicates with each sensor;

[0302] The data analysis and fusion unit is connected to the data acquisition module. The data analysis and fusion unit includes, in sequence: a stress mutation detection module, a multimodal data fusion module, and a risk assessment and prediction module.

[0303] A flexible emergency pressure relief device connected to the data analysis and fusion unit;

[0304] Downstream branch valve 5 control device that works in conjunction with the flexible emergency pressure relief device.

[0305] The data analysis and fusion unit sends control commands to the flexible emergency pressure relief device and downstream branch valve 5 based on the analysis results of the multimodal monitoring data, so as to realize the dynamic adjustment of the pressure status of the water transmission system.

[0306] For example, the anomaly detection and data fusion process of the present invention includes the following steps:

[0307] Collect multi-source monitoring data such as pressure, flow rate, acoustics, and temperature;

[0308] Sliding time window analysis is performed on pressure data to identify abrupt pressure events;

[0309] Frequency domain analysis of acoustic and pressure oscillation signals is performed to identify water hammer characteristics;

[0310] The Kalman filter method is used to predict traffic flow data and identify abrupt changes;

[0311] The anomaly criteria from different detection modules are fused together.

[0312] Determine the type of abnormal operating condition and its confidence level;

[0313] Predict the evolution trend and potential risks of abnormal operating conditions;

[0314] Pressure relief and linkage control commands are generated based on the risk assessment results.

[0315] For example, a monitoring point A is set up on the main water supply pipeline. A pressure sensor 1 and a flow sensor 2 are installed at monitoring point A and are connected in communication with the control system.

[0316] For example, when an abnormal pressure or flow rate is detected at monitoring point A:

[0317] The control system analyzes the pressure status of the main pipeline;

[0318] Based on the topology of the water supply network, identify multiple downstream branch pipes that have hydraulic relationships with monitoring point A;

[0319] Perform one or more of the following coordinated control operations on the downstream branch pipelines:

[0320] Adjust the opening degree of the branch pipe valve;

[0321] Start or adjust the pressure relief device at the branch pipe;

[0322] To allocate or limit the flow of branch pipes;

[0323] By coordinating the regulation of the main pipe and downstream branch pipes, the propagation and amplification of abnormal pressure can be suppressed.

[0324] For example, a flexible emergency pressure relief device is installed along a water pipeline and includes:

[0325] A shell structure connected to the water pipeline;

[0326] A flexible buffer medium cavity is set inside the shell to absorb transient pressure energy;

[0327] An elastic element or buffer assembly connected to a flexible buffer medium cavity;

[0328] An adjustable pressure relief valve is located on one side of the housing;

[0329] Pressure regulating mechanism connected to the pressure relief valve.

[0330] When the pressure inside the pipeline exceeds the set range, the flexible buffer medium cavity deforms to absorb some of the pressure energy; then the adjustable pressure relief valve opens step by step according to the control command to realize a continuously adjustable pressure relief process, thereby avoiding secondary water hammer caused by the traditional rigid pressure relief method.

[0331] In other embodiments, using the graph Laplacian matrix to propagate the predicted state of the first type of nodes to the second type of nodes along the network topology may also include:

[0332] Obtain the pre-configured propagation attenuation coefficient;

[0333] The diffusion operator matrix is ​​constructed by combining the propagation attenuation coefficient and the graph Laplace matrix;

[0334] Determine the topological distance between each target's second-type node and first-type node based on topological connectivity;

[0335] Using the predicted state of the first type of node as the initial boundary for deduction, the deduced state of the second type of node is calculated by using the matrix exponentiation operation of the diffusion operator matrix corresponding to the topological distance of the target node.

[0336] For example, the sampling rate of the acoustic signal is not less than twice the characteristic frequency of water hammer in the pipeline network. The specific rate can be adjusted according to the actual working conditions, and there is no single limitation on this. The same applies to other parameters and thresholds.

[0337] It should be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.

Claims

1. A flexible emergency pressure relief control method for water conveyance structures, characterized in that, include: Acquire multimodal monitoring signals from water conveyance structures; Extracting abnormal feature criteria from multimodal monitoring signals; Based on physical causal constraints, multi-source data fusion is performed on the anomaly feature criteria to obtain anomaly judgment conclusions that include anomaly type and confidence level. Based on the anomaly determination, time-domain state prediction is performed on the first type of nodes with monitoring equipment, and spatial topology deduction is performed on the second type of nodes without monitoring equipment to obtain the network-wide risk assessment results. Based on the anomaly determination conclusion and the network-wide risk assessment results, the flexible emergency pressure relief device is controlled to perform pressure relief actions, and coordinated control is implemented on downstream pipe sections that have a hydraulic relationship with the anomaly node.

2. The method of claim 1, wherein, Extracting abnormal feature criteria from multimodal monitoring signals, including: The sliding time window mechanism is used to detect the pressure change trend in multimodal monitoring signals and extract the pressure change criterion. Frequency domain transformation is performed on the multimodal monitoring signal, and the energy proportion of the characteristic frequency band is extracted to obtain the water hammer spectrum criterion; The traffic prediction residual is obtained by using a pre-built state tracking model, and traffic anomaly criteria are extracted.

3. The method of claim 1, wherein, In the process of fusing multi-source data based on physical causal constraints to obtain anomaly determination conclusions that include anomaly type and confidence level, a two-layer processing architecture is specifically adopted, including: The abnormal feature criteria are input into the preset front-end physical gate layer. By verifying the causal consistency between different abnormal feature criteria, the evidence source weight label corresponding to each abnormal feature criterion is output. The abnormal feature criteria and the corresponding evidence source weight labels are input into the post-evidence theory fusion layer to perform multi-source evidence synthesis calculation and output the abnormality judgment conclusion.

4. The method of claim 1, wherein, Based on the anomaly determination, time-domain state prediction is performed on the first type of nodes equipped with monitoring devices, specifically including: Obtain the pre-configured hydraulic parameters of the pipeline network topology; A discrete state-space prediction model is constructed based on the hydraulic parameters of the pipeline network topology. The discrete state-space prediction model includes a state transition matrix and an input control matrix. The initial state vector of the discrete state-space prediction model is updated based on multimodal monitoring signals; By using the state transition matrix and the input control matrix, a multi-step iterative deduction is performed on the initial state vector to output the multi-step predicted state distribution of the first type of node in the future time period.

5. The method of claim 1, wherein, Performing spatial topology simulation on Category II nodes without monitoring equipment yielded a network-wide risk assessment, specifically including: Obtain the predicted state of the first type of node, which is the output of time-domain state prediction for the first type of node with monitoring equipment; Based on the topological connection relationship and physical hydraulic impedance characteristics of the pipeline network nodes, a graph Laplace matrix is ​​constructed. The predicted state of the first type of node is propagated to the second type of node along the pipeline network topology using the graph Laplace matrix. The flow rate of the pipeline segment is calculated by combining the pressure and flow coupling relationship of the pipeline segment, so as to generate the risk assessment results of the whole network.

6. The method of claim 1, wherein, Criteria for extracting abnormal features from multimodal monitoring signals also include: Identify the current operating mode of the water supply network; Based on the operating condition mode, the sensitivity parameters of the criteria for each dimension of multimodal monitoring signals are adjusted collaboratively. The feature extraction window update points of each multimodal monitoring signal are aligned with a preset unified control cycle, so that the judgment indicators of different dimensions are output synchronously as abnormal feature criteria within the same time window.

7. The method of claim 1, wherein, Controlling the flexible emergency pressure relief device to perform pressure relief actions includes: When the risk assessment results of the entire network indicate that there is a risk of transient overpressure impact in the pipeline network, the flexible buffer medium cavity inside the flexible emergency pressure relief device is used to absorb the energy of the transient pressure wave through reversible elastic deformation. Based on the anomaly type and confidence level in the anomaly determination conclusion, the adjustable pressure relief channel of the flexible emergency pressure relief device is controlled to perform graded actions or continuously adjustable pressure relief opening operations. Based on the results of the overall network risk assessment, upper limit constraints and anti-shake control are implemented on the opening change rate and action holding time of the adjustable pressure relief channel.

8. The method of claim 2, wherein, Extract traffic anomaly criteria, including: Analyze the temporal variation patterns of traffic forecast residuals across multiple sliding windows; If the time-series change pattern exhibits a step-like abrupt change exceeding the step threshold, then output an anomaly criterion characterizing water load fluctuations. If the time-series change pattern shows a continuous offset that accumulates beyond the offset threshold, and is accompanied by abnormal temperature signals in the multimodal monitoring signals of a local pipe section, then an abnormality criterion characterizing leakage will be output.

9. A flexible emergency pressure relief control system for water conveyance structures, characterized in that, The method for implementing any one of claims 1 to 8 specifically includes: The signal acquisition module is used to acquire multimodal monitoring signals from water conveyance structures; The feature extraction module is used to extract abnormal feature criteria from multimodal monitoring signals; The data fusion module is used to perform multi-source data fusion on anomaly feature criteria based on physical causal constraints to obtain anomaly judgment conclusions that include anomaly type and confidence level. The risk assessment module is used to perform time-domain state prediction for the first type of nodes with monitoring equipment and spatial topology deduction for the second type of nodes without monitoring equipment based on the anomaly determination conclusion, so as to obtain the risk assessment results of the entire network. The linkage control module is used to control the flexible emergency pressure relief device to perform pressure relief actions based on the anomaly judgment conclusion and the risk assessment results of the entire network, and to implement linkage control on downstream pipe sections that have hydraulic correlation with the abnormal node.

10. A processing apparatus, characterized in that, It includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the method of any one of claims 1 to 8.

Images