A smart early warning system for fire hydrant monitoring based on anomaly analysis technology

Abstract

This invention relates to the field of monitoring and early warning technology, specifically to an intelligent early warning system for fire hydrants based on anomaly analysis technology. The system includes a flow velocity anomaly identification module, a node collaborative differential pressure verification module, a differential pressure trend independence discrimination module, a node degradation feature extraction module, and a risk level generation module. This invention, through the correlation judgment of flow velocity deviation and control signals, identifies the absence of signals in abnormal water intake behavior. Combined with a collaborative analysis mechanism of pressure changes in adjacent nodes, it can accurately identify hydraulic disturbance areas under non-human control. By analyzing the spatial independence of pressure response trends, and extracting the functional degradation level of the device based on historical data of opening / closing delay and response performance, it enhances the quantitative assessment capability of node functional status. When issuing risk warnings, it possesses the ability to finely classify risk levels, improving the accuracy of risk identification and enhancing the proactive detection capability of early-stage potential faults.

Description

Technical Field

[0001] This invention relates to the field of monitoring and early warning technology, and in particular to an intelligent early warning system for fire hydrant monitoring based on anomaly analysis technology. Background Technology

[0002] The field of monitoring and early warning technology primarily involves real-time monitoring of the operational status of the environment, facilities, equipment, or systems, and performing dynamic analysis based on the collected data to promptly identify potential anomalies or hazards. This technology typically includes sensor deployment, data acquisition modules, data communication networks, signal processing algorithms, status assessment models, and early warning decision-making mechanisms. Its aim is to issue early warning signals before risks occur, guiding timely intervention by response mechanisms. Applications cover multiple areas such as urban infrastructure safety, industrial equipment operation and maintenance, public safety, and natural disaster prevention. The implementation of this technology emphasizes real-time performance, high reliability, automated analysis capabilities, and scalability, often involving specific methods such as distributed node collaboration, edge computing, and fault diagnosis models.

[0003] Among them, the intelligent fire hydrant monitoring and early warning system is an intelligent monitoring system for urban fire protection facility management. It aims to monitor the operational status of fire hydrants in real time and automatically trigger early warning signals when abnormal use, abnormal water pressure, abnormal leakage, or unauthorized opening are detected. The system can be used for the safety assurance of urban fire protection infrastructure. By integrating multiple monitoring devices and intelligent analysis units, it improves the visualization management level and response efficiency of fire protection resources, effectively preventing problems such as water waste, delayed handling of equipment failures, and human-caused damage, thereby enhancing the integrity and intelligence of the urban public safety operation system.

[0004] Traditional early warning systems rely on multiple monitoring devices to trigger responses to abnormal situations, lacking a mechanism for comparing and analyzing flow velocity and control signals. This makes it difficult to detect false alarms caused by abnormal flow velocity without human intervention. In actual operation, the decline in response capability due to equipment aging is not dynamically assessed, causing fault signals from some nodes to be misinterpreted as environmental disturbances. Traditional systems lack a pressure coordination judgment process between nodes, making it difficult to eliminate misidentification caused by single-point fluctuations. When multiple nodes are simultaneously experiencing pressure differential changes, traditional systems lack the ability to distinguish the propagation path of hydraulic fluctuations from sudden leakage nodes, affecting the accuracy of anomaly location and the timeliness of response. This can easily lead to hidden dangers not being identified in advance, thereby exacerbating the risk of accidents. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose an intelligent early warning system for fire hydrant monitoring based on anomaly analysis technology.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent early warning system for fire hydrant monitoring based on anomaly analysis technology, the system comprising:

[0007] The flow velocity anomaly identification module acquires the flow velocity monitoring data of the fire hydrant monitoring node, compares the current flow velocity value of the fire hydrant monitoring node with the median value, calculates the deviation amplitude value, determines whether the deviation amplitude exceeds the single-point anomaly identification threshold and no control operation signal appears, and generates a flow velocity offset label.

[0008] The node collaborative differential pressure test module calls the flow velocity offset tag to obtain the pressure change rate of each node within the sampling period and compares it with abnormal nodes. It then filters out node groups with inconsistent directions or abrupt changes to obtain the flow velocity offset differential pressure response area.

[0009] The pressure difference trend independence discrimination module extracts the directional derivative sign sequence of water pressure change in each group based on the flow velocity offset pressure difference response area, compares the change direction of abnormal nodes with neighboring nodes, marks the spatial independence abrupt change behavior, and obtains the pressure difference abrupt change node set;

[0010] The node degradation feature extraction module calls the differential pressure change node set, calculates the average drift value of opening and closing time and the change rate of response delay based on the operation and maintenance data of each node, classifies the current functional degradation level of each fire hydrant node, and obtains a functional degradation status label set.

[0011] The risk level generation module, based on the functional attenuation state label set and the differential pressure change node set, and combined with trend characteristics, makes a graded judgment, classifies the risk warning level, implements fire hydrant risk warning for management personnel, and obtains real-time fire hydrant risk warning information.

[0012] The present invention improves upon the following: the flow velocity deviation label includes an abnormal node number for the flow velocity deviation magnitude, a time period identifier for which no control signal was triggered, and a high deviation probability time marker; the flow velocity deviation differential pressure response region includes a combination of nodes with inconsistent instantaneous differential pressure direction, a time synchronization conflict marker, and a local response mismatch region number; the differential pressure change node set includes a central node pressure drop abnormal attribute marker, a description of the non-response behavior of adjacent nodes, and a trend direction inconsistency index; the functional degradation status label set includes a structural functional level label, a performance degradation behavior type, and a segmented description of remaining response capacity; and the fire hydrant real-time risk warning information specifically includes a functional lower limit conflict level, a trend change intensity level, and a node warning level determination identifier.

[0013] The present invention is improved in that the flow velocity anomaly identification module includes:

[0014] The flow velocity baseline extraction submodule acquires the flow velocity monitoring data of the fire hydrant monitoring nodes, and extracts the flow velocity value of each node within the target time range at the current time point in combination with the corresponding sampling time point. It also extracts the median and standard deviation to obtain stable flow velocity statistical indicators.

[0015] The flow velocity deviation calculation submodule, based on the stable flow velocity statistical index, calls the flow velocity value corresponding to each node at the current time point, calculates the ratio of the absolute value of the deviation of the flow velocity value from the median value to the standard deviation of the fluctuation, calculates the deviation judgment value of the current node within the statistical window, calculates the single-point anomaly identification threshold based on the stable flow velocity statistical index, determines whether the deviation judgment value is greater than the single-point anomaly identification threshold, marks the anomaly, and obtains the suspicious flow velocity deviation identification group.

[0016] The control behavior exclusion submodule determines whether there is a fire hydrant valve control signal at the corresponding time point based on the suspicious flow rate deviation identifier group, filters out the time period data with control signal behavior, retains the deviation identifier without valve control, and generates flow rate deviation tags.

[0017] The present invention is improved in that the node collaborative differential pressure testing module includes:

[0018] The adjacent node extraction submodule calls the flow velocity offset tag, indexes the pipe segment structure diagram corresponding to the identified abnormal node number according to the abnormal node number, obtains the adjacent connected monitoring nodes, extracts the pressure data sequence of each adjacent node in the current sampling period, calculates the unit length pipe segment pressure difference value between the adjacent node and the abnormal node, and obtains the adjacent node pressure difference change group.

[0019] The pressure rate comparison submodule extracts the pressure change sequence of each node within the current sampling period based on the pressure difference change group of adjacent nodes, calculates the instantaneous pressure change rate of each sequence, and judges the difference ratio with the pressure change rate of the corresponding abnormal node. It marks whether it exceeds the instantaneous pressure rate difference threshold to obtain the pressure difference rate difference value group.

[0020] The gradient direction filtering submodule extracts the pressure difference change direction identifier of each node at the current time point based on the pressure difference rate difference value group, performs a consistency matching operation on the direction identifier, determines whether there are cases where the direction of adjacent nodes is inconsistent with or opposite to the direction of abnormal nodes, filters the node combination with contradictory direction characteristics, and generates the flow velocity offset pressure difference response region.

[0021] The present invention is improved in that the differential pressure trend independence discrimination module includes:

[0022] The initial feature extraction submodule obtains the location index of the first continuous decrease in water pressure value of each abnormal node within the sampling period based on the flow velocity offset pressure difference response region, and extracts multiple monitoring nodes adjacent to the abnormal node. It also extracts the first response change time index of each adjacent node within the same time segment, calculates the time difference between adjacent nodes and the abnormal node, and generates response time difference information.

[0023] The trend response calculation submodule calls the response time difference information, calculates the differential value sequence of water pressure change of nodes in a continuous period based on the water pressure sampling value corresponding to each group of nodes, extracts the direction of the differential sequence, marks the direction symbols to form the change direction combination between nodes, judges the synchronicity of the direction combination, and obtains the directional synchronicity feature information.

[0024] The independence screening and determination submodule determines whether the direction of water pressure change of the abnormal node conflicts with or is missing from the direction of the adjacent node based on the directional synchronization feature information. It then filters the node numbers with directional inconsistency or response delay features, and archives them as abnormal nodes to generate a pressure difference mutation node set.

[0025] The present invention is improved in that the node degradation feature extraction module includes:

[0026] The operation and maintenance index extraction submodule calls the differential pressure change node set, obtains the fire hydrant number corresponding to each node, extracts the usage frequency, opening and closing response delay, rated service life and number of maintenance records for each number, establishes the operation and maintenance index set corresponding to the node, and generates operation and maintenance parameter summary information.

[0027] Based on the summarized operation and maintenance parameter information, the performance drift calculation submodule calculates the average difference between the standard time corresponding to the opening and closing operation and the current delay, calculates the rate of change of response delay in adjacent cycles, and calculates and obtains the performance degradation index of the fire hydrant node.

[0028] The functional level determination submodule, based on the performance degradation index of the fire hydrant node, binds the functional degradation level code and number corresponding to each node according to the equipment remaining performance grading standard, and generates a functional degradation status label set.

[0029] The present invention is improved in that the risk level generation module includes:

[0030] The trend factor extraction submodule, based on the functional decay state label set and the differential pressure change node set, obtains the functional decay level and the corresponding water pressure trend direction combination of each abnormal node, extracts the continuous time segment of trend change and the time delay value of pressure recovery process, calculates the trend duration period and recovery response duration, and generates an abnormal trend feature parameter set.

[0031] The tolerance difference calculation submodule, based on the abnormal trend characteristic parameter group, calls the node remaining response capability index corresponding to the function decay level, and calculates and obtains the abnormal tolerance offset value by introducing the abnormal duration period and recovery delay time.

[0032] The risk level output submodule sets the risk level range based on the abnormal tolerance offset value and the average and standard deviation of the historical abnormal tolerance offset values, divides the risk warning level, encodes the current risk level of each node, and implements fire hydrant risk warning for management personnel to obtain real-time fire hydrant risk warning information.

[0033] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0034] In this invention, by correlating flow velocity deviation with control signals, abnormal water intake behavior can be identified without signals. Combined with a collaborative analysis mechanism of pressure changes at adjacent nodes, hydraulic disturbance areas under non-human control can be accurately identified. By analyzing the spatial independence of pressure response trends, it can determine whether abnormal behavior is isolated and eliminate interference from overall fluctuations in the pipeline system. Based on historical data of start-up and shutdown delays and response performance, the device's functional degradation level can be extracted, enhancing the quantitative assessment capability of node functional status. When identifying anomalies, equipment performance boundary information is extracted simultaneously and the magnitude and trend persistence of the anomalies are compared. When issuing risk warnings, it has the ability to finely classify risk levels, improving the accuracy of risk identification and enhancing the proactive detection capability of early-stage potential faults. Attached Figure Description

[0035] Figure 1 This is a system flowchart of the present invention;

[0036] Figure 2 This is a flowchart of the flow rate anomaly identification module of the present invention;

[0037] Figure 3 This is a flowchart of the node-coordinated differential pressure testing module of the present invention;

[0038] Figure 4 This is a flowchart of the differential pressure trend independence discrimination module of the present invention;

[0039] Figure 5 This is a flowchart of the node degradation feature extraction module of the present invention;

[0040] Figure 6 This is a flowchart of the risk level generation module of the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0042] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0043] Please see Figure 1 The present invention provides a technical solution: a fire hydrant monitoring intelligent early warning system based on anomaly analysis technology. The system includes a flow velocity anomaly identification module, a node collaborative differential pressure verification module, a differential pressure trend independence discrimination module, a node degradation feature extraction module, and a risk level generation module.

[0044] The flow velocity anomaly identification module acquires the flow velocity monitoring data of the fire hydrant monitoring nodes, calls the flow velocity value of each node within the target time range at the current time point, extracts the median value and standard deviation, compares the current flow velocity value of the fire hydrant monitoring node with the median value, calculates the deviation amplitude value, and compares it with whether there is a fire hydrant valve control signal at the corresponding time point. It determines whether the deviation amplitude exceeds the single-point anomaly identification threshold and no control operation signal appears, marks it as an abnormal node, and generates a flow velocity offset label.

[0045] The node collaborative differential pressure test module calls the flow velocity offset tag. Based on the identified abnormal node number, it extracts the relative differential pressure change and instantaneous pressure gradient direction of adjacent monitoring nodes on the associated pipe section. It obtains the pressure change rate of each node within the sampling period and compares it with the abnormal node to determine whether the pressure change direction between nodes is synchronized. It filters out node groups with inconsistent or abrupt changes in direction to obtain the flow velocity offset differential pressure response area.

[0046] The pressure difference trend independence discrimination module obtains the pressure drop start time of the abnormal node and the response delay of the surrounding nodes based on the flow velocity offset pressure difference response area. It extracts the directional derivative sign sequence of water pressure change in each group, compares the change direction of the abnormal node with that of the neighboring nodes, and if most of the neighboring nodes do not have synchronous trend changes or have opposite directions, the trend is marked as spatially independent abrupt change behavior, thus obtaining the pressure difference abrupt change node set.

[0047] The node degradation feature extraction module calls the differential pressure mutation node set. Based on the operation and maintenance data of each node, it extracts the usage frequency, opening and closing response delay, rated service life and number of maintenance records of the fire hydrant. It calculates the average opening and closing time drift and response delay change rate respectively. Based on the established equipment remaining performance classification standard, it classifies the current functional degradation level of each fire hydrant node and obtains the functional degradation status label set.

[0048] The mean opening and closing time drift refers to the average difference between the actual time required to complete the opening and closing operation and the standard time when the equipment is initially put into use; the response delay change rate refers to the rate of change of the delay between the equipment response start time and the control signal issuance time within a unit time period; the equipment residual performance grading standard is often used in water conservancy projects for equipment performance evaluation, including dimensions such as running time, number of runs, response characteristics, and maintenance frequency.

[0049] The risk level generation module extracts the functional decay level and spatial pressure trend characteristics corresponding to each abnormal node based on the functional decay status label set and the differential pressure change node set. Combining the abnormal duration and pressure recovery delay, it determines whether the abnormal behavior exceeds the tolerance range corresponding to the remaining performance level of the current equipment. It also performs a graded judgment based on the trend characteristics, classifies the risk warning level, implements fire hydrant risk warning for management personnel, and obtains real-time fire hydrant risk warning information.

[0050] The velocity deviation label includes the node number of the abnormal velocity deviation magnitude, the time period identifier of the period when the control signal was not triggered, and the high deviation probability time marker. The velocity deviation pressure difference response area includes the node combination of the instantaneous pressure difference direction inconsistency, the time synchronization conflict marker, and the local response mismatch area number. The pressure difference change node set includes the central node pressure drop abnormal attribute marker, the description of the non-response behavior of adjacent nodes, and the trend direction inconsistency index. The functional degradation status label set includes the structural functional level label, the performance degradation behavior type, and the segmented description of the remaining response capacity. The real-time risk warning information of fire hydrants specifically includes the functional lower limit conflict level, the trend change intensity level, and the node warning level judgment marker.

[0051] Please see Figure 2 The flow rate anomaly detection module includes:

[0052] The flow velocity baseline extraction submodule acquires the flow velocity monitoring data of the fire hydrant monitoring nodes, and extracts the flow velocity value of each node within the target time range at the current time point in combination with the corresponding sampling time point. It also extracts the median and standard deviation to obtain stable flow velocity statistical indicators.

[0053] The flow velocity baseline extraction submodule acquires the flow velocity monitoring data and sampling time points of the fire hydrant monitoring nodes. It needs to locate a fixed statistical window for each node before the current time point according to the sampling interval to obtain all historical flow velocity data within that window. This window is generally set to 10 minutes, with data collected once per minute, resulting in 10 sets of data. For example, the flow velocities recorded for node A01 between 10:00 and 10:10 are [5.1, 5.2, 5.3, 5.4, 5.0, 4.9, 5.3, 5.2, 5.1, 5.3]. The median value is calculated as the average of the fifth and sixth values: Calculating the standard deviation requires iterating through all values ​​in the dataset and taking the square root of their squared deviations from the mean. Assume the mean of this data segment is 5.08. The standard deviation is calculated as follows: the differences for each term are [-0.02, 0.02, 0.22, 0.32, -0.08, -0.18, 0.22, 0.12, 0.02, 0.22], and the average of the squares is 0.0308, which is 0.175. This yields the median and standard deviation of the current node's flow velocity statistical window, thus forming a stable flow velocity statistical index. The statistical window length and sampling frequency should be set according to the network flow fluctuation characteristics and regional hydraulic structure. For example, the maximum allowable sampling interval can be set to 1 minute to ensure that the sampling resolution meets the requirements for instantaneous fluctuation monitoring. The median and standard deviation serve as benchmark indicators for stable flow velocity and are not adjustable parameters. However, their numerical range directly reflects the degree of flow velocity fluctuation and needs to be combined with the historical fluctuation range of the pipeline flow to establish a reliable interval. For example, if the historical flow velocity fluctuation range of a certain node is 3.0~8.5... If the standard deviation exceeds 1.2, then the standard deviation is greater than 1.2. Nodes are considered high-fluctuation nodes. This range is determined by extracting and sorting data from the past 30 days for each node. The standard deviation of stable nodes should be controlled between 0.3 and 0.9. Within a certain range, so that a clear flow rate identification benchmark can be formed.

[0054] The flow velocity deviation calculation submodule, based on stable flow velocity statistics, calls the flow velocity value corresponding to each node at the current time point and calculates the ratio of the absolute value of the deviation of the flow velocity value from the median value to the standard deviation of the fluctuation, using the formula:

[0055] ;

[0056] The deviation judgment value of the current node within the statistical window is obtained through calculation. The single-point anomaly identification threshold is calculated based on the stable flow velocity statistical index. It is determined whether the deviation judgment value is greater than the single-point anomaly identification threshold, and anomaly is marked to obtain the suspicious flow velocity offset identification group.

[0057] in, This represents the velocity deviation judgment value, which is a dimensionless anomaly index. This represents the normalized value of the current flow rate, calculated by normalizing the current sampled value of this node against its past sampled value range. This represents the normalized value of the median flow velocity within the current time window, obtained by normalizing the median of the sampled values ​​within this window with the historical sampling range. This represents the standard deviation of the flow velocity within the current time window, reflecting the dispersion of the sampled values. This represents the normalized value of the average amplitude of short-period disturbances, obtained by comparing the mean range over this time period with the maximum historical fluctuation amplitude at each node. The normalized value representing the duration of continuous anomalies is calculated by dividing the duration of the anomaly by the set maximum allowable duration. The normalized value representing the frequency of valve opening and closing is obtained by dividing the number of opening and closing times within the statistical period by the upper limit of the standard operating frequency for this type of equipment. This represents the normalized value of the system's flow fluctuation in the current hourly period, which is calculated by normalizing the flow velocity standard deviation across all network nodes during this period.

[0058] The velocity deviation calculation submodule first normalizes the node's current velocity value against its historical range. The normalization method involves subtracting the historical minimum value from the current velocity value and then dividing by the difference between the maximum and minimum values. If the historical maximum value is 10... The minimum value is 0 The current flow rate is 5.3. The normalized value is 0.53, and the median value is 5.0. The normalization is set to 0.50. Both the standard deviation and the disturbance amplitude are normalized values. The disturbance amplitude is calculated by taking the average of the ranges within a short period and dividing it by the historical maximum range. For example, if the short-term range is 1.0 and the historical maximum range is 5.0, the disturbance amplitude is 0.2. Substituting these values ​​into the calculation formula:

[0059] The deviation items are:

[0060] ;

[0061] The duration of the anomaly was 40. The maximum allowed value is 120. Then the normalized value is:

[0062] ;

[0063] The valve actuation frequency is 4 times, with a maximum of 10 times. After normalization, it is:

[0064] ;

[0065] The second item is:

[0066] ;

[0067] The hourly flow velocity fluctuation was 0.6. The largest version across the entire network is 2.0. ,calculate:

[0068] ;

[0069] calculate:

[0070] ;

[0071] ;

[0072] The formula's calculation logic consists of three parts: the first term measures the intensity of the deviation of the current instantaneous flow velocity from the normal median value; the second term measures the combined risk factor of the duration of the anomaly and the frequency of valve operation; and the third term reflects the intensity of the overall system flow fluctuation within the current period. The sum of these three terms forms a single node's current anomaly severity index. This value is dimensionless and can be directly compared with a threshold. If the threshold is set to 0.6, the node is marked as an abnormal node. This threshold is set based on all nodes within the past month. The values ​​were distributed statistically, and an upper bound of the 90% confidence interval was set. Actual statistics showed that most nodes... The values ​​are concentrated between 0.3 and 0.55, so 0.6 is set as the anomaly judgment benchmark. This value does not change with a single node, but different levels can be set according to the regional characteristics. For example, it can be set to 0.7 in the high-voltage trunk area and 0.55 in the terminal area to improve sensitivity.

[0073] The control behavior exclusion submodule determines whether there is a fire hydrant valve control signal at the corresponding time point based on the suspicious flow rate deviation identifier group, filters out the time period data with control signal behavior, retains the deviation identifier without valve control, and generates flow rate deviation labels.

[0074] After the control behavior exclusion submodule calls the suspicious flow velocity offset identifier group obtained in the previous step, it needs to match the solenoid valve opening and closing behavior records of the corresponding nodes at each offset time point. It compares whether there is an "open" or "close" command signal within ±10 seconds before or after the current time point. If so, the record at that time point is marked as "accompanied by control signal"; otherwise, it is marked as "no accompanying control." For example, at the offset point of 602 seconds for node A01, if the node issues a solenoid valve "open" signal at 601.8 seconds, that time point will be excluded and will no longer participate in subsequent anomaly analysis. Only records of flow velocity offsets occurring without control behavior are retained. For example, if there is no opening or closing action record for A02 at the offset of 730 seconds, it is retained. This processing uses a matching method to compare the signal time with the offset time. By setting a ±Δt time tolerance window, the specific Δt value is set based on the system response delay experiment, and is generally set to 10~15 seconds in practice. This time window is a fixed value. The time interval is fixed and does not vary with nodes or time periods. It is set based on the routine response test report of the solenoid valve actuator. The experiment was conducted at 12 typical nodes and it was found that the maximum response delay did not exceed 8.3 seconds. Therefore, 10 seconds was set as a safe interval to ensure that the records are fully compared and to avoid introducing false exclusion phenomena, as shown in Table 1.

[0075] Table 1. Matching Flow Velocity Offset Points with Control Signals

[0076]

[0077] As shown in Table 1, records A01 and A03 with a time difference within ±10 seconds were determined to be caused by control behavior and were therefore excluded and no longer included in the anomaly analysis.

[0078] Please see Figure 3 The node-coordinated differential pressure testing module includes:

[0079] The adjacent node extraction submodule calls the flow velocity offset label, indexes the pipe segment structure diagram corresponding to the identified abnormal node number, obtains the adjacent connected monitoring nodes, extracts the pressure data sequence of each adjacent node in the current sampling period, calculates the unit length pipe segment pressure difference value between the adjacent node and the abnormal node, and obtains the adjacent node pressure difference change group.

[0080] After the adjacent node extraction submodule calls the abnormal node number marked in the velocity offset label, it needs to locate the pipe segment topology to which each number belongs in the system database. Following the topology map indexing method, it queries the upstream and downstream node information connected to the current node and extracts its pressure data sequence within the current sampling period. The system default sampling period is 60 seconds, once per second, for a total of 60 data points. After obtaining the pressure sequence of each pair of adjacent nodes, it sequentially iterates and calculates the pressure difference value of its corresponding unit pipe segment. The pressure difference value is calculated using the formula: Calculation, where and These represent the pressure values ​​of two adjacent nodes at that moment. This represents the straight-line pipe distance between two nodes, in meters; for example, the distance between nodes N01 and N02 is 120 meters. If the current instantaneous pressures are 310 kPa and 290 kPa respectively, then the pressure difference is... This process was repeated at all sampling points to obtain 60 instantaneous differential pressure values, which were then combined into differential pressure change groups. For each group, the sampling time and node pair number were recorded to facilitate subsequent identification of node pair combinations causing pressure fluctuations. In this module, the unit pipe segment distance... Derived from the GIS system or pipeline design drawings, the error does not exceed ±1m. The system's default rounding accuracy is 0.1m. This data is not used as a variable. However, if the distance between nodes is less than 20m, the pressure difference calculation is skipped. Because it is easily affected by the amplification of local disturbances, the minimum pipe length judgment criterion is: in the experiment of 10 short-distance nodes, it was found that the pressure difference fluctuation deviation rate was higher than 15% when it was less than 20m, so it was set as the rejection criterion.

[0081] The pressure rate comparison submodule extracts the pressure change sequence of each node within the current sampling period based on the pressure difference change group of adjacent nodes, calculates the instantaneous pressure change rate of each sequence, and judges the difference ratio with the pressure change rate of the corresponding abnormal node. It marks the difference value group based on whether it exceeds the instantaneous pressure rate difference threshold.

[0082] After the pressure rate comparison submodule calls the pressure difference change group obtained in the previous step, it sequentially extracts the pressure change sequence of each pair of nodes within the current sampling period. This sequence is an array of pressure values ​​sampled per second. The instantaneous rate of change needs to be obtained by first subtracting every two adjacent sampling points in this array and then dividing by the time interval. For example, the pressure sequence of node N02 within a 60-second sampling period is [290, 291, 293, 292, 291, 290, ...]. Then the rates of the first two terms are This yields 60 instantaneous rate values. The average rate of these values ​​is then calculated as a representative value. The difference between this average rate and the average rate at the corresponding anomalous node is calculated, and then divided by the anomalous node's rate value to obtain the difference ratio. For example, if the average pressure change rate at the anomalous node is 1.2... The current node value is 0.8. The ratio is The system defaults to marking a difference ratio greater than 0.3 as abnormal. This threshold is a fixed setting, which is derived from the following: in the historical data of 38 node groups, the maximum rate difference under normal working conditions is 28.7%. After considering a 5% margin, it is set to 30%. If it exceeds this, it is marked as a significant rate difference. In actual use, it is not recommended to increase this value, otherwise high-voltage nodes with reversed direction will be missed, as shown in Table 2.

[0083] Table 2. Statistics of Pressure Differential Rate Difference Ratio

[0084]

[0085] As shown in Table 2, the rate changes of nodes N02 and N05 are significantly different from those of N01, exceeding the set value of 0.3, and are therefore marked as nodes with significant rate differences.

[0086] The gradient direction filtering submodule extracts the pressure change direction identifier of each node at the current time point based on the pressure difference rate difference value group, performs a consistency matching operation on the direction identifier, determines whether there are cases where the direction of adjacent nodes is inconsistent with or opposite to the direction of abnormal nodes, filters the node combination with contradictory direction characteristics, and generates the flow velocity offset pressure difference response region.

[0087] After extracting the aforementioned rate difference value group, the gradient direction filtering submodule determines whether there is any reverse behavior based on the pressure change direction sampled at the current time point. The direction value is defined by the sign of the difference between the current sampling point and the previous sampling point; a positive value indicates pressure increase, and a negative value indicates pressure decrease. For example, if the pressure two seconds before and after the current node is 295... With 293 If the direction is negative, the change is -2. When the abnormal node is positive, the direction is opposite to that of the current node. Extract the direction sequence of all nodes at 60 sampling points and convert it into a symbol array, such as [+1,+1,-1,-1,+1,…]. Then compare it bit by bit with the direction sequence of the abnormal node and count the proportion of inconsistencies. If the inconsistency proportion exceeds the set threshold of 0.5, the node is marked as a group of nodes with opposite directions, forming a flow velocity offset pressure difference response region. The threshold of 0.5, i.e., half of the inconsistencies, is used as the criterion for the significance of the direction contradiction. The basis for this setting is: in the low-pressure branch experiment, the average consistency rate of normal fluctuation direction is 0.82. The system sets below 50% as a high direction mismatch behavior. This setting ensures that the selected nodes have significant direction differences, while eliminating the phase shift caused by slight disturbances. Finally, these nodes with contradictory directions are combined and output as the offset response region for subsequent modules to identify the pressure difference trend change region.

[0088] Please see Figure 4 The differential pressure trend independence discrimination module includes:

[0089] The initial feature extraction submodule obtains the location index of the first continuous decrease in water pressure value of each abnormal node within the sampling period based on the flow velocity offset pressure difference response area, and extracts multiple monitoring nodes adjacent to the abnormal node. It also extracts the first response change time index of each adjacent node within the same time segment, calculates the time difference between adjacent nodes and abnormal nodes, and generates response time difference information.

[0090] The initial feature extraction submodule reads the pressure sampling sequence point by point within the system sampling period based on the marked abnormal nodes in the velocity offset pressure difference response region. It compares the difference between each time point and the previous time point. If three or more consecutive sampling points are negative, the first occurrence is marked as the starting point of the water pressure drop. This drop must accumulate to a value exceeding 1.5. This is to eliminate minor perturbations; for example, the 60-second sampling pressure sequence of the anomalous node N01 is [310, 309.2, 308.5, 307.6, 307.0, ...]. Points 1 through 4 show a continuous decrease, with differences of -0.8, -0.7, and -0.9 respectively, accumulating to -2.4. If the initial index is 1 second, then the sampling sequences of the neighboring nodes N02, N03, and N04 are obtained. Within the same period, the first descent point is found, its index is extracted, and the difference is calculated with the index of the abnormal node to obtain the response delay. For example, if the first descent point for N02 is at the 3rd second, for N03 at the 1st second, and for N04 at the 5th second, then their response time differences are 2 seconds, 0 seconds, and 4 seconds, respectively. All response time differences for this node group are recorded to form response time difference information. In this module, the descent judgment threshold is 1.5 seconds. Based on historical data from 15 anomalous nodes, under continuous disturbances, the cumulative decrease at three points exceeded 1.5. The probability of 92.6% is used as the benchmark for abnormal startup identification, as shown in Table 3.

[0091] Table 3. Statistics on the time difference between the initial pressure drop at the abnormal node and the response of adjacent nodes.

[0092]

[0093] As shown in Table 3, N03 is a synchronous response node, while N02 and N04 have different degrees of delay. Among them, N04 has a significantly slower response and will be regarded as a potential trend deviation feature node in subsequent judgments.

[0094] The trend response calculation submodule calls the response time difference information, calculates the differential value sequence of water pressure change of nodes in a continuous period based on the water pressure sampling value corresponding to each group of nodes, extracts the direction of the differential sequence, marks the direction symbols to form the change direction combination between nodes, judges the synchronicity of the direction combination, and obtains the directional synchronicity feature information.

[0095] The trend response calculation submodule calls the pressure sequence of each abnormal node and its adjacent nodes recorded in the previous step. It takes 10 seconds of data after the starting point for trend judgment, forming a 10-point sampling segment. A first-order difference operation is performed on each sequence, subtracting the previous point from each point to obtain the pressure change. For example, the 10-second segment is [307.6, 307.1, 306.5, 306.3, 306.0, ...], and the difference is [-0.5, -0.6, -0.2, -0.3, ...]. The difference values ​​are then mapped to direction signs: negative is "-1", positive is "+1", and zero is "0", forming a direction sequence, such as [-1, -1, -1, -1, 0, +1, ...]. The direction sequence of the abnormal node is compared with that of each adjacent node, and the... Each item is paired, and the number of matching points in 10 directions is counted. If the consistency rate exceeds 70%, it is marked as "directionally synchronized"; if it is less than 30%, it is marked as "directionally reversed"; the rest are "directionally ambiguous". For example, if the directional consistency rate of N01 and N03 is 90%, but the consistency rate with N02 is only 20%, then N03 is regarded as a trend-following node, and N02 is marked as an out-of-direction response node. The synchronicity judgment threshold is set to 70% and 30%, which comes from high-frequency trend extraction experiments. In typical samples of pressure difference mutation, the synchronization rate of node groups was found to be higher than 0.75, so 70% is set as the lower limit of synchronization. The consistency rate of out-of-direction nodes is lower than 0.28 under the condition of reverse perturbation, so 30% is taken as the lower limit of judgment. The synchronicity information record will be used by the subsequent independence judgment submodule.

[0096] The independence screening and determination submodule determines whether the direction of water pressure change of abnormal nodes conflicts with or is missing from that of adjacent nodes based on the directional synchronization feature information. It then filters out node numbers with inconsistent directions or response delays, marks them as abnormal, and archives them to generate a set of nodes with sudden pressure changes.

[0097] The independence screening and determination submodule, based on the directional synchronization characteristic information obtained from the trend response calculation submodule, sequentially determines whether each anomalous node has at least two or more adjacent nodes exhibiting a "reverse direction" or "unclear direction" state. If this condition is met, the current anomalous node is marked as a "direction change isolated node," and its adjacent node numbers, response time difference, and directional state combination are recorded to construct a complete record of anomalous trend isolation characteristics. For example, the directional consistency rate of adjacent nodes N01, N02 and N04, is 20% and 40%, respectively, which is lower than the synchronization requirement. Although N03 is synchronized, its number is insufficient. Finally, N01 was marked as an independent trend node and entered the pressure differential mutation node set. The judgment criterion in this sub-module is: at least two neighboring nodes have conflicting directions or no response. The judgment basis is that in the network trend response pattern recognition sample, if the directions of most neighboring nodes are mismatched, it is very easy to cause the system pressure interference to fail to diffuse and form a local hydraulic fault structure. The experiment backtracked on 27 groups of node combinations and found that more than 85% of the trend isolated nodes subsequently formed substantial pressure differential mutations. Therefore, this system uses it as a key screening mark. The marking result will be output to the subsequent node degradation analysis module for joint risk calculation.

[0098] Please see Figure 5 The node degradation feature extraction module includes:

[0099] The operation and maintenance index extraction submodule calls the differential pressure change node set, obtains the fire hydrant number corresponding to each node, extracts the usage frequency, opening and closing response delay, rated service life and number of maintenance records for each number, establishes the operation and maintenance index set corresponding to the node, and generates operation and maintenance parameter summary information.

[0100] The maintenance index extraction submodule retrieves the associated fire hydrant equipment number for each node in the differential pressure change node set, and extracts the basic maintenance information associated with each device through database queries. This information includes: usage frequency (the actual number of opening and closing operations per unit time), opening and closing response delay (the time difference between the control signal being issued and the valve completing its opening / closing in each operation), rated service life (the safe service life set by the manufacturer in the equipment manual), and the number of maintenance records (the number of records of equipment repair, calibration, and component replacement registered in the system log). The extraction cycle is also specified. The system is set to a period of approximately 90 days. Each indicator is statistically analyzed within this period. The usage frequency is calculated by dividing the total number of operations by the number of days. The response latency is calculated by averaging the delay of each start-up and shutdown operation. The number of maintenance operations is accumulated. The rated service life is taken from the value marked in the instruction manual corresponding to the equipment model. The response latency needs to remove extreme data caused by communication abnormalities or human error. The filtering rule is that records with a latency greater than 3 times the standard upper limit will be discarded. This rule was determined after evaluating data from 95 different nodes. This threshold ensures that actual operating data is retained while avoiding misjudgment. Finally, a set of operation and maintenance indicators for each node is formed, as shown in Table 4.

[0101] Table 4. Extraction of Operation and Maintenance Indicators for Fire Hydrants

[0102]

[0103] As shown in Table 4, the service life and response latency of node H03 are both relatively high, and the number of maintenance visits also exceeds the recommended frequency, indicating that the subsequent performance degradation will be more pronounced.

[0104] The performance drift calculation submodule, based on the summarized operation and maintenance parameters, calculates the average difference between the standard time corresponding to the start / stop operation and the current latency, and calculates the rate of change of response latency between adjacent cycles using the following formula:

[0105] ;

[0106] Calculate and obtain the performance degradation index of the fire hydrant node;

[0107] in, This represents the performance degradation index of fire hydrant nodes, and is a dimensionless trend value. This represents the normalized value of the current average start / stop delay, obtained by normalizing the average start / stop delay over the current monitoring period with the upper limit of the design standard delay. This represents the normalized value of the design standard opening and closing delay, obtained by normalizing the target opening and closing time in the equipment manual with the industry standard range. This represents the normalized value of the standard deviation of the start-up and stop-down delay over the past three periods. It is obtained by normalizing the standard deviation of the start-up and stop-down delay under a sliding window with its periodic average fluctuation range. The normalized value representing the rate of change of response delay is obtained by dividing the difference in response of multiple on / off cycles within a unit time period by the theoretical upper limit of the change in on / off delay. The normalized value representing the service life is obtained by normalizing the actual operating years of the equipment against its rated life range. The normalized value representing the number of maintenance is obtained by normalizing the cumulative number of maintenance at the node with the upper limit of the recommended annual inspection frequency in the technical manual.

[0108] After calling the summary information of operation and maintenance parameters, the performance drift calculation submodule needs to normalize each parameter. The normalization method is as follows: current start / stop latency. Normalization is achieved by dividing the current mean by the standard latency limit, which is the standard on / off latency. Divide the target value given in the equipment model specification by the industry-allowed maximum response time (set to 6 seconds). For example, the average start / stop time for node H01 is 3.2 seconds, and the specification target is 2.5 seconds. After normalization... , The standard deviation of the start-stop delay within three cycles is 0.8 seconds, and the average fluctuation range of the cycles is 1.2 seconds. Response delay change rate The difference between the maximum and minimum response within a unit period is divided by the upper limit of the theoretical design delay (set to 4 seconds). For example, if the maximum is 4.2 seconds and the minimum is 2.6 seconds, then... The service life is unified as The recommended number of inspections is 4, and the maximum recommended annual inspection frequency is 5. Substitute all values ​​into the formula:

[0109] ;

[0110] ;

[0111] In this formula, the first term represents the intensity of the opening and closing response time drift, and the second term is a performance degradation weight term composed of response fluctuation, service life and maintenance conditions. The overall relationship is positively proportional. The larger the value, the more the equipment deviates from the initial standard performance. The response drift judgment standard is derived from measured sample data. A D value in the range of 0.3-0.5 indicates slight drift, 0.5-0.7 indicates moderate drift, and more than 0.7 indicates severe performance degradation.

[0112] The functional level determination submodule is based on the performance degradation index of fire hydrant nodes. According to the equipment remaining performance classification standard, it binds the functional degradation level code and number corresponding to each node to generate a functional degradation status label set.

[0113] The function level determination submodule calculates the performance degradation index based on each node. The system invokes the equipment remaining performance grading standard and classifies nodes into grades within a preset threshold range. Each node is assigned to a corresponding grade based on its D value. For example, a D value less than 0.3 is "Level 1 Normal", 0.3-0.5 is "Level 2 Mild", 0.5-0.7 is "Level 3 Moderate", and above 0.7 is "Level 4 Severe". The grade number is bound to the node number to form a functional degradation status label set. For example, node H01 corresponds to Level 2 Mild, and the label is "H01-G2". The label information includes the node number, grade code, and assessment period timestamp, and is output to the subsequent risk level generation module. The grade classification threshold in this module refers to the failure probability statistics of each grade node in the next 3 months in historical operating conditions. The failure probability of Level 1 nodes is less than 5%, Level 2 is between 6-12%, Level 3 is 13-25%, and Level 4 is higher than 30%. This constructs a layered evaluation standard for equipment operation reliability, which is used to drive the operation of the system partition early warning model.

[0114] Please see Figure 6 The risk level generation module includes:

[0115] The trend factor extraction submodule obtains the functional decay level and corresponding water pressure trend direction combination of each abnormal node based on the functional decay state label set and the differential pressure change node set, extracts the continuous time segment of trend change and the time delay value of pressure recovery process, calculates the trend duration period and recovery response time, and generates abnormal trend feature parameter set.

[0116] The trend factor extraction submodule performs node matching based on the functional decay state label set and the differential pressure change node set. Each identified abnormal node is associated with its functional decay level and the water pressure trend direction within the current sampling period. First, the water pressure sequence of the node within a specified detection window (e.g., the past 60 seconds) is retrieved, and the sequence is subjected to first-order differencing to identify continuous decreasing segments and locate their start and end times, thereby extracting the duration of the abnormal trend. Determine the starting point of the pressure recovery process after the termination of the continuous descent state, and take the moment when the pressure returns to the starting point of the descent as the completion point of the recovery. The time difference between the two points is the recovery delay time. For example, if the pressure at node N12 continuously decreases between 600 and 615 seconds, lasting for 15 seconds, and recovers to its original value at 642 seconds, then the recovery delay is 27 seconds. If similar patterns appear in multiple trend segments, the most significant duration needs to be selected. This significance is determined by the cumulative value of the decrease. The recovery point is defined as a recovery of more than 90%. All extracted abnormal trend feature parameters are bound by node numbers to form an abnormal trend feature parameter group. The maximum abnormal duration is determined by the node equipment type and location. The limit for fire hydrants on the main pipeline is 30 seconds, and for fire hydrants at the end, it is 20 seconds. The average recovery delay is the arithmetic mean of the recovery times of all adjacent nodes on the same branch. If the recovery process spans multiple cycles, the earliest recovery segment is taken as the recovery base point to eliminate the influence of subsequent secondary interference.

[0117] The tolerance difference calculation submodule, based on the abnormal trend characteristic parameter group, calls the node's remaining response capability index corresponding to the function decay level. By introducing the abnormal duration period and recovery delay time, it uses the following formula:

[0118] ;

[0119] Calculate and obtain the anomaly tolerance offset value;

[0120] in, Indicates the anomaly tolerance offset value. This represents the normalized value indicating the duration of the anomaly, obtained by dividing the duration by the trend duration limit set for the node type. This represents the normalized value of the recovery delay, obtained by dividing the recovery time by the average delay within the current node group. This represents the normalized value of the node's remaining response capacity, obtained by dividing the node's remaining start / stop response capacity by the upper limit of the reference capacity for similar nodes. The factor representing the degree of variation in trend direction is obtained by dividing the number of sign differences between the current node's water pressure change direction and the directions of all its neighboring nodes by the total number of neighboring nodes. This represents the device function attenuation level factor, which is obtained by dividing the current node's function attenuation level by the maximum function attenuation level.

[0121] The tolerance difference calculation submodule takes the trend parameter group as input and sequentially calls the preset remaining response capability index for each node's corresponding function attenuation level to calculate the normalized value of the abnormal duration and recovery delay time, respectively:

[0122] ;

[0123] in and These represent the trend duration and the recovery delay, respectively. Set a duration limit for the node type. The average recovery delay of neighboring nodes is calculated; then the normalized value of the node's remaining response capability is calculated:

[0124] ;

[0125] If node N12 has a remaining response time of 2.3 seconds and a maximum reference capability limit of 5 seconds, then Directional variation factor The value is calculated by dividing the number of differences between the direction sequence of this node and the direction sequences of all its neighboring nodes by the number of neighboring nodes. For example, if the direction difference is 2 and there are a total of 5 neighboring nodes, then... Functional degradation level factor Divide the current level value by the maximum level. For example, if the current level is 3 and the maximum is 4, then... Substitute all parameters into the formula:

[0126] ;

[0127] set up ,but:

[0128] ;

[0129] The formula is calculated as follows: the first part represents the ratio of the degree of trend anomaly to the remaining capacity; the second part is the coupling fluctuation factor between directional distortion and equipment functional degradation; the product of the two constitutes the comprehensive offset intensity index. Ultimately, this is used to assess whether the current state of a node has significantly exceeded its carrying capacity, as shown in Table 5.

[0130] Table 5. Calculation Table of Abnormal Trends and Response Capability Deviation

[0131]

[0132] As shown in Table 5, the N14 offset intensity is significantly exceeded, and its directional variation and high-function degradation together cause a strong overcapacity risk.

[0133] The risk level output submodule sets the risk level range based on the abnormal tolerance offset value and the average and standard deviation of the historical abnormal tolerance offset values, divides the risk warning level, encodes the current risk level of each node, and implements fire hydrant risk warning for management personnel to obtain real-time fire hydrant risk warning information.

[0134] The risk level output submodule uses the calculated anomaly tolerance offset value. Based on this, a tiered matching operation is performed to map the risk level ranges set by the system: It is classified as "Level 1 Low Risk". It is classified as "Level 2 Medium Risk". It is classified as "Level 3 High Risk". The risk level is classified as "Level 4 Extremely High Risk". Each level is represented by a label G1 to G4, which are combined with the node number to form the final risk warning label, such as "N12-G3". After the label is generated, it is sent to the risk warning platform simultaneously, and marked with a timestamp and assessment cycle number, which is provided to the management personnel to implement early warning intervention. The level division standard in this module refers to the system's historical failure rate database. When the R value is higher than 1.3, the probability of failure triggering within the past 30 days exceeds 34%, so it is set as the extremely high risk limit. The risk output label can directly drive operational strategies such as remote detection of fire hydrants, on-site inspection, or mandatory re-inspection.

[0135] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A fire hydrant monitoring intelligent early warning system based on anomaly analysis technology, characterized in that, The system includes: The flow velocity anomaly identification module acquires the flow velocity monitoring data of the fire hydrant monitoring node, compares the current flow velocity value of the fire hydrant monitoring node with the median value, calculates the deviation amplitude value, determines whether the deviation amplitude exceeds the single-point anomaly identification threshold and no control operation signal appears, and generates a flow velocity offset label. The node collaborative differential pressure test module calls the flow velocity offset tag to obtain the pressure change rate of each node within the sampling period and compares it with abnormal nodes. It then filters out node groups with inconsistent directions or abrupt changes to obtain the flow velocity offset differential pressure response area. The node-coordinated differential pressure testing module includes: The adjacent node extraction submodule calls the flow velocity offset tag, indexes the pipe segment structure diagram corresponding to the identified abnormal node number according to the abnormal node number, obtains the adjacent connected monitoring nodes, extracts the pressure data sequence of each adjacent node in the current sampling period, calculates the unit length pipe segment pressure difference value between the adjacent node and the abnormal node, and obtains the adjacent node pressure difference change group. The pressure rate comparison submodule extracts the pressure change sequence of each node within the current sampling period based on the pressure difference change group of adjacent nodes, calculates the instantaneous pressure change rate of each sequence, and judges the difference ratio with the pressure change rate of the corresponding abnormal node. It marks whether it exceeds the instantaneous pressure rate difference threshold to obtain the pressure difference rate difference value group. The gradient direction filtering submodule extracts the pressure difference change direction identifier of each node at the current time point based on the pressure difference rate difference value group, performs a consistency matching operation on the direction identifier, determines whether there are cases where the direction of adjacent nodes is inconsistent with or opposite to the direction of abnormal nodes, filters the node combination with contradictory direction characteristics, and generates the flow velocity offset pressure difference response region. The pressure difference trend independence discrimination module extracts the direction symbol sequence of water pressure change in each group based on the flow velocity offset pressure difference response area, compares the change direction of abnormal nodes with neighboring nodes, marks the spatial independence mutation behavior, and obtains the pressure difference mutation node set. The node degradation feature extraction module calls the differential pressure change node set, calculates the average drift value of opening and closing time and the change rate of response delay based on the operation and maintenance data of each node, classifies the current functional degradation level of each fire hydrant node, and obtains a functional degradation status label set. The risk level generation module, based on the functional attenuation state label set and the differential pressure change node set, and combined with trend characteristics, makes a graded judgment, classifies the risk warning level, implements fire hydrant risk warning for management personnel, and obtains real-time fire hydrant risk warning information.

2. The intelligent early warning system for fire hydrant monitoring based on anomaly analysis technology according to claim 1, characterized in that, The velocity deviation label includes the node number of the abnormal velocity deviation magnitude, the time period identifier of the period when the control signal was not triggered, and the high deviation probability time marker. The velocity deviation pressure difference response area includes the node combination of instantaneous pressure difference direction inconsistency, the time synchronization conflict marker, and the local response mismatch area number. The pressure difference mutation node set includes the central node pressure drop abnormal attribute marker, the description of the non-response behavior of adjacent nodes, and the trend direction inconsistency index. The functional degradation status label set includes the structural functional level label, the performance degradation behavior type, and the segmented description of the remaining response capacity. The fire hydrant real-time risk warning information specifically includes the functional lower limit conflict level, the trend mutation intensity level, and the node warning level judgment identifier.

3. The intelligent early warning system for fire hydrant monitoring based on anomaly analysis technology according to claim 2, characterized in that, The flow velocity anomaly identification module includes: The flow velocity baseline extraction submodule acquires the flow velocity monitoring data of the fire hydrant monitoring nodes, and extracts the flow velocity value of each node within the target time range at the current time point in combination with the corresponding sampling time point. It also extracts the median and standard deviation to obtain stable flow velocity statistical indicators. The flow velocity deviation calculation submodule, based on the stable flow velocity statistical index, calls the flow velocity value corresponding to each node at the current time point, calculates the ratio of the absolute value of the deviation of the flow velocity value from the median value to the standard deviation of the fluctuation, calculates the deviation judgment value of the current node within the statistical window, calculates the single-point anomaly identification threshold based on the stable flow velocity statistical index, determines whether the deviation judgment value is greater than the single-point anomaly identification threshold, marks the anomaly, and obtains the suspicious flow velocity deviation identification group. The control behavior exclusion submodule determines whether there is a fire hydrant valve control signal at the corresponding time point based on the suspicious flow rate deviation identifier group, filters out the time period data with control signal behavior, retains the deviation identifier without valve control, and generates flow rate deviation tags.

4. The intelligent early warning system for fire hydrant monitoring based on anomaly analysis technology according to claim 1, characterized in that, The differential pressure trend independence discrimination module includes: The initial feature extraction submodule obtains the location index of the first continuous decrease in water pressure value of each abnormal node within the sampling period based on the flow velocity offset pressure difference response region, and extracts multiple monitoring nodes adjacent to the abnormal node. It also extracts the first response change time index of each adjacent node within the same time segment, calculates the time difference between adjacent nodes and the abnormal node, and generates response time difference information. The trend response calculation submodule calls the response time difference information, calculates the differential value sequence of water pressure change of nodes in a continuous period based on the water pressure sampling value corresponding to each group of nodes, extracts the direction of the differential sequence, marks the direction symbols to form the change direction combination between nodes, judges the synchronicity of the direction combination, and obtains the directional synchronicity feature information. The independence screening and determination submodule determines whether the direction of water pressure change of the abnormal node conflicts with or is missing from the direction of the adjacent node based on the directional synchronization feature information. It then filters the node numbers with directional inconsistency or response delay features, and archives them as abnormal nodes to generate a pressure difference mutation node set.

5. The intelligent early warning system for fire hydrant monitoring based on anomaly analysis technology according to claim 4, characterized in that, The node degradation feature extraction module includes: The operation and maintenance index extraction submodule calls the differential pressure change node set, obtains the fire hydrant number corresponding to each node, extracts the usage frequency, opening and closing response delay, rated service life and number of maintenance records for each number, establishes the operation and maintenance index set corresponding to the node, and generates operation and maintenance parameter summary information. Based on the summarized operation and maintenance parameter information, the performance drift calculation submodule calculates the average difference between the standard time corresponding to the opening and closing operation and the current delay, calculates the rate of change of response delay in adjacent cycles, and calculates and obtains the performance degradation index of the fire hydrant node. The functional level determination submodule, based on the performance degradation index of the fire hydrant node, binds the functional degradation level code and number corresponding to each node according to the equipment remaining performance grading standard, and generates a functional degradation status label set.

6. The intelligent early warning system for fire hydrant monitoring based on anomaly analysis technology according to claim 5, characterized in that, The risk level generation module includes: The trend factor extraction submodule, based on the functional decay state label set and the differential pressure change node set, obtains the functional decay level and the corresponding water pressure trend direction combination of each abnormal node, extracts the continuous time segment of trend change and the time delay value of pressure recovery process, calculates the trend duration period and recovery response duration, and generates an abnormal trend feature parameter set. The tolerance difference calculation submodule, based on the abnormal trend characteristic parameter group, calls the node remaining response capability index corresponding to the function decay level, and calculates and obtains the abnormal tolerance offset value by introducing the abnormal duration period and recovery delay time. The risk level output submodule sets the risk level range based on the abnormal tolerance offset value and the average and standard deviation of the historical abnormal tolerance offset values, divides the risk warning level, encodes the current risk level of each node, and implements fire hydrant risk warning for management personnel to obtain real-time fire hydrant risk warning information.

Images