Fire-fighting water supply pipe network dynamic pressure stabilization control method based on working condition self-recognition

Abstract

The application discloses a kind of dynamic pressure-stabilizing control methods of fire-fighting water supply pipe network based on working condition self-identification, and relates to fire-fighting water supply control technical field.The application obtains pressure time series data and flow switch signal;Through working condition self-identification model, sliding window analysis is carried out, mean, variance and slope are extracted, and working condition is identified as pressure-stabilizing maintaining, suspected leakage, real water use or water hammer disturbance in combination with flow signal;Predictive control module dynamically adjusts pressure-stabilizing strategy according to working condition: when pressure-stabilizing maintaining, it is predicted in advance to compensate pressure, when suspected leakage, pump group is locked and alarm is delayed, when real water use, fire-fighting main pump is started, and when water hammer disturbance, it is shielded and adjusted.The application realizes fine working condition identification and self-adaptive pressure-stabilizing control, effectively avoids misjudgment and misoperation, improves system reliability and intelligent level, and is suitable for pressure-stabilizing control of various fire-fighting water supply systems.

Description

Technical Field

[0001] This invention relates to the field of fire-fighting water supply control technology. More specifically, this invention relates to a dynamic pressure stabilization control method for fire-fighting water supply networks based on self-identification of operating conditions. Background Technology

[0002] Fire-fighting water supply networks are core facilities for fire safety in buildings and industrial sites. Their pressure stabilization systems maintain network pressure within a set range, ensuring sufficient pressure and flow can be quickly provided for firefighting. Pressure stabilization control of fire-fighting water supply networks is crucial for ensuring the reliability of fire-fighting water supply; its core function is to maintain stable network pressure within a preset range, ensuring a rapid response to water supply needs in the event of a fire.

[0003] The existing pressure stabilization control of fire water supply networks mostly adopts fixed threshold control methods, such as fixed value PID regulation or simple pressure upper and lower limit start and stop control. All of these methods collect the network pressure through pressure transmitters. When the pressure is lower than the preset threshold, the pressure stabilizing pump is started to replenish the pressure, and when it is higher than the threshold, the pressure replenishment is stopped.

[0004] However, existing technologies have significant shortcomings: in actual pipeline operation, there are various complex conditions such as normal leakage (e.g., micro-seepage in buried pipelines), fire-fighting water supply, and water hammer impact. These conditions have significant differences in pressure change characteristics, but existing control methods cannot accurately identify these different conditions. They are prone to misjudging minor pipeline leaks and water hammer interference as normal pressure stabilization or fire-fighting water supply, often leading to frequent start-stop of pressure stabilizing pumps, mistakenly starting the main pump as if it were leaking water supply, or exacerbating fluctuations due to incorrect adjustments under water hammer interference. At the same time, pressure stabilization control lacks predictability, only adjusting after the pressure deviates from the threshold, resulting in response lag and difficulty in maintaining stable pipeline pressure. Furthermore, using uniform control for different conditions cannot adapt to the needs of different scenarios such as pressure stabilization, leakage, and water hammer, resulting in low pressure stabilization accuracy.

[0005] The main reason for these problems is that existing control methods do not differentiate operating conditions by combining the characteristics of pipeline pressure changes and flow signals, and the control strategies lack specificity, making it difficult to cope with the complex operating states of pipeline networks. Such problems not only cause energy waste and equipment wear, but may also delay rescue efforts in real fires. Therefore, how to accurately identify pipeline network operating conditions and implement targeted pressure stabilization control has become a technical challenge that urgently needs to be solved in this field. Summary of the Invention

[0006] This invention provides a dynamic pressure stabilization control method for fire water supply networks based on self-identification of operating conditions. It achieves refined operating condition identification and adaptive pressure stabilization control, effectively avoids misjudgment and malfunction, improves system reliability and intelligence, and is applicable to pressure stabilization control of various fire water supply systems.

[0007] To achieve these objectives and other advantages according to the present invention, a dynamic pressure stabilization control method for fire-fighting water supply networks based on self-identification of operating conditions is provided, comprising the following steps: The system acquires real-time pressure timing data from pressure transmitters installed on the fire water supply network at a fixed frequency, and real-time flow switch signals from flow switches installed on the network. The system then transmits the pressure timing data and flow switch signals to a field controller with a built-in working condition self-identification model and predictive control module. The operating condition self-identification model performs sliding window analysis on the received pressure time series data, extracts the pressure mean, pressure fluctuation variance, and pressure change trend slope within the current time window, and combines this with the status of the flow switch signal to identify the current pipeline operating condition as at least one of the following: pressure stabilization, suspected leakage, actual water use, and water hammer interference. The pressure stabilization condition corresponds to a pressure fluctuation variance less than a preset steady-state variance threshold and the flow switch signal being off. The suspected leakage condition corresponds to a pressure change trend slope that is continuously negative and the absolute value of the slope is greater than a preset small leakage slope threshold but less than a fire-fighting water slope threshold, while the flow switch signal is off. The actual water use condition corresponds to a pressure change trend slope that is greater than or equal to a fire-fighting water slope threshold and the flow switch signal changing from off to closed. The water hammer interference condition corresponds to a pressure fluctuation variance greater than or equal to a steady-state variance threshold but the pressure mean has not undergone a unilateral continuous change and the flow switch signal remains off. The predictive control module adjusts the dynamic pressure stabilization control method based on the identified current pipeline network conditions: When the pressure is identified as a stable operating condition, the predictive control module uses the current average pressure as a benchmark and combines it with historical pressure data for the same period to predict the natural pressure decay trend of the next cycle. Based on the prediction results, it adjusts the output frequency of the frequency converter in advance and drives the pressure-stabilizing pump group to replenish pressure in a predictive compensation manner, so that the pipeline pressure is stabilized within the preset dynamic target range. When a suspected leakage condition is identified, the predictive control module locks the start of the pressure stabilizing pump group and controls the timer to start accumulating the suspected leakage duration. After the duration exceeds the preset leakage confirmation delay, the leakage is confirmed and a leakage alarm signal is issued. When a real water usage condition is identified, the predictive control module immediately locks the pressure-stabilizing pump set and issues a command to start the main fire pump. When a water hammer disturbance is identified, the predictive control module maintains the current operating state of the pressure-stabilizing pump group and shields the immediate adjustment response to pressure fluctuations.

[0008] Preferably, before extracting features from the pressure time series data, the working condition self-identification model first performs noise reduction preprocessing on the original pressure time series data based on Kalman filtering or wavelet transform to filter out high-frequency noise components caused by water hammer or electromagnetic interference in the pipeline network, and obtains a smooth pressure curve. The subsequent pressure mean, pressure fluctuation variance and pressure change trend slope are all extracted based on the smooth pressure curve. The operating condition self-identification model adaptively selects the noise reduction algorithm based on the current pipeline noise type: Calculate the noise characteristic indices of the raw pressure time series data within the current sliding window, including the peak value C and the proportion of high-frequency energy E, where the peak value C = p max / p rms p max p represents the maximum pressure within the window. rms The root mean square value of pressure within the window is given. The high-frequency energy ratio E is obtained by calculating the ratio of high-frequency detail coefficient energy to total energy after wavelet decomposition of the original pressure time series data. When the peak coefficient C is greater than the preset peak threshold C th And C th When the value is 3~5, it is determined that the pressure spike noise caused by water hammer impact is dominant, and wavelet transform is preferentially used for noise reduction; when the proportion of high-frequency energy E is greater than the preset energy threshold E th And E th When the value is between 0.3 and 0.5, it is determined that random high-frequency noise caused by electromagnetic interference is dominant, and Kalman filtering is preferred for noise reduction. When using wavelet transform, the db4 wavelet basis is adopted, and the decomposition level is 3-5. Soft thresholding is applied to the high-frequency coefficients after decomposition. The threshold calculation formula is as follows: In the formula, s The noise standard deviation is calculated from the background noise of the raw pressure data. N The number of pressure data points within a single sliding window is used to process the data and then reconstruct a smooth pressure curve using inverse wavelet transform. When Kalman filtering is selected, noise removal is achieved through iterative prediction and updating. The state transition matrix A is 0.98~0.99, the process noise covariance Q is 0.0001~0.0005, and the observation noise covariance R is 0.001~0.005. The values ​​of A, Q, and R are pre-calibrated based on the time constant of pipeline pressure change and the measurement error characteristics of the pressure transmitter.

[0009] Preferably, in the sliding window analysis, the determination of the sliding window length is related to the characteristic parameters of the pipeline network, specifically: For a suspected leakage condition, the sliding window length L1 satisfies L1 ≥ 2×(L0 / v), where L0 is the water supply distance at the farthest end of the pipe network, v is the propagation speed of the pressure wave in the pipe and is determined according to the pipe material and medium temperature of the pipe network, and the value range is 1000 - 1400 m / s; For a water hammer interference condition, the sliding window length L2 satisfies L2 ≤ 0.5 × (L0 / v); For a steady pressure maintenance condition, the sliding window length L3 is taken as L2 < L3 < L1, and the specific value is determined by on-site calibration or autocorrelation analysis of historical pressure data. The on-site calibration collects historical pressure data at three different time periods and takes the window length when the autocorrelation coefficient is the largest. The autocorrelation analysis uses the autocorrelation function with a lag order of 5 - 10 to calculate.

[0010] Preferably, in the process of the predictive control module confirming the occurrence of leakage, a double leakage verification is adopted. The specific steps are as follows: When the working condition self-identification model identifies a suspected leakage condition, the predictive control module locks the start of the pressure stabilizing pump group, and at the same time controls the timer to start accumulating the suspected leakage duration, and synchronously monitors two auxiliary indicators: ① The slope of the pressure change trend in the current time window is k1, and the absolute value of the deviation |k1 - k2| between it and the slope k2 of the pressure change trend in the adjacent previous time window is less than the preset slope consistency threshold Δk. The value of Δk is 0.0005 - 0.001 MPa / min, where the adjacent previous time window is the previous sliding window with the same duration as the current window and no overlap; ② The absolute value of the second derivative of the pressure time series data is continuously less than the preset pressure acceleration threshold a, and the value of a is 0.0001 - 0.0005 MPa / min 2 ; When the water supply distance L0 at the farthest end of the pipe network ≥ 500 m, the predictive control module simultaneously monitors Index ① and Index ②; When the water supply distance L0 at the farthest end of the pipe network < 500 m, the predictive control module only monitors Index ①; When the suspected leakage duration exceeds the preset leakage confirmation delay T1 and all the selected monitoring indicators are satisfied, it is confirmed that leakage has occurred and a leakage alarm signal is issued.

[0011] Preferably, when it is identified as a suspected leakage condition, the predictive control module also calculates an estimated leakage flow value based on the slope value of the pressure change trend in the current time window and the duration of the flow switch signal being in the off state, and issues a leakage alarm signal corresponding to the preset level according to the level to which the estimated leakage flow value belongs; The estimated leakage flow value Q The calculation formula of Q = k · | k 1∣· t , In the formula, Q The estimated leakage flow rate is expressed in cubic meters per second (m³). 3 / h, k This is the leakage coefficient of the pipeline network, in meters. 3 / (h·MPa), determined through on-site calibration based on the pipe diameter and pipe material, with a value range of 0.05~0.2. k 1 represents the slope of the pressure change trend within the current time window, in MPa / min; a negative value indicates pressure decay. t The duration for which the flow switch signal is in the off state, in minutes; The preset leakage flow rate is divided into three levels: Level 1 leakage is 0 < Q ≤ 0.5m 3 / h, a yellow alarm signal is issued, indicating a secondary leakage of 0.5 < Q ≤ 2.0m 3 / h, an orange alarm signal is issued, indicating a level three leak. Q >2.0m 3 / h, a red alarm signal is issued, and staff are simultaneously notified.

[0012] Preferably, when the pressure stabilization condition is identified, the predictive control module acquires the cumulative running time data of multiple pressure stabilizing pumps in the pressure stabilizing pump group during the advance adjustment process, and selects the pressure stabilizing pump with the shortest cumulative running time and no current fault as the main operating pump based on the cumulative running time data, so as to achieve wear balance among pump groups. Assume the pressure-stabilizing pump group has a total of n Taiwan and n ≥ 2, no. i The cumulative running time of the pressure stabilizing pump is T i The unit is h. i = 1,2,..., n Calculate the difference between the cumulative running time of each pump and the average cumulative running time: ,

[0013] In the formula, Let Δ be the average cumulative runtime of all pressure-stabilizing pumps. T iIf the wear difference exceeds the preset threshold and the value ranges from 20 to 50 hours, the pump with the shortest duration will be activated first. Simultaneously, the cumulative runtime of each pump will be updated after each adjustment cycle of 1 to 2 hours. T i When the difference in the cumulative running time between any two pumps exceeds 100 hours, the main and standby pumps will be forcibly switched. Each time the pressure stabilizing pump starts, if the pump with the shortest cumulative running time is in a fault state, the pump with the second shortest running time will be selected in sequence.

[0014] Preferably, the predictive control module includes a parameter self-tuning unit, which can be turned on or off by the operator through a human-machine interface; When the parameter self-tuning unit is activated, it periodically or under specific operating conditions records and analyzes historical pressure fluctuation data of the pipeline network under conditions of no water usage and leakage, and dynamically updates the preset steady-state variance threshold δ based on this historical data. 2 The threshold values ​​for the slope of minute leaks (k3) and the threshold values ​​for the slope of fire-fighting water supply (k4) are also included. The specific update process is as follows: Triggered periodically every 7 days, or triggered when pipeline maintenance is completed, ambient temperature changes exceed ±8℃, or pipeline pressure loss changes exceed 15%, collecting pressure time-series data from the past 24 hours under conditions of no water usage and no leakage in the pipeline network, recorded as follows: p j , j = 1,2,..., M , M For the number of data points; Calculate the mean of the pressure data during this period. , And update the steady-state variance threshold. , In the formula, β is the fluctuation margin coefficient, which ranges from 1.1 to 1.5 and is pre-calibrated according to the measurement error characteristics of the pressure transmitter; Linear fitting was performed on the pressure data during this period to obtain the natural pressure decay slope k5, in MPa / min. The micro-leakage slope threshold k3 was updated to 3·|k5|, and the fire water slope threshold k4 was updated to 10·|k5|. After each threshold update, the parameter self-tuning unit stores the updated threshold in a local cache and compares it with the previous three historical thresholds. If the updated threshold exceeds ±30% of the average of the historical thresholds, the update is deemed invalid, the original threshold is maintained, and an anomaly log is recorded. If the update is valid, the updated threshold must be within a preset reasonable range: steady-state variance threshold δ. 2 The value is 0.001~0.01MPa. 2The threshold for the slope of minute leakage, k3, is -0.01 to -0.001 MPa / min, and the threshold for the slope of fire-fighting water supply, k4, is ≤ -0.1 MPa / min.

[0015] When the parameter self-tuning unit is turned off, the preset steady-state variance threshold δ 2 The threshold values ​​for minute leakage slope k3 and fire water slope k4 are set to factory default values ​​or manually calibrated by staff through the human-machine interface.

[0016] Preferably, when the ambient temperature of the pipeline network changes by less than ±3℃ within 24 hours and the monthly change in pressure loss is less than 5%, the parameter self-tuning unit remains in the off state, and the preset steady-state variance threshold δ is used. 2 The threshold values ​​for minute leakage slope k3 and fire water slope k4 are calibrated manually on a regular basis, with a calibration cycle of 3 to 6 months.

[0017] Preferably, the field controller monitors the operating status of the pressure transmitter, flow switch, frequency converter, and pressure-stabilizing pump set in real time, and executes corresponding fault handling methods based on the monitoring results: When a pressure transmitter fault is detected, the field controller suspends the use of real-time pressure timing data and instead uses the predicted value based on historical pressure data from the same period, and issues a pressure transmitter fault alarm. When a flow switch fault is detected, the field controller forces the flow switch signal to be in an open state, performs operating condition identification separately based on pressure timing data, and issues a flow switch fault alarm. When a frequency converter fault is detected, the field controller immediately locks the corresponding voltage regulator pump driven by the frequency converter. If there are other available frequency converters and voltage regulator pumps in the voltage regulator pump group, it switches to the backup frequency converter and the corresponding voltage regulator pump. If there are no available backup devices, a frequency converter fault alarm is issued and the current voltage regulator pump group status is maintained. When a fault is detected in a pressure-stabilizing pump in the pressure-stabilizing pump group, the field controller, based on wear equalization, skips the faulty pump in subsequent startups and selects the available pump with the second shortest cumulative running time as the main operating pump, while issuing a pressure-stabilizing pump fault alarm.

[0018] Preferably, the field controller has a built-in data storage unit that stores and cyclically overwrites the following data in real time, with a storage period of not less than 90 days: Pressure time-series data and their corresponding timestamps; Flow switch signal and its status change time; The working condition identification results for each time window output by the working condition self-identification model; The predictive control module outputs the inverter's adjustment frequency, the pressure-stabilizing pump's start / stop commands, and the execution time. The cumulative running time of each pump in the pressure-stabilizing pump set and the time of each start-up and shutdown; When a suspected leakage condition is identified, record the duration T1, slope k1, acceleration a, leakage confirmation result, estimated leakage flow rate Q, and corresponding alarm level. When the parameter self-tuning unit is enabled, record the time of each threshold update, the threshold before the update, the threshold after the update, and the basis for the update. The field controller detects all equipment fault types, fault times, recovery times, and fault handling actions performed. The field controller supports querying the above historical data through the human-machine interface or remote monitoring platform, and exporting it in the form of trend curves or reports. The exported data includes the complete process of operating condition changes and control response records.

[0019] The present invention has at least the following beneficial effects: First, this invention collects pressure time-series data and flow switching signals, and combines sliding window analysis to extract the pressure mean, variance, and slope. It accurately identifies pipeline operating conditions into four types: pressure stabilization, suspected leakage, actual water use, and water hammer interference. For each condition, a differentiated pressure stabilization control strategy is adopted. In pressure stabilization, it achieves predictive pressure replenishment; in suspected leakage, it locks the pump unit and delays the alarm; in actual water use, it immediately starts the main fire pump; and in water hammer interference, it shields the adjustment response. This fundamentally solves the problems of erroneous main pump start-up and frequent start-stop of pressure-stabilizing pumps caused by confused operating conditions in traditional control systems. It effectively avoids equipment misoperation, saves energy, and ensures the reliability of fire-fighting water supply.

[0020] Secondly, this invention preprocesses the original pressure time series data with noise reduction based on Kalman filtering or wavelet transform, and adaptively selects the noise reduction algorithm according to the peak coefficient and the proportion of high-frequency energy. This effectively filters out high-frequency noise caused by water hammer impact and electromagnetic interference, making the subsequently extracted pressure mean, variance, and slope closer to the real physical changes. At the same time, the sliding window length is correlated with the distance to the farthest end of the pipeline and the pressure wave velocity, setting an appropriate window length for different working conditions. This ensures the stability of leakage detection, the sensitivity of water hammer capture, and the response speed of pressure stabilization, significantly improving the accuracy and robustness of working condition identification.

[0021] Third, this invention employs dual verification of slope consistency and pressure acceleration during the confirmation of suspected leakage conditions, and selectively activates verification indicators based on the scale of the pipeline network. This effectively distinguishes between minor leaks and instantaneous pressure fluctuations in the pipeline network, reducing the probability of false triggering of leakage alarms. At the same time, it estimates the leakage flow rate by pressure attenuation slope and duration, establishing a three-level alarm mechanism that enables staff to make quick decisions based on the severity of the leakage. Minor leaks can be scheduled for maintenance, while severe leaks can be handled urgently, providing accurate quantitative basis for operation and maintenance.

[0022] Fourth, under the pressure stabilization condition, this invention prioritizes the pressure-stabilizing pump with the shortest cumulative running time and no faults as the main operating pump, and forces rotation when the time difference between pumps is too large, thereby achieving wear balance of the pump group's running time. This effectively prevents excessive wear of a single pump and corrosion of the standby pump due to long-term idleness, and extends the service life of the entire pump group. At the same time, the steady-state variance threshold, the micro-leakage slope threshold, and the fire water slope threshold are dynamically updated through the parameter self-tuning unit, so that the thresholds can adapt to long-term drift due to pipeline aging, environmental changes, etc., maintaining high accuracy of operating condition identification and avoiding the problem of fixed threshold failure.

[0023] Fifth, this invention monitors the operating status of pressure transmitters, flow switches, frequency converters, and pressure-stabilizing pump sets in real time, and implements differentiated fault-tolerant handling strategies for different equipment failures. This ensures that the system can still operate basically even when some equipment fails, guaranteeing the continuity of pressure stabilization control. At the same time, the built-in data storage unit completely records key data such as pressure, flow rate, operating condition identification results, control commands, equipment running time, leakage alarms, and fault events, with a storage time of no less than 90 days. This provides a detailed data foundation for fault backtracking, system optimization, and performance evaluation, significantly improving the reliability and maintainability of the system.

[0024] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Detailed Implementation

[0025] The present invention will now be described in further detail to enable those skilled in the art to implement it based on the description.

[0026] It should be understood that terms such as "having," "comprising," and "including" as used herein do not exclude the presence or addition of one or more other elements or combinations thereof. It should be noted that the experimental methods described in the following embodiments, unless otherwise specified, are conventional methods and therefore should not be construed as limiting the invention.

[0027] This invention provides a dynamic pressure stabilization control method for fire-fighting water supply networks based on self-identification of operating conditions, comprising the following steps: Pressure transmitters are installed at key nodes in the fire water supply network (such as the outlet of the pressure stabilizing pump and the end of the network) to collect network pressure data (reflecting pressure changes over time). Flow switches are installed on the main pipeline to monitor whether there is significant water flow in the network (the contact closes when water flows through the flow switch, otherwise it opens to determine if water is being used). The pressure time-series data is acquired in real time by the pressure transmitters installed on the fire water supply network at a fixed frequency of 1~10Hz, which is adjusted according to the network size and pressure change rate. The acquired pressure time-series data is a continuous sequence of pressure values, containing real-time information on network pressure changes, as well as flow switch signals acquired in real time by the flow switches installed on the network. These signals only include two states: open and closed. The open state indicates no significant water use in the network, while the closed state indicates significant water use. The pressure time-series data and flow switch signals are transmitted to a field controller with a built-in self-identification model and predictive control module. The self-identification model for operating conditions performs sliding window analysis on the received pressure time-series data. The sliding window is a fixed-length time period, the length of which is set according to the subsequent operating condition identification requirements. The window slides continuously, with each slide distance being one data acquisition cycle. The window moves forward continuously over time. Through sliding window analysis, the model extracts the pressure mean (the average of all pressure data within the current window, representing the overall pressure level of the pipeline network within the current window), pressure fluctuation variance (representing the degree of pressure dispersion within the window; the smaller the variance, the more stable the pressure), and pressure change trend slope (obtained by linear fitting of the pressure data within the window, representing the rate and direction of pressure change; a negative slope indicates pressure decay, and a positive slope indicates pressure rise). Combined with the state of the flow switch signal (open or closed), the model identifies the current pipeline network operating condition as at least one of the following: pressure stabilization, suspected leakage, actual water use, and water hammer interference. The pressure stabilization condition corresponds to a state where the pressure fluctuation variance is less than a preset steady-state variance threshold and the flow switch signal is open, indicating that the pipeline network is under pressure stabilization. The network is currently experiencing no water usage or leakage, and the pressure is in a relatively stable state, with only minor natural attenuation or fluctuations. Suspected leakage conditions correspond to a pressure change trend slope that is continuously negative (i.e., the pressure is continuously decreasing) and the absolute value of the slope is greater than the preset minor leakage slope threshold but less than the fire water slope threshold. At the same time, the flow switch signal is in an open state, indicating that there is a slow pressure drop in the pipeline network, but the rate of decrease is insufficient to trigger the fire water judgment, which may indicate a minor leak. Actual water usage conditions correspond to a pressure change trend slope that is greater than or equal to the fire water slope threshold, and the flow switch signal changes from an open state to a closed state, indicating a sharp pressure drop and water flow. This is actual fire water usage (a situation where a large amount of water is needed, such as a fire). Water hammer interference conditions correspond to a pressure fluctuation variance that is greater than or equal to the steady-state variance threshold, but the pressure mean does not change unilaterally (i.e., the pressure fluctuates up and down but the average value remains basically unchanged), and the flow switch signal remains in an open state. This indicates that there is water hammer interference in the pipeline network, such as instantaneous pressure shock waves caused by rapid valve opening and closing or pump start-up and shutdown. This is interference rather than actual water usage. The predictive control module adjusts the dynamic pressure stabilization control method based on the identified current pipeline network conditions: When the system is identified as a pressure stabilization and maintenance condition, the predictive control module uses the current average pressure as a benchmark and combines it with historical pressure data from the same period (such as pressure data from the same period in the past week, by analyzing the changing patterns of historical data) to predict the natural pressure decay trend of the next cycle. Based on the prediction results, the module adjusts the output frequency of the frequency converter in advance to drive the pressure stabilizing pump group to supplement pressure in a predictive compensation manner, so that the pipeline pressure is stabilized within the preset dynamic target range (the dynamic target range is 0.8~1.2MPa), avoiding starting only when the pressure is too low; When a suspected leakage condition is identified, the predictive control module locks the start of the pressure-stabilizing pump unit (to prevent frequent pump start-stops from masking the leakage) and controls the timer to start accumulating the suspected leakage duration. Once the duration exceeds the preset leakage confirmation delay, the leakage confirmation delay ends. The value is 30~60s, which can be flexibly adjusted according to the pipeline network scale and leakage detection accuracy. It is used to distinguish between instantaneous pressure fluctuations and actual leakage, confirm the occurrence of leakage and issue a leakage alarm signal to remind staff to investigate. When a real water usage condition is identified, the predictive control module immediately locks the pressure-stabilizing pump group and issues a command to start the main fire pump to ensure timely water supply. When water hammer interference is detected, the predictive control module maintains the current operating state of the pressure-stabilizing pump group and shields the immediate adjustment response to pressure fluctuations to avoid aggravating the water hammer effect due to frequent adjustments. After the water hammer interference disappears, it automatically resumes normal adjustment.

[0028] In an example of pressure stabilization control of a residential community's fire water supply network. 1. Pipeline parameters: The furthest water supply distance of the fire water supply pipeline in the residential area is L0 = 300m. The pipeline material is galvanized steel pipe. The pressure wave propagation speed is v = 1200m / s. The dynamic target range is set to 0.9~1.1MPa.

[0029] 2. Equipment parameters: The pressure transmitter is installed at the outlet of the pressure-stabilizing pump, and the acquisition frequency is set to 5Hz; the flow switch is installed on the main pipeline, and the contact closing threshold is 5L / s (closes when there is obvious water use); the field controller has a built-in self-identification model of operating conditions and a predictive control module; the pressure-stabilizing pump set contains 2 pressure-stabilizing pumps, and the frequency converter is used to adjust the pump set frequency; the leakage confirmation delay is set to 45s.

[0030] 3. Preset threshold setting: Steady-state variance threshold = 0.005MPa 2 The threshold for minute leakage slope is -0.005 MPa / min, and the threshold for fire water slope is -0.1 MPa / min.

[0031] 4. Data Acquisition Phase The pressure transmitter acquires pressure data from the outlet of the pressure-stabilizing pump in real time at a frequency of 5Hz, forming continuous pressure time-series data (one pressure value is acquired every 0.2s); the flow switch monitors the water flow status of the main pipeline in real time, opening the contact when there is no water use and closing the contact when there is significant water use. Both types of data are transmitted to the field controller in real time via wired transmission.

[0032] 5. Operating condition identification and control implementation (simulating four operating conditions one by one) (1) Pressure stabilization condition: During a certain period, the flow switch remains open. The field controller analyzes the pressure through a sliding window (window length set to 0.5s) and extracts the average pressure within the current window as 1.0MPa and the pressure fluctuation variance as 0.003MPa. 2 The pressure change trend slope is -0.001MPa / min. The operating condition self-identification model identifies it as a pressure stabilization condition. The predictive control module uses 1.0MPa as a benchmark and retrieves the pressure data for the same period in the past week. It finds that the natural pressure decay rate during this period is about 0.002MPa / min. It predicts that the pressure will drop to 0.998MPa in the next cycle (1min). Based on this, the predictive control module adjusts the inverter output frequency in advance and drives the pressure stabilizing pump group to replenish pressure at a low frequency, maintaining the pressure within the dynamic target range of 0.9~1.1MPa. This avoids the pressure dropping below the threshold before pressure replenishment is initiated, thus solving the response lag problem.

[0033] (2) Suspected leakage condition: During a certain period, the flow switch remained in the off state. The sliding window analysis showed that the slope of the pressure change trend was -0.006 MPa / min, and the pressure fluctuation variance was 0.004 MPa. 2 The operating condition self-identification model identified a suspected leakage condition. The predictive control module immediately locked the start of the pressure stabilizing pump group (to prevent the pressure stabilizing pump from covering up the leakage). At the same time, the timer started to accumulate the duration of the suspected leakage. After 45 seconds, the accumulated time of the timer reached the target. The field controller confirmed that the leakage had occurred and issued an audible and visual alarm signal to remind the staff to check for micro-leakage points in the pipeline network. It was found that the leakage was at the buried pipeline interface.

[0034] (3) Real water use conditions: Simulate a fire scenario, fire hydrants are opened for water use, the flow switch detects water flow (greater than 5L / s), and the contact changes from open to closed; at the same time, the sliding window analysis shows that the slope of the pressure change trend is -0.12MPa / min, the working condition self-identification model immediately identifies it as a real water use condition, the predictive control module quickly locks the pressure stabilizing pump group, and at the same time issues a command to start the fire main pump. After the fire main pump starts, it provides sufficient pressure and flow to the pipeline network to ensure the water supply needs of fire rescue.

[0035] (4) Water hammer interference condition: The simulated pipeline valve closes rapidly, causing water hammer interference. At this time, the flow switch remains open. The sliding window analysis shows that the pressure fluctuation variance is 0.006MPa. 2 (Greater than the steady-state variance threshold of 0.005 MPa²), but the average pressure remained around 1.0 MPa (no continuous unilateral change, only fluctuations). The operating condition self-identification model identified it as a water hammer disturbance condition. The predictive control module maintained the current operating state of the pressure-stabilizing pump group, shielding the immediate adjustment response to pressure fluctuations to avoid frequent adjustments exacerbating the water hammer impact. After 30 seconds, the water hammer disturbance disappeared, and the pressure fluctuation variance dropped to 0.003 MPa.2 The predictive control module automatically resumes normal adjustment mode.

[0036] In the above technical solution, by collecting pressure time-series data and flow switch signals, and combining sliding window analysis to extract pressure characteristics, the system accurately identifies four operating conditions: pressure stabilization, suspected leakage, actual water use, and water hammer interference. Differentiated pressure stabilization control is then implemented for each condition, fundamentally solving problems such as accidental start-up of the main pump and frequent start-stop of the pressure-stabilizing pump caused by confusing operating conditions in traditional control methods. In particular, the advanced predictive pressure replenishment under the pressure stabilization condition effectively solves the problem of delayed pressure stabilization response, improving the stability of the pipeline network pressure. Targeted controls such as locking the pressure-stabilizing pump group and starting the fire-fighting main pump prevent equipment malfunctions, save energy, and ensure the reliability of fire-fighting water supply. This solution is suitable for pressure stabilization control scenarios in various fire-fighting water supply networks.

[0037] The original pressure signal time series data contains spike noise caused by water hammer impact and high-frequency noise from electromagnetic interference generated by equipment such as frequency converters and motors, resulting in inaccurate extraction of operating condition identification features and a high misjudgment rate. In another technical solution, before extracting features from the pressure time series data, the operating condition self-identification model first performs noise reduction preprocessing on the original pressure time series data based on Kalman filter or wavelet transform to filter out high-frequency noise components caused by water hammer or electromagnetic interference in the pipeline network and obtain a smooth pressure curve. Kalman filter is mainly used to filter out random high-frequency noise caused by electromagnetic interference, and wavelet transform is mainly used to filter out pressure spike noise caused by water hammer impact. The operating condition self-identification model automatically selects the appropriate noise reduction algorithm according to the noise type of the current pipeline network through an adaptive algorithm selection mechanism. The subsequent pressure mean, pressure fluctuation variance, and pressure change trend slope are all extracted based on the smooth pressure curve. The operating condition self-identification model adaptively selects the noise reduction algorithm based on the current pipeline noise type: Noise characteristic analysis is performed on the raw pressure time series data, and the noise characteristic indicators of the raw pressure time series data within the current sliding window are calculated, including the peak value C and the proportion of high-frequency energy E, where the peak value C = p max / p rms p max p represents the maximum pressure within the window. rmsThe peak value is the root mean square value of pressure within the window. The larger the peak value, the greater the probability of peak signals in the pressure data, which is more likely to be noise caused by water hammer impact. The high-frequency energy ratio E is obtained by performing wavelet decomposition on the original pressure time series data (e.g., using the db4 wavelet basis, decomposing 3 to 5 layers), decomposing the data into high-frequency detail coefficients and low-frequency approximation coefficients, and calculating the ratio of high-frequency detail coefficient energy to total energy (energy of all layer detail coefficients plus approximation coefficient energy). The higher the high-frequency energy ratio, the more random high-frequency noise there is in the pressure data, which is more likely to be noise caused by electromagnetic interference. When the peak coefficient C is greater than the preset peak threshold C th And C th When the value is 3~5, it is determined that the pressure spike noise caused by water hammer impact is dominant, and wavelet transform is preferentially used for noise reduction; when the proportion of high-frequency energy E is greater than the preset energy threshold E th And E th When the value is 0.3~0.5, it is determined that random high-frequency noise caused by electromagnetic interference is dominant, and Kalman filtering is preferred for noise reduction. When both types of noise exist at the same time, the result of the peak coefficient C is taken as the standard, and water hammer peak noise is processed first because water hammer noise has a greater impact on pressure feature extraction. When neither type of noise is satisfied, it means that the noise level is low, and noise reduction can be omitted or the default method can be used. When using wavelet transform, the db4 wavelet basis is employed, which possesses excellent time-domain and frequency-domain localization characteristics, making it suitable for processing abrupt signals. The decomposition level is 3-5 levels, adjustable according to noise intensity; stronger noise necessitates more levels. Soft thresholding is applied to the decomposed high-frequency coefficients, setting the absolute values ​​of coefficients less than the threshold to zero and subtracting the threshold from those greater than the threshold to filter out high-frequency noise. The threshold calculation formula is as follows: To ensure noise reduction while preserving the true trend of pressure data changes, the formula is as follows: s The noise standard deviation is calculated from the background noise of the original pressure data. The calculation method is as follows: the original pressure data is decomposed into a 1-level db4 wavelet basis, and the high-frequency detail coefficients of the first level are taken. d 1. Calculation d The median absolute deviation of 1, MAD = med(∣ d 1( i )-med( d 1)∣), and then obtain the noise standard deviation using σ ≈ MAD / 0.6745. N The number of pressure data points within a single sliding window is determined by soft thresholding, which sets coefficients with absolute values ​​less than λ to zero and shrinks coefficients greater than λ towards zero to suppress noise. After processing, the pressure curve is reconstructed through inverse wavelet transform to obtain a smooth pressure curve. When using Kalman filtering, a state-space model is established, and noise is filtered out through iterative prediction and updating. By establishing the pressure state equation and observation equation, the pressure prediction value is continuously corrected to obtain an accurate pressure estimate. The state transition matrix A takes a value of 0.98~0.99, reflecting the stability of the pipeline pressure state. The closer the value is to 1, the smoother the pressure state change, which is consistent with the characteristics of the pressure stabilization condition of the fire protection pipeline network. The process noise covariance Q takes a value of 0.0001~0.0005, describing the noise intensity generated by the pressure fluctuation of the pipeline itself. The smaller the value, the more stable the pressure of the pipeline itself. The observation noise covariance R takes a value of 0.001~0.005, describing the noise intensity generated by the measurement error of the pressure transmitter. The values ​​of A, Q, and R are pre-calibrated based on the time constant of the pipeline pressure change and the measurement error characteristics of the pressure transmitter. Kalman filtering estimates the real pressure value in real time through prediction and updating iteration, and outputs a smooth curve.

[0038] In the above technical solution, by performing noise reduction preprocessing on the original pressure time series data, high-frequency noise such as water hammer and electromagnetic interference is effectively filtered out, ensuring the smoothness and accuracy of the pressure data. This makes the subsequently extracted pressure mean, variance and slope closer to the real physical changes, adapting to different pipeline noise scenarios. Compared with fixed noise reduction algorithms, the noise reduction effect is better, further improving the accuracy and stability of working condition identification.

[0039] The sliding window length is a crucial parameter for operating condition identification. A fixed sliding window length cannot adapt to the identification needs of different operating conditions. A window that is too short is easily affected by instantaneous fluctuations, while a window that is too long results in a delayed response, leading to low accuracy and slow response speed in operating condition identification. In another technical solution, the determination of the sliding window length in sliding window analysis is related to the characteristic parameters of the pipeline network (the furthest water supply distance L0 at the pipeline end and the pressure wave propagation velocity v), specifically: For suspected leakage conditions, since the pressure decay caused by leakage is a slow process, a sufficiently long window is required to accurately extract the slope of the pressure change trend. The determination of the sliding window length L1 needs to capture the slow pressure drop trend. The window should be long enough to cover at least one complete pressure wave round trip cycle, that is, the time it takes for the pressure wave to travel from the pump station to the far end and then reflect back. This can avoid inaccurate pressure trend extraction due to the window being too short and reduce misjudgment of leakage. The sliding window length L1 satisfies L1≥ 2×(L0 / v), where L0 is the water supply distance of the farthest fire-fighting facility (such as the farthest fire hydrant) in the pipe network, and v is the propagation speed of the pressure wave in the pipe, which is determined according to the pipe material and medium temperature, and the value range is 1000~1400 m / s. For the water hammer interference condition, since the water hammer interference is an instantaneous pressure mutation, a too long window will cause multiple water hammer signals to be superimposed, making it impossible to accurately identify the water hammer condition. The determination of the sliding window length L2 needs to meet the requirement of quickly capturing the pressure fluctuation caused by the water hammer. The sliding window length L2 satisfies L2 ≤ 0.5 × (L0 / v), that is, it does not exceed half of the one-way propagation time, which can quickly capture the instantaneous pressure fluctuation, timely identify the water hammer interference, and avoid the impact of the water hammer on the pressure stabilization control; For the pressure stabilization and maintenance condition, it is necessary to ensure that it can accurately reflect the stable state of the pressure and also have a certain response speed to avoid response lag caused by a too long window and misjudgment of pressure fluctuation caused by a too short window. The value of the sliding window length L3 is L2 < L3 < L1. The specific value is determined by on-site calibration or autocorrelation analysis of historical pressure data. The on-site calibration is to collect historical pressure data at 3 different time periods (such as daytime, nighttime, weekend), calculate the pressure variance stability under different window lengths respectively, and take the window length when the autocorrelation coefficient is the largest. The larger the autocorrelation coefficient, the stronger the correlation of the pressure data within the window and the more it can reflect the stable state of the pressure. The autocorrelation analysis is calculated using the autocorrelation function with a lag order of 5 - 10, and the lag order can be adjusted according to the acquisition frequency of the pressure data to determine the window length that can accurately reflect the stable state of the pressure through autocorrelation analysis.

[0040] In the above technical solution, the sliding window length is associated with the distance to the farthest end of the pipe network and the pressure wave velocity, and a suitable sliding window length is set for different conditions. A longer window is set for the suspected leakage condition to ensure the accurate extraction of the pressure change trend and avoid misjudgment of leakage. A shorter window is set for the water hammer interference condition to achieve rapid identification and response to the water hammer interference. A medium-length window is set for the pressure stabilization and maintenance condition to balance the identification accuracy and response speed, which overall improves the accuracy and adaptability of the condition identification.

[0041] The suspected leakage condition only confirms the leakage based on the duration, and the confirmation method is single, which is easily interfered by instantaneous pressure fluctuations, misjudging the pressure fluctuation as leakage and causing false triggering of the leakage alarm. In another technical solution, when the condition self-identification model determines that it is a suspected leakage condition, the prediction control module does not immediately confirm the leakage. Only when the suspected leakage state lasts for a certain time and meets the preset auxiliary indicators, the leakage is confirmed. The process of the prediction control module confirming the occurrence of leakage adopts double verification of leakage. The specific steps are as follows: When the condition self-identification model identifies a suspected leakage condition, the prediction control module locks the start of the pressure stabilizing pump group to avoid continuous pressure supplementation by the pressure stabilizing pump, covering up the leakage problem and causing the leakage degree to expand. At the same time, it controls the timer to start accumulating the duration of the suspected leakage. The timing starts from the moment when the condition is identified as a suspected leakage until the condition changes or the leakage is confirmed, and two auxiliary indicators are monitored synchronously: ① Consistency of pressure change slope: The slope of the pressure change trend within the current time window is k1, and the slope of the adjacent window (with the same duration as the current window and no overlap) is denoted as k2. The absolute value of the deviation between the pressure change trend slope k1 and the pressure change trend slope k2 of the adjacent window, |k1-k2|, is less than the preset slope consistency threshold Δk. Δk ranges from 0.0005 to 0.001 MPa / min, indicating that the pressure drop rate is stable and consistent, which is a typical characteristic of leakage (the pressure drop caused by leakage is usually relatively uniform). If the slope changes greatly, it may be due to instantaneous interference. ② Pressure acceleration: Calculate the second derivative of the pressure time series data (i.e., the rate of change of pressure, reflecting the acceleration of pressure change). If its absolute value is consistently less than the preset pressure acceleration threshold 'a' within the window, where 'a' ranges from 0.0001 to 0.0005 MPa / min... 2 The smaller the value, the stricter the requirement for pressure acceleration. It can be adjusted according to the accuracy requirements of leakage detection. It indicates that the pressure drop process is smooth and there is no sudden acceleration or deceleration. Using water or water hammer often causes sudden pressure changes and a large acceleration. The verification indicators are selected according to the scale of the pipeline network. For large pipeline networks (the farthest water supply distance L0 ≥ 500m), due to the long pipeline and slow pressure wave propagation, the pressure fluctuation may be more complex. When leakage occurs, the pressure drop may be accompanied by small fluctuations. It is necessary to verify two indicators at the same time to ensure accuracy. The predictive control module monitors indicators ① and ② simultaneously. For small pipe networks (the furthest water supply distance L0 < 500m), pressure waves propagate quickly, and the pressure drop during leakage is almost linear. Pressure fluctuations are relatively simple and can be effectively judged using only indicator ①. Skipping indicator ② can simplify the calculation. The predictive control module only monitors indicator ①. If the suspected leakage lasts for more than the preset leakage confirmation delay T1 and all selected monitoring indicators are met, the leakage is confirmed and a leakage alarm signal is issued. If any condition is not met during this period, the timer is reset, the system re-monitors, and the leakage confirmation process is canceled.

[0042] In the above technical solution, a dual leakage verification method is adopted, combining two auxiliary indicators: pressure slope consistency and pressure acceleration. This effectively distinguishes between minor leaks and pressure fluctuations in the pipeline network, reduces the probability of false alarms, and improves the accuracy of leakage confirmation. Different verification indicators are adapted according to the differences in pipeline network size. For small pipeline networks, the verification process is simplified, reducing the system's computational load, while for large pipeline networks, the verification requirements are strengthened to ensure reliable leakage identification. By setting a leakage confirmation delay, false alarms caused by instantaneous pressure fluctuations are further avoided, providing staff with accurate leakage early warning information.

[0043] Suspected leakage conditions can only trigger an alarm signal, but cannot quantify the degree of leakage. In another technical solution, when a suspected leakage condition is identified, the predictive control module, while locking the pressure-stabilizing pump group and starting the timer, begins to calculate the estimated leakage flow rate, which approximately reflects the actual leakage situation of the pipeline network. Specifically, it is based on the slope value of the pressure change trend within the current time window. k 1. The duration during which the flow switch signal is in the off state. t Calculate an estimated leakage flow rate. Q And based on the preset level to which the leakage flow rate value belongs, a corresponding leakage alarm signal is issued; Estimated leakage flow rate Q slope value of pressure change trend k 1. Duration of disconnection t and pipeline leakage coefficient k Positive correlation: the faster the pressure decays and the longer the disconnection lasts, the higher the leakage coefficient of the pipeline network. k The larger the volume, the higher the estimated leakage flow rate. Q The larger the value, the higher the estimated leakage flow rate. Q The calculation formula is: Q = k · | k 1∣· t , In the formula, Q The estimated leakage flow rate is expressed in cubic meters per second (m³). 3 / h, k This is the leakage coefficient of the pipeline network, in meters. 3 / (h·MPa), determined through on-site calibration based on the pipe diameter and material, is used to perform tests at known leakage points, measuring the pressure drop rate and actual leakage flow rate, and then calculating the pressure. k The larger the pipe diameter, the worse the pipe's sealing performance. k The larger the value, the greater the range, from 0.05 to 0.2. k 1 represents the slope of the pressure change trend within the current time window, in MPa / min. A negative value indicates pressure decay; the larger the absolute value, the faster the pressure decays, and the more severe the leakage may be. t The duration for which the flow switch signal is in the off state is accumulated from the moment the flow switch changes to the off state until the flow switch state changes or leakage is confirmed, in minutes. The preset leakage flow rate is divided into three levels: Level 1 leakage is 0 < Q ≤ 0.5m 3 At this point, the leakage is minor and has little impact on the pipeline network. A yellow alarm signal is issued to remind staff to regularly check for leaks. No emergency treatment is required. Secondary leakage is defined as 0.5 < Q ≤ 2.0m3 At this point, the leakage is moderate, and if not addressed promptly, it may worsen, triggering an orange alarm signal to alert staff to investigate and handle the situation as soon as possible. Level 3 leakage is... Q >2.0m 3 / h, at which point the leakage is severe and may cause a sharp drop in pipeline pressure, affecting the safety of fire water supply. A red alarm signal will be issued and a staff linkage notification will be triggered simultaneously. The leakage information will be pushed to the staff's mobile APP or sent by SMS to ensure that the staff can arrive at the scene in time to deal with the situation and minimize the losses caused by the leakage. Throughout the process, the estimated leakage flow rate is updated in real time. After each sliding window analysis is completed, the calculation is repeated to ensure the real-time quantification of leakage. At the same time, the estimated leakage flow rate, the corresponding time, and the alarm level are recorded for each calculation.

[0044] In the above technical solution, leakage flow is estimated by pressure attenuation slope and duration, and a three-level alarm mechanism is adopted. Different leakage levels correspond to different alarm signals and handling methods, enabling staff to make quick decisions based on the severity of leakage (such as planned maintenance for minor leakage and emergency handling for severe leakage), ensuring the safe operation of the pipeline network.

[0045] Under pressure stabilization conditions, the operation of the pressure stabilizing pump group is uneven, and some pump bodies wear out too quickly and their service life is shortened. In another technical solution, when the pressure stabilization condition is identified, the predictive control module acquires the cumulative running time data of multiple pressure stabilizing pumps in the pressure stabilizing pump group during the process of making advance adjustments. Based on the cumulative running time data, the pressure stabilizing pump with the shortest cumulative running time and no current fault is selected as the main operating pump to achieve wear balance among the pump groups. Assume the pressure-stabilizing pump group has a total of n Taiwan and n ≥ 2, no. i The cumulative running time of the pressure stabilizing pump is T i The unit is h. i = 1,2,..., n Calculate the difference between the cumulative running time of each pump and the average cumulative running time: ,

[0046] In the formula, The average cumulative runtime of all pressure-stabilizing pumps. Calculate the cumulative runtime of each pressure-stabilizing pump. T i With average cumulative runtime The larger the absolute value of the difference, the greater the deviation of the pump's operating time from the average level, and the more it needs to be prioritized to balance wear. If the Δ value of a certain pressure-stabilizing pump is... T i If the wear difference exceeds the preset threshold and the value ranges from 20 to 50 hours, the wear deviation of the pump is considered to be large. The pump with the shortest duration should be used first to balance the wear. It is also necessary to consider whether the pump is faulty. Each time the pressure stabilizing pump needs to be started, the controller first filters out the currently available pumps that are not faulty, and then selects the pump with the shortest cumulative running time as the main running pump. If the pump with the shortest duration is faulty, the pump with the second shortest duration is selected, and so on. At the same time, the cumulative runtime of each pump is updated after each adjustment cycle is completed. T i The adjustment cycle is 1-2 hours. If the adjustment cycle is too short, it will increase the system's computational load. If it is too long, it will be unable to balance wear in time. When the difference in the cumulative running time of any two pumps exceeds 100 hours, the main and standby pumps will be forcibly switched. Even if the main pump is still running, it will be switched to the standby pump and the pump with the shorter running time will be activated to avoid excessive wear of a single pump. For example, a pressure-stabilizing pump group contains two pressure-stabilizing pumps, namely pump 1 and pump 2. The cumulative running time of pump 1 is 100 hours, the cumulative running time of pump 2 is 80 hours, and the average cumulative running time is 90 hours. The difference between the two is 20 hours. If the wear difference threshold is set to 20 hours, then pump 2 will be started first. Every 1.5-hour adjustment cycle, the running time is updated. If pump 2 runs for 1.5 hours and the cumulative running time becomes 81.5 hours, while pump 1 is still 100 hours, the difference becomes 18.5 hours. At this time, pump 2 will no longer be started first, and the normal selection mode will be restored.

[0047] In the above technical solution, by prioritizing the pressure-stabilizing pump with the shortest cumulative running time as the main operating pump, the balanced distribution of pump set running time is achieved, solving the problem of excessive wear of some pump bodies. When the time difference between pumps is too large, forced rotation is performed. At the same time, fault conditions are considered to effectively prevent excessive wear of single pumps and corrosion of standby pumps due to long-term idleness, thereby extending the service life of the entire pump set and reducing replacement costs.

[0048] The preset steady-state variance threshold, micro-leakage slope threshold, and fire water slope threshold are key parameters for operating condition identification. As the pipeline network ages, scales form, and the ambient temperature changes, these thresholds may deviate from their optimal values, leading to a decrease in identification accuracy. Fixed preset thresholds for each operating condition cannot adapt to changes in the pipeline environment and equipment status, resulting in decreased operating condition identification accuracy and poorer pressure stabilization effect. In another technical solution, the predictive control module includes a parameter self-tuning unit, which can be turned on or off by the operator through a human-machine interface. When the parameter self-tuning unit is activated, it periodically or under specific operating conditions records and analyzes historical pressure fluctuation data of the pipeline network under conditions of no water usage and leakage, and dynamically updates the preset steady-state variance threshold δ based on this historical data. 2 The threshold values ​​for the slope of minute leaks (k3) and the threshold values ​​for the slope of fire-fighting water supply (k4) are also included. The specific update process is as follows: Periodic trigger: Triggered once every 7 days, or Event triggering: Triggered when pipeline maintenance is completed, ambient temperature changes exceed ±8℃, or pipeline pressure loss changes exceed 15%. These conditions can all lead to changes in pipeline pressure characteristics. After triggering, pressure time-series data of the pipeline network under no water usage and no leakage conditions over the past 24 hours are collected. The flow switch signal and pressure change trend are used to determine whether the pipeline network is in a state of no water usage and no leakage. Specifically, if the flow switch remains continuously open and the pressure fluctuation variance is less than 1.5 times the current steady-state variance threshold for more than one consecutive hour, it can be considered to be in a stable state. The collected pressure time-series data are recorded as a series of continuous pressure values, denoted as... p j , j = 1,2,..., M The number of data points is M , M The size is determined by the sampling frequency and sampling duration. The sampling frequency is 1~10Hz, and the number of sampling points in 24 hours is 86,400~864,000. The data filtering conditions are that the flow switch is kept open and there is no unilateral change in pressure, to ensure that the data can truly reflect the natural fluctuation characteristics of the pipeline network. Calculate the mean of the pressure data during this period. , And update the steady-state variance threshold. , In the formula, β is the fluctuation margin coefficient, which ranges from 1.1 to 1.5. It is a safety margin reserved for normal pressure fluctuations and is pre-calibrated according to the measurement error characteristics of the pressure transmitter. After updating the steady-state variance threshold, the pressure data within this period is linearly fitted to obtain the pressure natural decay slope k5, which reflects the sealing performance of the pipeline network and the characteristics of natural pressure loss. The unit is MPa / min. The micro-leakage slope threshold k3 is updated to 3·|k5|, and the fire water slope threshold k4 is updated to 10·|k5|. This ensures that the sharp pressure drop caused by real fire water can be quickly and accurately identified, avoiding confusion with leakage and interference. After each threshold update, the parameter self-tuning unit stores the updated threshold in a local cache and compares it with the previous three historical thresholds. If the updated threshold exceeds ±30% of the historical threshold average, the update is deemed invalid, automatically discarded, and the original threshold is maintained, with an exception log recorded. If the update is valid, the updated threshold must be within a preset reasonable range: steady-state variance threshold δ. 2 The value is 0.001~0.01MPa. 2 The threshold for the slope of minute leakage, k3, is -0.01 to -0.001 MPa / min, and the threshold for the slope of fire-fighting water supply, k4, is ≤ -0.1 MPa / min.

[0049] When the parameter self-tuning unit is turned off, the preset steady-state variance threshold δ 2 The threshold values ​​for minute leakage slope k3 and fire water slope k4 are set to factory default values ​​or manually calibrated by staff through the human-machine interface.

[0050] In the above technical solution, the steady-state variance threshold, the micro-leakage slope threshold, and the fire water slope threshold are dynamically updated through the parameter self-tuning unit. This solves the problem that fixed thresholds cannot adapt to changes in the pipeline environment and equipment status, improving the accuracy of operating condition identification and the system's adaptability. Two update methods are set: periodic triggering and specific operating condition triggering. This ensures the timeliness of threshold updates while avoiding unnecessary update operations and reducing system energy consumption. Threshold validity verification and reasonable range limitation ensure the reliability of the updated thresholds and prevent abnormal thresholds from affecting system operation. A manual calibration backup scheme is provided, improving the system's flexibility and reliability and adapting to different operation and maintenance needs.

[0051] The system can be configured to enable a parameter self-tuning unit, allowing the system to dynamically update core thresholds (steady-state variance, micro-leakage slope, and fire water slope) based on actual pipeline network operating data (times without water usage). This enables the thresholds to adapt to long-term drift caused by pipeline aging, scaling, and changes in ambient temperature, maintaining high accuracy in operating condition identification and avoiding the problem of fixed threshold failure.

[0052] In practical applications, some fire protection pipe networks are located indoors, where the ambient temperature remains stable year-round, and the pressure loss of the pipe network changes very little over time. For such stable scenarios, frequent self-tuning may be unnecessary and could even increase the system burden. In another technical solution, when the ambient temperature of the pipe network changes by less than ±3℃ within 24 hours and the monthly change in pressure loss is less than 5%, the environment in which the pipe network is located is considered stable, its own operating state changes smoothly, and there is no need for frequent threshold updates. The parameter self-tuning unit remains in the off state, and the preset steady-state variance threshold δ is used. 2 The threshold values ​​for minute leakage slope k3 and fire water slope k4 are calibrated manually on a regular basis, with a calibration cycle of 3 to 6 months.

[0053] In the above technical solution, the system remains in a closed state when the pipeline environment and pressure loss change little, avoiding unnecessary threshold updates, reducing system computational load and energy consumption. The manual calibration cycle and requirements are set to ensure that the threshold remains accurate when the self-tuning unit is closed, thus ensuring the effectiveness of operating condition identification and pressure stabilization control. Through regular manual calibration, threshold deviations can be corrected in a timely manner to adapt to subtle changes in the pipeline network during long-term operation, further improving the stability and reliability of the system.

[0054] In another technical solution, the field controller monitors the operating status of the pressure transmitter, flow switch, frequency converter, and pressure-stabilizing pump set in real time. Monitoring methods include: periodic self-testing, signal anomaly detection (such as signal exceeding the range or prolonged lack of change), and communication interruption detection. Based on the monitoring results, corresponding fault handling methods are executed. When a pressure transmitter malfunction is detected, the malfunction criteria include signal interruption, output value overflow, long-term constant pressure, and exceeding the reasonable pressure range of the pipeline network. The field controller suspends the use of real-time pressure time-series data and instead uses the predicted value based on historical pressure data of the same period. The historical pressure data of the same period is selected as the average pressure and trend of the same period in the past 7 days, and a pressure transmitter malfunction alarm is issued. When a flow switch fault is detected, the fault criteria include long-term signal jamming, no reasonable state transition, continuous false triggering or continuous non-action. The field controller forces the flow switch signal to be judged as disconnected, and completes the working condition identification entirely based on the mean, variance and slope characteristics of the pressure time series data. It no longer relies on the flow signal to assist in the judgment, and performs working condition identification independently based on the pressure time series data, while issuing a flow switch fault alarm. When a frequency converter fault is detected, the fault criteria include communication interruption, overcurrent, overload, and start-up failure. The field controller immediately locks the corresponding voltage regulator pump driven by the frequency converter, prohibiting it from driving the corresponding voltage regulator pump. If there are other available frequency converters and voltage regulator pumps in the voltage regulator pump group, it switches to the backup frequency converter and the corresponding voltage regulator pump. If there is no available backup equipment, a frequency converter fault alarm is issued and the current voltage regulator pump group status is maintained. When a fault is detected in a pressure-stabilizing pump in the pressure-stabilizing pump group, the fault criteria include overload, overheating, start-up timeout, and abnormal noise feedback. Based on wear balance, the field controller skips the faulty pump in subsequent starts and no longer includes it in the main operating pump selection range. It selects the second shortest cumulative running time from the available pumps as the main operating pump to ensure that the pump group balancing logic is not affected by the fault. At the same time, a pressure-stabilizing pump fault alarm is issued. All fault handling actions are completed within 1 second to ensure that the pressure stabilization control of the fire protection pipeline network is not interrupted. After the fault is recovered, the system automatically releases the lockout and shield and restores the normal full equipment operation mode.

[0055] In the above technical solution, by monitoring the operating status of various equipment in real time, timely detection of equipment failures and execution of corresponding handling methods are achieved, which solves the problem of the inability to respond to equipment failures in a timely manner and the resulting system paralysis. This ensures the continuous and stable operation of the pressure stabilization system, and differentiates the handling of different equipment failures, further improving the reliability of the pressure stabilization control of the fire water supply network.

[0056] In another technical solution, the field controller has a built-in data storage unit that stores and cyclically overwrites the following data in real time, with a storage period of no less than 90 days. All key operational data are written and cyclically overwritten in chronological order, and the stored content comprehensively covers four categories: operating condition identification, control output, equipment operation, and fault alarms. Pressure time series data and its corresponding timestamps: Pressure time series data and timestamps are stored synchronously, recording the pipeline pressure value and acquisition time at each moment; Flow switch signal and its state change time: Record the two states of open and closed, and the exact time of each state change; The working condition identification results of each time window output by the self-identification model: record the identification results of each sliding window, and clearly mark the four working conditions: pressure stabilization, suspected leakage, actual water use, and water hammer interference; The predictive control module outputs the inverter's adjustment frequency, the pressure-stabilizing pump's start / stop commands, and the execution time. The operating data of the pressure stabilizing pump set is stored separately, including the cumulative running time of each pump and the time of each start-up and shutdown; When a suspected leakage condition is identified, record the duration T1, slope k1, acceleration a, leakage confirmation result, estimated leakage flow rate Q, and corresponding alarm level. When the parameter self-tuning unit is enabled, record the time of each threshold update, the threshold before the update, the threshold after the update, and the basis for the update. The field controller detects all equipment fault types, fault times, recovery times, and fault handling actions performed. Data storage adopts a structured format, supporting fast retrieval and filtering. The field controller provides two data access methods: the field controller supports querying the above historical data through the human-machine interface or remote monitoring platform. The query can be filtered by time, operating condition type, fault type, threshold update event, and exported in the form of trend curves or reports. Specifically, continuous data such as pressure, frequency, and runtime are displayed as trend curves, while discrete data such as operating conditions, alarms, faults, and commands are displayed as reports. The exported data contains a complete operating condition change process and control response records, which can be directly used for operation and maintenance review, fault analysis, and fire protection acceptance. When the storage unit approaches its capacity limit, it automatically performs a cyclic overwrite, deleting the complete data of the earliest day to ensure that the latest 90 days of data are complete and available.

[0057] In the above technical solution, key data such as pressure, flow rate, operating condition identification results, control commands, equipment running time, leakage alarms and fault events are fully recorded through the built-in data storage unit, providing a reliable basis for fault tracing, facilitating staff to analyze the cause of faults and formulate targeted optimization measures. The storage time of no less than 90 days is set to ensure the integrity and traceability of the data, thereby improving the practicality and reliability of the system.

[0058] This invention provides a dynamic pressure stabilization control system for fire-fighting water supply networks based on self-identification of operating conditions, comprising: Pressure transmitters are installed at key nodes in fire water supply networks (such as the outlet of pressure-stabilizing pumps and the end of the network) to collect pressure time-series data in real time at a fixed frequency of 1~10Hz. This data is a continuous numerical sequence that reflects the dynamic change process of network pressure. A flow switch is installed on the pipeline to collect flow switch signals in real time. The signal has only two states: open (no water use) and closed (water use). It is used to help determine whether there is a significant water use event. The field controller has a built-in self-identification model for operating conditions and a predictive control module. It receives data from pressure transmitters and flow switches, performs real-time analysis and decision-making based on the built-in model, and outputs control commands. Specifically, the self-identification model performs sliding window analysis on the received pressure time-series data, extracting the average pressure, pressure fluctuation variance, and pressure change trend slope within the current time window. Combined with the status of the flow switch signal, it identifies the current pipeline operating condition as at least one of the following: pressure stabilization, suspected leakage, actual water usage, and water hammer disturbance. The predictive control module adjusts the dynamic pressure stabilization control mode based on the identified current pipeline operating condition. The frequency converter, controlled by the field controller, drives the pressure-stabilizing pump group to operate by adjusting the output frequency, thereby achieving precise regulation of the pipeline pressure; A pressure-stabilizing pump set, including at least one pressure-stabilizing pump, is used to replenish the pressure of the pipeline network under the drive of a frequency converter to maintain stable pressure; The main fire pump is an independent water supply pump set used to start water supply under actual water use conditions. It is started by command from the field controller to provide a large flow of water for fire fighting.

[0059] The on-site controller's built-in operating condition self-identification model performs sliding window analysis on the received pressure time-series data. Within each sliding window, the model extracts three core features: Average pressure: The average value of all pressure data within the window, reflecting the current overall pressure level of the pipeline network; Pressure fluctuation variance: The degree of dispersion of pressure data within a window, reflecting the stability of pressure; Pressure change trend slope: obtained by linear fitting of pressure data within the window, reflecting the rate and direction of pressure change (negative values ​​indicate decay, positive values ​​indicate increase).

[0060] The model combines the above characteristics with the state (open / closed) of the flow switch signal, and based on preset threshold conditions, accurately identifies the current pipeline network operating condition into one of the following four types: Pressure stabilization condition: The pressure fluctuation variance is less than the steady-state variance threshold, and the flow switch is off, indicating that there is no water usage or leakage in the pipeline network, and it is in a stable state. Suspected leakage condition: The slope of the pressure change trend is continuously negative, and its absolute value is greater than the minor leakage slope threshold but less than the fire water slope threshold. In addition, the flow switch is open, indicating that there is a slow pressure decay in the pipeline network, and minor leakage may occur. Real-world water usage conditions: The absolute value of the slope of the pressure change trend is greater than or equal to the slope threshold for fire-fighting water, and the flow switch changes from open to closed, indicating a sharp drop in pressure and water flow, which is real-world fire-fighting water usage. Water hammer disturbance condition: The pressure fluctuation variance is greater than or equal to the steady-state variance threshold, but the pressure mean does not change continuously on one side, and the flow switch remains open, indicating the presence of instantaneous disturbances such as water hammer impact.

[0061] The predictive control module dynamically adjusts the voltage regulation control method based on the identification results: Pressure stabilization mode: Based on the current average pressure, combined with historical pressure data for the same period, predict the natural decay trend of the next cycle, adjust the output frequency of the frequency converter in advance, drive the pressure stabilizing pump group to replenish pressure in a predictive compensation manner, so that the pipeline pressure is stabilized within the preset dynamic target range (such as 0.8~1.2MPa). Suspected leakage: Immediately shut down the start of the pressure stabilizing pump unit and start a timer to accumulate the duration of suspected leakage. Once the duration exceeds the leakage confirmation delay (e.g., 30~60s), leakage is confirmed and an alarm signal is issued. Real-world water usage scenario: Immediately shut down the pressure-stabilizing pump unit and simultaneously issue a command to start the main fire pump to ensure fire water supply; Water hammer interference condition: Maintain the current operating status of the pressure stabilizing pump group, shield the immediate adjustment response to pressure fluctuations, and automatically restore normal adjustment after the interference disappears.

[0062] The number of devices and processing scale described herein are for the purpose of simplifying the description of the invention. Applications, modifications, and variations of the invention will be readily apparent to those skilled in the art.

[0063] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. Other modifications can be easily made by those skilled in the art. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details shown and described herein.

Claims

1. A dynamic pressure stabilization control method for fire-fighting water supply networks based on self-identification of operating conditions, characterized in that, Includes the following steps: The system acquires real-time pressure timing data from pressure transmitters installed on the fire water supply network at a fixed frequency, and real-time flow switch signals from flow switches installed on the network. The system then transmits the pressure timing data and flow switch signals to a field controller with a built-in working condition self-identification model and predictive control module. The operating condition self-identification model performs sliding window analysis on the received pressure time series data, extracts the pressure mean, pressure fluctuation variance, and pressure change trend slope within the current time window, and combines this with the status of the flow switch signal to identify the current pipeline operating condition as at least one of the following: pressure stabilization, suspected leakage, actual water use, and water hammer interference. The pressure stabilization condition corresponds to a pressure fluctuation variance less than a preset steady-state variance threshold and the flow switch signal being off. The suspected leakage condition corresponds to a pressure change trend slope that is continuously negative and the absolute value of the slope is greater than a preset small leakage slope threshold but less than a fire-fighting water slope threshold, while the flow switch signal is off. The actual water use condition corresponds to a pressure change trend slope that is greater than or equal to a fire-fighting water slope threshold and the flow switch signal changing from off to closed. The water hammer interference condition corresponds to a pressure fluctuation variance greater than or equal to a steady-state variance threshold but the pressure mean has not undergone a unilateral continuous change and the flow switch signal remains off. The predictive control module adjusts the dynamic pressure stabilization control method based on the identified current pipeline network conditions: When the pressure is identified as a stable operating condition, the predictive control module uses the current average pressure as a benchmark and combines it with historical pressure data for the same period to predict the natural pressure decay trend of the next cycle. Based on the prediction results, it adjusts the output frequency of the frequency converter in advance and drives the pressure-stabilizing pump group to replenish pressure in a predictive compensation manner, so that the pipeline pressure is stabilized within the preset dynamic target range. When a suspected leakage condition is identified, the predictive control module locks the start of the pressure stabilizing pump group and controls the timer to start accumulating the suspected leakage duration. After the duration exceeds the preset leakage confirmation delay, the leakage is confirmed and a leakage alarm signal is issued. When a real water usage condition is identified, the predictive control module immediately locks the pressure-stabilizing pump set and issues a command to start the main fire pump. When a water hammer disturbance is identified, the predictive control module maintains the current operating state of the pressure-stabilizing pump group and shields the immediate adjustment response to pressure fluctuations.

2. The dynamic pressure stabilization control method for fire-fighting water supply networks based on self-identification of operating conditions as described in claim 1, characterized in that, Before extracting features from the pressure time series data, the working condition self-identification model first performs noise reduction preprocessing on the original pressure time series data based on Kalman filtering or wavelet transform to filter out high-frequency noise components caused by water hammer or electromagnetic interference in the pipeline network and obtain a smooth pressure curve. The subsequent pressure mean, pressure fluctuation variance and pressure change trend slope are all extracted based on the smooth pressure curve. The operating condition self-identification model adaptively selects the noise reduction algorithm based on the current pipeline noise type: Calculate the noise characteristic indices of the raw pressure time series data within the current sliding window, including the peak value C and the proportion of high-frequency energy E, where the peak value C = p max / p rms p max p represents the maximum pressure within the window. rms The root mean square value of pressure within the window is given. The high-frequency energy ratio E is obtained by calculating the ratio of high-frequency detail coefficient energy to total energy after wavelet decomposition of the original pressure time series data. When the peak coefficient C is greater than the preset peak threshold C th And C th When the value is 3~5, it is determined that the pressure spike noise caused by water hammer impact is dominant, and wavelet transform is preferentially used for noise reduction; when the proportion of high-frequency energy E is greater than the preset energy threshold E th And E th When the value is between 0.3 and 0.5, it is determined that random high-frequency noise caused by electromagnetic interference is dominant, and Kalman filtering is preferred for noise reduction. When using wavelet transform, the db4 wavelet basis is adopted, and the decomposition level is 3-5. Soft thresholding is applied to the high-frequency coefficients after decomposition. The threshold calculation formula is as follows: In the formula, σ The noise standard deviation is calculated from the background noise of the raw pressure data. N The number of pressure data points within a single sliding window is used to process the data and then reconstruct a smooth pressure curve using inverse wavelet transform. When Kalman filtering is selected, noise removal is achieved through iterative prediction and updating. The state transition matrix A is 0.98~0.99, the process noise covariance Q is 0.0001~0.0005, and the observation noise covariance R is 0.001~0.

005. The values ​​of A, Q, and R are pre-calibrated based on the time constant of pipeline pressure change and the measurement error characteristics of the pressure transmitter.

3. The dynamic pressure stabilization control method for fire-fighting water supply networks based on self-identification of operating conditions as described in claim 1, characterized in that, In sliding window analysis, the determination of the sliding window length is related to the characteristic parameters of the pipeline network, specifically: For suspected leakage conditions, the sliding window length L1 satisfies L1 ≥ 2×(L0 / v), where L0 is the water supply distance at the farthest end of the pipe network, and v is the propagation speed of the pressure wave in the pipe, which is determined according to the pipe material and medium temperature, and the value range is 1000~1400 m / s. For water hammer interference conditions, the sliding window length L2 satisfies L2 ≤ 0.5 × (L0 / v); For the pressure stabilization operation, the sliding window length L3 is L2 < L3 < L1. The specific value is determined by field calibration or by autocorrelation analysis based on historical pressure data. Field calibration is performed by collecting three sets of historical pressure data from different time periods and taking the window length when the autocorrelation coefficient is at its maximum. The autocorrelation analysis is performed using an autocorrelation function with a lag order of 5 to 10.

4. The dynamic pressure stabilization control method for fire-fighting water supply networks based on self-identification of operating conditions according to claim 1, characterized in that, The predictive control module confirms the leakage process by employing dual leakage verification, with the following specific steps: When the operating condition self-identification model identifies a suspected leakage condition, the predictive control module locks the start of the pressure-stabilizing pump unit, simultaneously controls the timer to start accumulating the suspected leakage duration, and monitors two auxiliary indicators concurrently: ① The slope of the pressure change trend within the current time window is k1, and the absolute value of the deviation |k1-k2| from the slope of the pressure change trend of the adjacent previous time window is less than the preset slope consistency threshold Δk, where Δk is 0.0005~0.001MPa / min, and the adjacent previous time window is the previous sliding window with the same duration as the current window and without overlap. ② The absolute value of the second derivative of the pressure time series data is continuously less than the preset pressure acceleration threshold 'a' within the window, where 'a' ranges from 0.0001 to 0.0005 MPa / min. 2 ; When the furthest water supply distance L0 in the pipeline network is ≥ 500m, the predictive control module monitors both indicator ① and indicator ② simultaneously. When the furthest water supply distance L0 in the pipeline network is less than 500m, the predictive control module only monitors indicator ①. When the suspected leakage lasts for more than the preset leakage confirmation delay T1, and all selected monitoring indicators are met, the leakage is confirmed and a leakage alarm signal is issued.

5. The dynamic pressure stabilization control method for fire-fighting water supply networks based on self-identification of operating conditions according to claim 1, characterized in that, When a suspected leakage condition is identified, the predictive control module also calculates an estimated leakage flow rate based on the slope value of the pressure change trend within the current time window and the duration of the flow switch signal being in the off state, and issues a corresponding leakage alarm signal according to the preset level to which the leakage flow rate value belongs. Estimated leakage flow rate Q The calculation formula is: Q = k ·∣ k 1∣· t ， In the formula, Q The estimated leakage flow rate is expressed in cubic meters per second (m³). 3 / h, k This is the leakage coefficient of the pipeline network, in meters. 3 / (h·MPa), determined through on-site calibration based on the pipe diameter and pipe material, with a value range of 0.05~0.

2. k 1 represents the slope of the pressure change trend within the current time window, in MPa / min; a negative value indicates pressure decay. t The duration for which the flow switch signal is in the off state, in minutes; The preset leakage flow rate is divided into three levels: Level 1 leakage is 0 < Q ≤ 0.5m 3 / h, a yellow alarm signal is issued, indicating a secondary leakage of 0.5 < Q ≤ 2.0m 3 / h, an orange alarm signal is issued, indicating a level three leak. Q > 2.0m 3 / h, a red alarm signal is issued, and staff are simultaneously notified.

6. The dynamic pressure stabilization control method for fire-fighting water supply networks based on self-identification of operating conditions according to claim 1, characterized in that, When the pressure stabilization condition is identified, the predictive control module acquires the cumulative running time data of multiple pressure stabilizing pumps in the pressure stabilizing pump group during the advance adjustment process. Based on the cumulative running time data, the pressure stabilizing pump with the shortest cumulative running time and no current fault is selected as the main operating pump to achieve wear balance among pump groups. Assume the pressure-stabilizing pump group has a total of n Taiwan and n ≥ 2, no. i The cumulative running time of the pressure stabilizing pump is T i The unit is h. i = 1,2,..., n Calculate the difference between the cumulative running time of each pump and the average cumulative running time: ， In the formula, Let Δ be the average cumulative runtime of all pressure-stabilizing pumps. T i If the wear difference exceeds the preset threshold and the value ranges from 20 to 50 hours, the pump with the shortest duration will be activated first. Simultaneously, the cumulative runtime of each pump will be updated after each adjustment cycle of 1 to 2 hours. T i When the difference in the cumulative running time between any two pumps exceeds 100 hours, the main and standby pumps will be forcibly switched. Each time the pressure stabilizing pump starts, if the pump with the shortest cumulative running time is in a fault state, the pump with the second shortest running time will be selected in sequence.

7. The dynamic pressure stabilization control method for fire-fighting water supply networks based on self-identification of operating conditions according to claim 1, characterized in that, The predictive control module includes a parameter self-tuning unit, which can be turned on or off by the operator through a human-machine interface; When the parameter self-tuning unit is activated, it periodically or under specific operating conditions records and analyzes historical pressure fluctuation data of the pipeline network under conditions of no water usage and leakage, and dynamically updates the preset steady-state variance threshold δ based on this historical data. 2 The threshold values ​​for minute leakage slope k3 and fire water slope k4; The specific update process is as follows: Triggered periodically every 7 days, or triggered when pipeline maintenance is completed, ambient temperature changes exceed ±8℃, or pipeline pressure loss changes exceed 15%, collecting pressure time-series data from the past 24 hours under conditions of no water usage and no leakage in the pipeline network, recorded as follows: p j , j =1,2,..., M , M For the number of data points; Calculate the mean of the pressure data during this period. , And update the steady-state variance threshold. , In the formula, β is the fluctuation margin coefficient, which ranges from 1.1 to 1.5 and is pre-calibrated according to the measurement error characteristics of the pressure transmitter; Linear fitting was performed on the pressure data during this period to obtain the natural pressure decay slope k5, in MPa / min. The micro-leakage slope threshold k3 was updated to 3·|k5|, and the fire water slope threshold k4 was updated to 10·|k5|. After each threshold update, the parameter self-tuning unit stores the updated threshold in a local cache and compares it with the previous three historical thresholds. If the updated threshold exceeds ±30% of the average of the historical thresholds, the update is deemed invalid, the original threshold is maintained, and an anomaly log is recorded. If the update is valid, the updated threshold must be within a preset reasonable range: steady-state variance threshold δ. 2 The value is 0.001~0.01MPa. 2 The threshold for the slope of minute leakage, k3, is -0.01 to -0.001 MPa / min, and the threshold for the slope of fire-fighting water supply, k4, is ≤ -0.1 MPa / min. When the parameter self-tuning unit is turned off, the preset steady-state variance threshold δ 2 The threshold values ​​for minute leakage slope k3 and fire water slope k4 are set to factory default values ​​or manually calibrated by staff through the human-machine interface.

8. The dynamic pressure stabilization control method for fire water supply networks based on self-identification of operating conditions according to claim 7, characterized in that, When the ambient temperature of the pipeline network changes by less than ±3℃ within 24 hours and the monthly change in pressure loss is less than 5%, the parameter self-tuning unit remains in the off state, and the preset steady-state variance threshold δ 2 The threshold values ​​for minute leakage slope k3 and fire water slope k4 are calibrated manually on a regular basis, with a calibration cycle of 3 to 6 months.

9. The dynamic pressure stabilization control method for fire-fighting water supply networks based on self-identification of operating conditions according to claim 6, characterized in that, The field controller monitors the operating status of the pressure transmitter, flow switch, frequency converter, and pressure-stabilizing pump set in real time, and executes corresponding fault handling methods based on the monitoring results: When a pressure transmitter fault is detected, the field controller suspends the use of real-time pressure timing data and instead uses the predicted value based on historical pressure data from the same period, and issues a pressure transmitter fault alarm. When a flow switch fault is detected, the field controller forces the flow switch signal to be in an open state, performs operating condition identification separately based on pressure timing data, and issues a flow switch fault alarm. When a frequency converter fault is detected, the field controller immediately locks the corresponding voltage regulator pump driven by the frequency converter. If there are other available frequency converters and voltage regulator pumps in the voltage regulator pump group, it switches to the backup frequency converter and the corresponding voltage regulator pump. If there are no available backup devices, a frequency converter fault alarm is issued and the current voltage regulator pump group status is maintained. When a fault is detected in a pressure-stabilizing pump in the pressure-stabilizing pump group, the field controller, based on wear equalization, skips the faulty pump in subsequent startups and selects the available pump with the second shortest cumulative running time as the main operating pump, while issuing a pressure-stabilizing pump fault alarm.

10. The dynamic pressure stabilization control method for fire-fighting water supply networks based on self-identification of operating conditions according to claim 5, characterized in that, The field controller has a built-in data storage unit that stores and cyclically overwrites the following data in real time, with a storage period of no less than 90 days: Pressure time-series data and their corresponding timestamps; Flow switch signal and its status change time; The working condition identification results for each time window output by the working condition self-identification model; The predictive control module outputs the inverter's adjustment frequency, the pressure-stabilizing pump's start / stop commands, and the execution time. The cumulative running time of each pump in the pressure-stabilizing pump set and the time of each start-up and shutdown; When a suspected leakage condition is identified, record the duration T1, slope k1, acceleration a, leakage confirmation result, estimated leakage flow rate Q, and corresponding alarm level. When the parameter self-tuning unit is enabled, record the time of each threshold update, the threshold before the update, the threshold after the update, and the basis for the update. The field controller detects all equipment fault types, fault times, recovery times, and fault handling actions performed. The field controller supports querying the above historical data through the human-machine interface or remote monitoring platform, and exporting it in the form of trend curves or reports. The exported data includes the complete process of operating condition changes and control response records.

Images