Deep well hot water submersible pump axial force working condition adaptive balancing method

Abstract

The application discloses a deep well hot water submersible pump axial force working condition adaptive balance method. The method obtains a pump characteristic curve, impeller stage group parameters and a water temperature compensation coefficient table, calibrates a characteristic curve function, collects pump outlet pressure, flow, motor input power, rotating speed and water temperature operation parameters and pre-processes, estimates total axial force based on the characteristic curve according to an impeller stage group axial force model, corrects the total axial force after verification by the motor input power and cross-checking by the outlet pressure, corrects balance cavity balance force according to the water temperature compensation coefficient, determines residual load of a thrust bearing according to the corrected total axial force and the corrected balance force, and adjusts balance cavity throttle valve opening degree to minimize the residual load. The method can estimate axial force on line without downhole sensors and adaptively adjusts balance according to working conditions.

Description

Technical Field

[0001] This application relates to the field of deep well hot water submersible pump technology, and in particular to an adaptive balancing method for axial force conditions of deep well hot water submersible pumps. Background Technology

[0002] Deep-well hot water submersible pumps operate in high-temperature and high-pressure environments such as geothermal wells. Due to limited downhole space and high ambient temperatures, directly installing force or pressure sensors downhole to measure axial force presents challenges such as poor reliability, high cost, and difficult wiring, making real-time monitoring of downhole axial force difficult over long periods. Furthermore, the pump's flow rate and head vary with well conditions, causing significant fluctuations in axial force. Traditional passive balancing structures with fixed balance discs or drums cannot adjust their balancing capacity according to operating conditions, leading to increased residual load on the thrust bearing and a higher risk of overload under varying conditions. In addition, the balancing clearance expands thermally with rising water temperature under hot water conditions, causing deviations in the calculation of the balancing force in the balancing chamber, which cannot be corrected by fixed balancing devices, further exacerbating the load on the thrust bearing. Therefore, how to estimate downhole axial force online based on easily measurable signals from the surface and adaptively adjust the balancing according to operating conditions without the constraints of downhole sensors has become a pressing problem in this field. Summary of the Invention

[0003] To address the shortcomings of existing technologies in deep well hot water submersible pumps, such as the difficulty in online sensing of axial force under conditions without downhole sensors, the inability of passive balancing structures to adaptively adjust to changing operating conditions, and thermal drift due to hot water conditions leading to balancing force deviations, this application provides an adaptive axial force balancing method for deep well hot water submersible pumps. This method uses an impeller stage axial force model based on the pump characteristic curve to estimate the total downhole axial force online from multiple operating parameters at the surface. The estimation results are verified using motor input power and cross-checked with outlet pressure. A water temperature compensation coefficient is used to correct for thermal drift of the balancing force. Finally, the opening of the throttle valve in the balancing chamber is adjusted according to the residual load to minimize the residual load on the thrust bearing. Specifically, this application provides the following technical solutions: Firstly, this application provides an adaptive balancing method for axial force conditions of a deep well hot water submersible pump, including: Obtain the pump characteristic curve, impeller stage parameters, and water temperature compensation coefficient table, and calibrate the characteristic curve function; Collect pump outlet pressure, flow rate, motor input power, speed and water temperature operating parameters, and preprocess the operating parameters; Based on the aforementioned characteristic curves, the total axial force is estimated according to the impeller stage axial force model. The axial force of each stage is a function of the pressure difference between the impeller inlet and outlet and the projected area of ​​the hub, and is accumulated stage by stage. The total axial force is verified using the motor input power and cross-checked using the outlet pressure. The total axial force is then corrected based on the verification results. Determine the balancing force of the balancing chamber throttle valve at the current opening, and correct the balancing force according to the water temperature compensation coefficient; The residual load of the thrust bearing is determined based on the corrected total axial force and the corrected balancing force. The opening of the throttle valve in the balancing chamber is adjusted to minimize the residual load.

[0004] Optionally, the preprocessing includes: performing sliding window filtering on each operating parameter to remove outliers that exceed the limit and retaining the most recent valid value; when the number of continuously removed outliers exceeds the window length, setting a signal quality degradation flag and reducing the verification weight.

[0005] Optionally, estimating the total axial force includes: converting the characteristic curve to the current rotational speed according to the similarity law, and determining the total head by fitting a function.

[0006] Optionally, the estimation of total axial force further includes: distributing the total head to the inlet and outlet pressure differences of each stage impeller according to the stage group distribution coefficient, and ensuring that the inlet and outlet pressure differences of each stage satisfy the energy conservation constraint.

[0007] Optionally, the estimation of the total axial force further includes: determining the axial force of each stage based on the inlet and outlet pressure difference and the hub projection area of ​​each stage impeller, and accumulating the axial forces of each stage to the total axial force.

[0008] Optionally, the step of verifying the total axial force using the motor input power includes: determining the total head based on the flow rate, the characteristic curve, and the current rotational speed; determining the hydraulic power based on the flow rate and the total head; converting the hydraulic power into an estimated value of the motor input power using an efficiency chain; verifying the total axial force according to the relative deviation between the estimated value of the motor input power and the actual motor input power, and correcting it according to the deviation limit; and reducing the verification weight when the outlet pressure exceeds the cross-verification limit.

[0009] Optionally, the step of correcting the balancing force according to the water temperature compensation coefficient includes: determining the water temperature compensation coefficient according to the measured water temperature using an interpolation table or a reference formula; when the measured water temperature exceeds the applicable range of the interpolation table or the reference formula, taking the endpoint value as the water temperature compensation coefficient and outputting an out-of-bounds alarm.

[0010] Optionally, adjusting the opening of the balance chamber throttle valve includes: determining the signed balance error based on the corrected total axial force and the corrected balance force; determining the opening increment according to the signed error proportional-integral control law; applying opening limit and rate of change limit to the opening increment, and freezing the integral accumulation during limit; and determining convergence and maintaining the current opening when the absolute value of the signed balance error is not greater than the convergence threshold.

[0011] Optionally, the method further includes: outputting a warning signal or performing speed reduction when the corrected total axial force is not less than a warning threshold or the residual load is not less than a residual load warning threshold; performing a linkage shutdown when the corrected total axial force is not less than a protection threshold or the residual load is not less than a residual load protection threshold; and performing warning cancellation according to the hysteresis interval, and managing the transfer between the warning state and the protection state through a hierarchical protection state machine.

[0012] Secondly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0013] Beneficial effects: This application migrates the axial force measurement link from downhole to the surface, using the pump characteristic curve as the core mechanism and the impeller stage axial force model as the calculation carrier. It uses easily measurable signals from the surface to estimate the total downhole axial force online, so that axial force sensing in deep well high-temperature and high-pressure scenarios no longer depends on downhole sensors. The estimated result is verified by the measured value of motor input power and cross-checked by the outlet pressure, forming a second information channel to suppress estimation deviations caused by characteristic curve aging or stage parameter drift. The balance force deviation caused by thermal expansion of the balance gap is corrected by the water temperature compensation coefficient, avoiding the underestimation of residual load. The residual load is determined based on the corrected total axial force and the corrected balance force, and the opening of the balance chamber throttle valve is adjusted to minimize the residual load. Closed-loop active balancing replaces passive fixed balancing, so that the balancing capability adapts to the working conditions and reduces the risk of thrust bearing overload under varying working conditions. Attached Figure Description

[0014] Figure 1 A schematic diagram of the overall process of the adaptive balancing method for axial force conditions of a deep well hot water submersible pump provided in the embodiments of this application.

[0015] Figure 2 This is a schematic diagram of the sub-process for collecting and preprocessing operating parameters provided in an embodiment of this application.

[0016] Figure 3 This is a schematic diagram of the soft measurement estimation subprocess for the axial force of the impeller stage provided in an embodiment of this application.

[0017] Figure 4 This is a schematic diagram of the motor power verification and model correction sub-process provided in the embodiments of this application.

[0018] Figure 5 This is a schematic diagram of the adaptive regulating sub-process of the balance chamber throttle valve provided in an embodiment of this application.

[0019] Figure 6 This is a schematic diagram of the system module architecture provided for an embodiment of this application.

[0020] Figure 7This is a schematic diagram of the family of pump characteristic curves and similarity conversion provided in the embodiments of this application.

[0021] Figure 8 This is a schematic diagram of the water temperature compensation coefficient curve provided in the embodiments of this application.

[0022] Figure 9 This is a schematic diagram of the residual load convergence curve provided in an embodiment of this application.

[0023] Figure 10 A schematic diagram of the connection architecture of the deep well hot water submersible pump, ground sensor group, controller and throttle valve device provided in the embodiments of this application.

[0024] Attached reference numerals: 1. Deep well hot water submersible pump; 2. Balance chamber; 3. Throttling valve; 4. Thrust bearing; 5. Surface sensor group; 6. Parameter library module; 7. Signal acquisition module; 8. Soft measurement engine; 9. Calibration module; 10. Compensation module; 11. Adjustment and execution module; 12. Protection module; 13. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0027] The adaptive axial force balancing method for deep well hot water submersible pumps is applicable to deep well hot water submersible pumps, geothermal well circulating pumps, mining hot water pumps, and various multi-stage submersible pumps with balancing chambers and throttle valve structures. For example... Figure 10 As shown, the system deploys a controller 5 and a ground sensor group 6 at the surface end, and a balance chamber 2, a throttle valve 3, and a thrust bearing 4 downhole. The controller 5 is connected to the downhole actuator and the ground sensor group 6 via cables. The downhole environment of deep well hot water submersible pumps is high temperature and high pressure, with water temperatures reaching 60℃ to 120℃. Downhole space is limited, and directly installing force sensors results in poor reliability and difficult maintenance. Therefore, this method uses easily measurable signals from the surface end as input, and estimates the downhole axial force online using a physical mechanism model. For example... Figure 1 As shown, this method includes the following steps: S000, obtain the pump characteristic curve, impeller stage parameters and water temperature compensation coefficient table, and calibrate the characteristic curve function.

[0028] The pump nameplate parameters, factory characteristic curve discrete points, water temperature compensation coefficient table, and control threshold table stored in parameter library module 7 are read into controller 5. The pump nameplate parameters include the number of impeller stages Z and the projected area of ​​each stage hub. Rated speed Rated flow rate Rated head With the effective working area of ​​the balance cavity During loading, verify the relationship between the impeller stage number Z and the hub projected area. To ensure accuracy and consistency with the pump nameplate, the temperature range of the calibration compensation coefficient table should cover 60℃ to 120℃. Parameter loading is performed sequentially and verified item by item: first, load the pump nameplate parameters and verify their completeness; then load the discrete points of the factory characteristic curve and verify that the number of points is no less than 5 and the head decreases monotonically with flow rate; next, load the water temperature compensation coefficient table and verify that the temperature nodes increase monotonically and the coefficient values ​​fall within the range of 0.9 to 1.1; finally, load the control threshold table and verify that the warning threshold is less than the protection threshold. If any verification fails, loading is terminated and the corresponding failure flag is set, and the calibration process is not initiated. After loading, a version number and timestamp are added to the calibration results. The version number increments with parameter changes, and the timestamp is used for event logs and maintenance traceability. If a parameter version change is detected in any subsequent cycle, the soft measurement and adjustment steps are reinitialized with the new parameters.

[0029] Take the factory characteristic curve at the rated speed Discrete points below ( , ), j is taken from 1 to M, and converted to the current actual speed n according to the similarity law, such as Figure 7 As shown, the family of characteristic curves, after conversion, covers the operating speed range: , A quadratic polynomial fit is performed on the converted discrete points to obtain the characteristic curve function. The coefficients a, b, and c are determined by least squares fitting, and the root mean square of the fitting residuals is less than 2% of the rated head. Least squares fitting aims to minimize the sum of squares of the head deviations at each discrete point; that is, it seeks solutions that minimize the sum of squares of deviations for coefficients a, b, and c. The objective function is quadratic with respect to the coefficients, and its extreme points satisfy a linear canonical equation, which can be solved directly without iterative search. Before fitting, the factory discrete points are converted to the current speed according to the similarity law. The physical basis of the similarity law is that the impeller outlet circumferential speed is proportional to the speed, and the head is proportional to the square of the circumferential speed. Therefore, the converted discrete points correspond one-to-one with the original factory points. Fitting is performed in the converted coordinate space, making the fitting function directly applicable to the current speed. The root mean square of the fitting residuals is the square root of the mean of the sum of squares of the residuals at each point, i.e., the root mean square error. ,in The residuals for each discrete point are set to a threshold of 2% of the rated head, based on the ability of the quadratic polynomial to fit the smooth characteristic curve and the measurement uncertainty of the factory test points. When the residuals exceed the limit, the factory curve is judged to be abnormal or the fitting order is insufficient. In this case, a cubic polynomial can be used or the number of discrete points can be increased before refitting.

[0030] For example, the rated speed of an 8-stage deep well hot water submersible pump 2900 rpm, rated flow rate for Rated head The head is 120m. The factory curve uses five discrete points: (0,140), (25,132), (50,120), (75,102), and (100,80). Least square fitting yields a = -0.00394, b = -0.206, and c = 139.9, with a root mean square residual of approximately 0.22m, which is less than 2% of the rated head (2.4m). The calibration is successful. When the actual speed is 2800rpm, the zero-flow head, converted to the square of the speed ratio, is 130.4m, which can be used in subsequent steps to calculate the total head from the flow rate at any operating speed. The calibrated characteristic curve function covers both rated speed and variable speed operation scenarios. During operation, it is converted in real time according to the current speed, eliminating the need for separate calibration for each speed, reducing calibration workload, and ensuring the consistency of estimation across the entire speed range.

[0031] In an alternative implementation, piecewise linear interpolation can be used directly to establish characteristic curves for the factory discrete points without explicit fitting coefficients. If the pump is equipped with a frequency converter and the speed is accurately known, it can be converted according to the similarity law. The implementation of piecewise linear interpolation is as follows: the converted discrete points are arranged in ascending order of flow rate, and adjacent points are connected by a straight line. During operation, the interval is located according to the current flow rate, and the total head is calculated by linear interpolation. Its advantage is that it completely retains the factory measurement point information and has no fitting residuals. Its disadvantage is that the slope is discontinuous at the endpoints of the interval, making it sensitive to flow fluctuations. Polynomial fitting, on the other hand, is smooth overall and is more suitable as a continuous input for the control law. The results of piecewise linear interpolation and polynomial fitting are consistent at the factory measurement points. Both methods can be used to realize the characteristic curve function, and the choice is made by the calibration steps based on the quality of the factory data.

[0032] After performing a monotonicity check on the water temperature compensation coefficient table, a piecewise linear interpolation function is established. (T). The compensation coefficient table provides compensation coefficients of 1.00, 0.97, 0.94, and 0.91 at four points: 60℃, 80℃, 100℃, and 120℃, respectively, for the reference water temperature. The calibration temperature is set to 60℃, consistent with the first point of the calibration table. The bench calibration process for the compensation coefficient table is as follows: On the test bench, keep the pump flow rate and throttle valve opening constant, and adjust the hot water temperature in the pump to four calibration points in sequence: 60℃, 80℃, 100℃, and 120℃. After the temperature stabilizes at each point, measure the pressure in the balance chamber, and calculate the ratio of the balance force at that temperature to the reference balance force at 60℃ as the compensation coefficient. Repeat the measurement at each calibration point and take the average value. The curves between the calibration points are connected by piecewise linear interpolation. The bench calibration is completed at the factory stage, and the calibration data is written into the parameter database and delivered with the pump. During the field operation stage, only the calibration results are loaded, and no recalibration is performed.

[0033] The equivalent coefficient of thermal expansion β is determined by bench calibration, with a value ranging from 1e-4 to 2e-3 degrees Celsius. In this embodiment, 0.0015 degrees Celsius is used, and four-point verification is performed. Consistent point-by-point with the interpolation table, for example The calibration method for β involves measuring the pressure change in the equilibrium chamber at different water temperatures under fixed flow rates and opening degrees, and then extrapolating the result. Its value is based on the linear expansion coefficient of stainless steel (approximately 1.7e-5 degrees Celsius) and the relative change in the equilibrium gap (on the order of 1e-4 to 1e-3 degrees Celsius). Compensation coefficient. The values ​​range from 0.9 to 1.1. The interpolation function provides coefficient mapping for the water temperature compensation step, while using an explicit β formula for cross-validation. The boundary conditions for the interpolation function are as follows: when the measured water temperature is 60℃ lower than the lower limit of the calibration table, the initial coefficient is set to 1.00 and a low-temperature boundary crossing flag is set; when the measured water temperature is 120℃ higher than the upper limit of the calibration table, the final coefficient is set to 0.91 and a high-temperature boundary crossing flag is set; during the boundary crossing period, no interpolation is pushed outward to prevent the compensation coefficient from deviating from the physically reasonable range of 0.9 to 1.1. The interpolation function and the β formula are consistent point by point within the calibration range. During the operation phase, the interpolation function is used first, and the β formula is used as a cross-validation method after loading. When the deviation between the two exceeds 0.01, a compensation table consistency anomaly flag is set, which is recorded by the protection step.

[0034] After calibration, the model self-test baseline value is calculated and the model self-test flag is output. Before calibration is complete, subsequent steps are in a waiting state. The model self-test baseline value includes three baseline points: zero flow head, rated flow head, and full opening balance force. After loading, the values ​​are recalculated according to the current parameters and compared with the factory baseline. If the relative deviation is no greater than 2%, the self-test is considered passed. If the deviation of any baseline point exceeds the limit, a self-test failure flag is set and a diagnostic code is output. The soft measurement engine 9 does not start, the protection module 13 records the initialization fault event, and the system refuses to enter the estimation process. For example, if the impeller stage number Z is found to be 0 or the hub projection area is found to be 0 during loading, the self-test failure flag is set and a diagnostic code is output. The soft measurement engine 9 does not start, the protection module 13 records the initialization fault event, and the system refuses to enter the estimation process. If the value is not greater than 0, the parameter loader sets the parameter loading failure flag, the soft measurement engine 9 does not start, the protection module 13 records the initialization fault event, and the system refuses to enter the estimation process. Output status includes the calibrated characteristic curve function, the group parameter set {Z = 8, ...} for The system includes a compensation interpolation function, a control threshold table (early warning threshold 28000N, protection threshold 32000N), and model self-check flags. Output status is written to a shared memory calibration data slot for the soft measurement engine 9, compensation module 11, and protection module 13 to read during their respective execution cycles. The calibration data slot remains unchanged until the parameter version is updated; each consumer module determines whether reloading is needed based on the version number.

[0035] S100 collects pump outlet pressure, flow rate, motor input power, speed and water temperature operating parameters, and preprocesses the operating parameters.

[0036] Signal acquisition module 8 uses a fixed sampling period The raw signals output by ground sensor group 6 are sampled synchronously. The raw signals include the outlet pressure. Flow rate Q, motor input power The rotational speed n and water temperature T are also considered. The configuration of ground sensor group 6 is as follows: an outlet pressure transmitter is installed in the pump outlet pipe section, with a range of 0 to 5 MPa, outputting a 4 to 20 mA analog signal; the flow meter is either electromagnetic or ultrasonic, with a range of 0 to... The motor input power is measured by a three-phase power transmitter with a range of 0 to 45 kW; the speed is provided by the inverter communication interface or encoder, with a range of 2700 to 3000 rpm; the water temperature sensor is installed at the pump outlet or wellhead return water pipe section, with a range of 0 to 150℃. Each signal is converted from an A / D converter and then enters a circular buffer with a window length L of 5. A sliding window filter is performed on each signal, and the average of the five most recent samples within the window is used as the filtered output to suppress high-frequency noise from the sensor and preserve the trend of operating conditions. Figure 2 As shown, the filtered signal continues to undergo amplitude limiting detection, comparing the current sampled value with the previous valid sampled value. If the difference exceeds the set limit, it is judged as an outlier and removed, and the most recent valid value is retained, while the lost-point count is accumulated. If the sampled value exceeds the sensor's range, it is judged as invalid. After aligning all signals according to the sampling time, a pre-processed synchronization signal is output. , , , , The data is valid. The sampling alignment of each signal is as follows: based on the sampling period. Based on the start time, each sampling channel simultaneously triggers sample-and-hold. After A / D conversion, the converted values ​​of each channel, along with the same timestamp, are written to a circular buffer. The timestamp monotonically increases according to the sampling period. If a channel is occupied or times out during conversion, the data for that channel in this period is marked as missing, filled with the most recent valid value, and a missing point of 1 is recorded. Simultaneously, the timestamp for that channel is maintained at the value of the previous period to prevent time misalignment from entering subsequent calculations. Sampling Period A sampling frequency of 100ms is chosen, corresponding to a sampling frequency of 10Hz. This provides a margin compared to the bandwidth of varying operating conditions, balancing real-time estimation with filtering effectiveness.

[0037] The mean form of sliding window filtering is: A window length L of 5 is a common compromise in engineering. An excessively long window introduces phase lag and affects the closed-loop response of the throttle valve, while a window that is too short results in insufficient filtering. A length of 5 ensures that the filter cutoff frequency is approximately one-fifth of the sampling frequency. In alternative implementations, median filtering can be used, taking the median value within the window as the filter output, which is more robust to impulse noise; alternatively, a first-order low-pass filter can be used, i.e., the current filter value... α is set to 0.2. The applicable boundaries of the three filtering methods are as follows: sliding window mean filtering has the best suppression of Gaussian white noise and is suitable for scenarios where steady-state noise dominates; median filtering is robust to impulse interference and is suitable for scenarios where there are electromagnetic interference spikes or signal interruptions; first-order low-pass filtering has the least computational cost and is suitable for scenarios where controller resources are limited. In actual deployment, the selection is based on the noise type: during the calibration phase, signal samples are collected in both shutdown and operation states, and the noise standard deviation and impulse amplitude distribution are statistically analyzed. When the noise is mainly Gaussian, sliding window mean filtering is selected; when the impulse component is dominant, median filtering is selected; and when computational resources are limited, first-order low-pass filtering is selected. The selection results and window parameters are written into the parameter library for loading during the operation phase. If the measured noise is large, the window length can be adjusted during the calibration phase according to the noise standard deviation collected in the shutdown state.

[0038] The criteria for amplitude limiting detection are ,in Calibrate according to the sensor's range and the rate of change of operating conditions, such as flow rate limiting. A 5% threshold of the rated flow rate is used per sampling cycle. The flow rate change rate under normal operating conditions is much lower than this value, while the flow rate jump during a disconnection is much higher, making them clearly distinguishable numerically. The limiting parameters for each signal are independently calibrated according to their respective range and operating characteristics: outlet pressure limiting is 3% of the range per sampling cycle, motor input power limiting is 8% of the rated power per sampling cycle, speed limiting is 2% of the rated speed per sampling cycle, and water temperature limiting is 2% of the range per sampling cycle. These limiting parameters are loaded during the calibration process, based on noise statistics collected under shutdown conditions and the maximum rate of change under normal operating conditions, ensuring that normal fluctuations are not misjudged and abnormal jumps are not missed. When the number of consecutive lost points exceeds the window length, a signal quality degradation flag is set, notifying the verification process to reduce the verification weight. The rules for handling dropped signal counts and signal quality degradation are as follows: When the cumulative number of dropped signals for a single signal is within the window length L, only the most recent valid value is retained, and downstream processing remains unchanged; when the cumulative number of dropped signals exceeds L, the signal is determined to be degraded, a degradation flag is set, and the verification process reduces the corresponding verification weight of the signal. For example, when the motor input power signal degrades, the efficiency chain verification weight is reduced to 50% of the normal value; when the outlet pressure signal degrades, the cross-verification weight is reduced to 50% of the normal value. The estimation process continues to calculate using the remaining normal signals to ensure that the system can operate at a degraded level under partial sensor failures. The degradation flag is cleared after the signal returns to normal and there are no dropped signals for several consecutive cycles, and the verification weight is restored synchronously.

[0039] For example, sampling period Take 100ms, and collect the raw value at the current time { 1.85 MPa, Q is , The power output is 21.6kW, the rpm is 2801rpm, and the temperature (T) is 95.3℃. The five closest points Q within the flow filtering window are [48.5, 48.0, 48.2, 47.9, 48.2]. mean for The flow rate in the previous cycle during the amplitude limiting detection was... , for The difference between the current value and the previous value for If the flow rate is less than the limit, the pass is granted. Counterexample: If the flow rate at a certain point jumps to... The difference from the previous value for greater than It was judged as abnormal and removed. Maintain and account for point 1. Convert outlet pressure to head. ρ is the density of hot water at 95℃. , This value includes static head and pipeline losses, and is used for cross-checking the outlet pressure in the power calibration step. The total head used in the soft measurement step is obtained from the characteristic curve. The difference between the outlet pressure-converted head and the characteristic curve head comes from two items: static head and pipeline losses. Static head is the liquid column pressure corresponding to the pump installation depth and is independent of flow rate; pipeline losses increase approximately with the square of the flow rate and are calibrated by pipeline characteristics. Cross-checking compares the converted head with the characteristic curve head plus the two corrections, reflecting only the deviation of the characteristic curve itself after eliminating installation and pipeline factors. The output status is passed to subsequent steps, including { 1.85MPa for , 2801rpm For 21.6kW, The value is 95.3℃, valid is TRUE, and missing points are 0. The signal validity flag is a summary of the valid states of all signals. When the quality of any signal is determined to be degraded, valid is set to partially valid; when all signals are normal, valid is set to fully valid. The soft measurement engine 9 and the verification module 10 determine whether to use the data of this cycle in the calculation according to the level of valid. When partially valid, the normal signal is used for calculation and the corresponding verification weight is reduced. When completely invalid, the calculation of this cycle is skipped and the output of the previous cycle is maintained.

[0040] S200, based on the characteristic curve, the total axial force is estimated according to the impeller stage axial force model. The axial force of each stage is a function of the pressure difference between the impeller inlet and outlet and the hub projected area, and is accumulated stage by stage.

[0041] Soft Measurement Engine 9 reads preprocessed traffic from shared memory With rotational speed The calibrated characteristic curve function is called, and the characteristic curve is converted to the current speed according to the similarity law. The total head H under the current operating condition is determined by the fitting function. The total head is distributed to the inlet and outlet pressure difference of each impeller according to the stage group allocation rule. The axial force of each stage is determined according to the inlet and outlet pressure difference of each stage and the hub projection area. The axial forces of each stage are accumulated step by step to form the total axial force. .like Figure 3 As shown, the estimation process sequentially involves calculating the total head, pressure differential distribution, single-stage axial force calculation, and step-by-step accumulation, followed by a physical rationality check. The computational workload of the estimation process is concentrated in the single-stage axial force calculation. When the number of stages Z is 8, the calculation volume per cycle is approximately tens of floating-point operations, which is far less than the controller's cycle budget and can be completed in real time. The estimation process adopts a pipeline structure, with the calculation of total head, pressure differential distribution, and single-stage calculation executed sequentially. Each stage of the calculation is independent and can be parallelized and accelerated stage by stage.

[0042] The total head is calculated using the quadratic polynomial obtained from calibration. Where a, b, and c are calibration coefficients. In the stage group pressure difference distribution, the inlet and outlet pressure difference of the i-th stage... Single-stage head Coefficients are allocated according to grade group. distribute, The value range is from 0.9 times (1 / Z) to 1.1 times (1 / Z). In this embodiment, equal allocation is used. and satisfy The energy conservation constraint. The implementation method for unequal allocation is as follows: Due to differences in import and export structures, the allocation coefficients of the first and last stages are slightly adjusted based on the benchmark values. The first stage is set to 1.1 times the benchmark, and the last stage is set to 0.9 times the benchmark. The intermediate stages maintain equal allocation, and the adjustment still satisfies the following requirements. Energy conservation constraints; if the pump's factory output curve provides the actual distribution ratios for each stage, the factory data can be directly used for loading. Factory data has higher priority than the allocation benchmark. Another alternative implementation is to allocate according to the geometric dimensions of each impeller stage, meaning the pressure difference between the inlet and outlet of each stage is proportional to the square of the impeller's outer diameter. Determined by the square ratio of the outer diameter of each impeller stage, it is applicable to multi-stage pumps with inconsistent impeller sizes at each stage; geometric size allocation and equal allocation degenerate into the same result when the impeller sizes at each stage are consistent.

[0043] ρ is based on the actual measured water temperature According to the hot water property table, in this embodiment, the temperature was taken at 95.3℃. Axial force at each level , This is the impeller axial force correction factor, with a value ranging from 0.8 to 1.2; in this embodiment, it is set to 1.0. A warning sign is displayed when the pressure exceeds this range. The physical mechanism of this model is as follows: the areas affected by fluid pressure on the impeller inlet and outlet sides are different; the hub projection area corresponding to the low-pressure zone on the inlet side is smaller than that on the high-pressure zone on the outlet side; the pressure difference between the front and rear cover plates cannot be completely canceled out axially; the net axial force is approximately equal to... ; It is used to absorb the equivalent deviation caused by hub taper, interstage leakage, and differences in flow channel geometry. Its empirical range of 0.8 to 1.2 is determined by engineering statistics in the field of impeller axial force calculation. The closer the value is to 1, the closer the axial force of that stage impeller is to the ideal product of pressure difference and area. The physical meaning of progressive accumulation is that the axial forces generated by each impeller in a multi-stage pump are superimposed in the same axial direction. The axial force of each impeller acts on the pump shaft and is transmitted to the thrust bearing; therefore, the total axial force... For example, if impeller wear in a certain stage increases clearance and inter-stage leakage, the pressure difference between the inlet and outlet of that stage decreases, and the axial force of that single stage decreases accordingly. The remaining stages are unaffected, and the total axial force decreases cumulatively stage by stage. The soft-sensor results are continuously traceable as the operating conditions change. Total Axial Force The estimated result must fall within the physical reasonable range of 0 to the protection threshold; if it exceeds this range, the estimation is invalidated. The criteria for physical reasonableness checking include: total head H greater than 0, and pressure difference between inlet and outlet at each stage. Greater than 0, total axial force The value should not exceed the protection threshold, and the ratio of the cumulative result at each level to the average value of a single level should not exceed the reasonable range corresponding to level Z. If any one of these conditions is not met, the estimation invalidation flag should be set and the protection step should be initiated to record the soft measurement abnormality event. The throttle valve adjustment step should maintain the opening of the previous cycle and should not execute any new adjustment action to prevent misadjustment based on invalid estimation values.

[0044] For example, continuing the aforementioned interface state, for , The rpm is 2801, and Z is 8. for ρ take The calibration coefficients are a = -0.00394, b = -0.206, and c = 139.9. Speed ​​ratio. Total head Step-by-step calculation yields , and ,total Single-stage head Single-stage inlet and outlet pressure difference In the energy conservation check, the sum of the pressure differences at each level , The two are consistent within the rounding error range, and the verification passes. Single-stage axial force. Total axial force If the value falls within the physically plausible domain, the reliability flag is set to TRUE. Counterexample: If the speed sensor malfunctions... If the calculated total head is 0, the physical rationality check for a value greater than 0 is not met, triggering the estimation invalidity flag. The process then proceeds to the protection step, recording the abnormal soft-sensor input event, and does not enter the throttle valve adjustment step. The output status is passed to subsequent steps, including when H is 111.8m. The axial force vector is 5272N, with each step having a specific value. The system uses a reliability flag for estimation. This flag is read by both the verification module 10 and the protection module 13. If the estimation is reliable, the verification module performs normal power verification; if it is unreliable, the verification module skips the current cycle verification and retains the previous cycle's correction result. The protection module 13 processes the unreliable estimate using a conservative strategy. The soft measurement engine 9 maintains a cache of the most recent estimation result, employing a double-buffer structure: the previous cycle's H, the step-by-step vector, and... The buffer is maintained until the calculation for the current cycle is completed before being overwritten, to avoid the throttle valve adjustment step reading partially updated data; the switching of the double buffer is performed uniformly at the end of the cycle, and all quantities are updated synchronously at the moment of switching to ensure that the estimation results read by downstream modules are consistent. The cache also records the timestamp and speed value of the estimation moment, so that the verification step can compare power and axial force by time alignment.

[0045] S300, the total axial force is verified using the motor input power and cross-checked using the outlet pressure, and the total axial force is corrected according to the verification result.

[0046] Verification module 10 reads traffic Total head H, outlet pressure With motor input power .Depend on Calculate hydraulic power with H , Convert to cubic meters per second; then from The estimated motor input power is calculated using a four-term efficiency chain. The four sub-efficiency chains include hydraulic efficiency. Mechanical efficiency Volumetric efficiency With motor efficiency , .Will Compared with actual measurement Compare and calculate the relative deviation. The estimate is considered reliable when δ is not greater than the tolerance ε; when ε is greater than the tolerance ε, the estimate is considered reliable. Within the interval, the total head is corrected according to the direction of deviation and the total axial force is recalculated; δ is greater than The model is then assessed for degradation, and a degradation flag is set. In the cross-checking of outlet pressure, [the following is used]: Converting head , with total mileage Jingyangcheng and pipeline losses The sum (i.e. When the relative deviation exceeds the limit, a pressure inconsistency flag is set and the verification weight is reduced. The physical meaning of the outlet pressure cross-verification is as follows: the outlet pressure transmitter directly measures the actual pressure at the pump outlet, and its corresponding head includes the static head and pipeline losses. It is independent of the head estimated by the characteristic curve. If the two are consistent, it means that the characteristic curve is still accurate under the current operating conditions. When the pressure verification exceeds the limit but the power verification is normal, the characteristic curve aging or stage parameter drift is suspected first, and the verification weight is reduced proportionally. When both the pressure verification and power verification exceed the limit, the estimation model is judged to be distorted as a whole, and a degradation flag is set directly to avoid misjudgment from a single channel.

[0047] like Figure 4As shown, the verification process proceeds sequentially as follows: hydraulic power calculation, efficiency chain conversion, deviation comparison, graded correction, and pressure cross-verification. The physical motivation for power verification lies in energy conservation: the pump converts the electrical energy input from the motor into the hydraulic power of the fluid through four efficiency chains: hydraulic, mechanical, volumetric, and motor. This hydraulic power is determined by flow rate, head, and density. If the characteristic curve, due to wear and aging or stage parameter drift, causes the estimated head to be too low, the estimated motor input power derived from the estimated head will also be too low, resulting in a systematic positive deviation from the measured power. The direction of this deviation indicates the correction direction, and the deviation amplitude, after efficiency chain conversion, becomes a proportional estimate of the head deviation. Therefore, power verification serves both as a reliability assessment and as a correction measure and direction, ensuring that the model, after correction, matches the measured energy input.

[0048] hydraulic efficiency Mechanical efficiency is calibrated according to the pump's factory efficiency curve. The value ranges from 0.95 to 0.98, indicating volumetric efficiency. The value ranges from 0.95 to 0.97, indicating motor efficiency. The values ​​are based on the motor nameplate and load rate, ranging from 0.93 to 0.96. The physical meanings of the four sub-efficiencies are: hydraulic efficiency represents the proportion of hydraulic losses in the impeller and guide vane flow channels to the total energy; mechanical efficiency represents mechanical friction losses from bearings and packing; volumetric efficiency represents interstage leakage and backflow losses from sealing gaps; and motor efficiency represents the copper losses, iron losses, and stray losses of the motor windings. The product of these four sub-efficiencies is the overall pump unit efficiency. The value range is 0.68 to 0.78. The deviation tolerance ε is 5%, and the correction limit is... 15% is taken, all based on factory test calibration. Pressure calibration tolerance. Take 10%, static head With pipeline loss Loading is determined by the installation site. When the flow rate is far from the optimal efficiency point (below 30% of the rated flow), the extrapolation error of the efficiency curve increases, and the calibration weight is downgraded. The correction action is constrained by amplitude to prevent over-adjustment: correction ratio. The power data is determined to be abnormal if the correction ratio is outside the range of 0.85 to 1.15, and the correction is not accepted. A degradation flag is then set to prevent the power transmitter malfunction from causing the estimation results to be biased.

[0049] When the verification conclusion is degradation, the throttle valve adjustment step pauses adjustment and maintains the current opening, while the protection step records the model degradation event. The verification module performs trend judgment on the relative deviation δ sequence of the most recent N periods, where N is 10. When the deviation continuously increases and the most recent value is greater than the tolerance ε, even if the correction limit has not yet been reached, the verification is initiated. A degradation warning sign is also set to indicate that there is a slow drift in the characteristic curve or stage parameters, allowing the protection steps to intervene in advance. Trend judgment uses a sliding window buffer with a window length of 10 cycles, approximately 1 second, balancing response speed and noise suppression. In an alternative implementation, a shaft power meter can be added to directly measure the shaft power, or the motor input power can be calculated using the power factor and voltage / current, serving as an independent verification channel for motor input power estimation. The shaft power meter directly measures the pump shaft input power, skipping the motor efficiency step, shortening the verification chain to hydraulic power converted by hydraulic, mechanical, and volumetric efficiencies, reducing the deviation introduced by motor efficiency uncertainty. The power factor method calculates the motor input power using voltage, current, and power factor, eliminating the need for a dedicated power transmitter. Its accuracy is affected by power factor measurement errors, making it suitable for applications with lower accuracy requirements. When multiple channels coexist, the verification conclusion is a weighted sum of the conclusions from each channel, with weights allocated according to the measurement uncertainty of each channel, improving verification robustness.

[0050] For example, continuing the aforementioned interface state, for Right now H is 111.8m. It is 1.85 MPa. It is 21.6kW, ρ is taken as , Take 0.82, Take 0.96, Take 0.96, Take 0.95. Hydraulic power. The product of the four components' efficiency . . If the value falls within the 5% to 15% range, it is determined that the correction has been made. The correction percentage is as follows: Total head after correction Total axial force after correction Export pressure is being cross-checked. Jingyangcheng Take 70m, pipeline loss Take 14m, relative deviation The pressure check passes if the deviation is no greater than 10%. Counterexample: If the actual measurement... The value is 19.7 kW, and δ is 0.3%, which is no greater than 5%, so the estimate is considered reliable. H is equal to 111.8m. Equals 5272N, no calibration required. Counterexample: If measured... It is 14kW. If the correction limit is greater than 15%, the model is deemed to have degraded, a degradation flag is set, the throttle valve adjustment step is paused and the current opening is maintained, the protection step records the model degradation event and suggests maintenance, the soft measurement result is marked as unreliable and does not enter the adjustment link.

[0051] The verification conclusions are categorized into three levels: reliable, calibrated, and degraded. The reliable level corresponds to a deviation (δ) of no more than 5%, the calibrated level corresponds to a deviation (δ) falling within the 5% to 15% range, and the degraded level corresponds to a deviation (δ) greater than 15% or a calibration ratio exceeding the limit. These three levels correspond to three subsequent actions: maintain, calibrate, and pause, forming a graded verification closed loop. The recovery logic for the degraded state is as follows: After the degradation flag is set, the verification module continuously monitors the power deviation and pressure verification results. If the deviation falls back to within the tolerance ε for several consecutive cycles and the pressure verification passes, the model is deemed to have automatically recovered, the degradation flag is cleared, and adjustment is resumed. If the deviation continues to exceed the limit, the degraded state is maintained, and maintenance shutdown for inspection is recommended to prevent operation with defects. Output status includes verification conclusions of "calibrated," "corrected," etc. It is 122.9m. The value is 5794N, the reliability level is calibrated, and the pressure verification mark is passed.

[0052] S400, determine the balancing force of the throttle valve in the balancing chamber at the current opening, and correct the balancing force according to the water temperature compensation coefficient.

[0053] Compensation module 11 reads the measured water temperature Reference balancing force corresponding to the current throttle valve opening ,Depend on The water temperature compensation coefficient can be obtained by looking up the interpolation function of the compensation coefficient. , to the reference balancing force and Multiplying yields the corrected balanced forces Reference balancing force The balancing force corresponding to the current throttle valve opening at the reference temperature for the balancing chamber structure is given by the opening pressure model of the previous cycle of the throttle valve adjustment step, with the calibration reference value used for the first cycle. The balancing chamber pressure model describes the relationship between the throttle valve opening and the balancing chamber pressure. This embodiment uses a linear approximation; the balancing chamber pressure is at its maximum when the throttle valve is fully closed. For every 1% increase in opening, the balancing chamber pressure decreases linearly according to the ratio corresponding to the opening force slope m, which is the reference balancing force. ,in To balance the chamber pressure, the opening pressure model was obtained by point-by-point measurement under fixed flow rate and water temperature through bench calibration. The calibration results were written into the parameter library for use by the compensation module and the regulation module, and were consistent with the calibration results of the opening force slope m of the regulation execution module 12. Figure 8 As shown, the compensation coefficient decreases monotonically with increasing water temperature. The value ranges from 0.9 to 1.1.

[0054] The compensation coefficient interpolation function uses piecewise linear interpolation, based on measured water temperature. Falling into the calibration range , ]hour, Piecewise linear interpolation has a constant computational complexity, requiring only one interval location and one linear operation, making it suitable for real-time execution of controller cycle tasks. Interval location uses binary search, with a fixed number of four calibration points and no more than two searches, resulting in negligible computational overhead. The interpolation results are precisely consistent with the calibration values ​​at the calibration nodes. The error between nodes is determined by the linear assumption of the thermal expansion equivalence coefficient. The interpolation error corresponding to a 20℃ interval between calibration points is less than 0.5% as verified by bench testing. The calibration table is provided by bench calibration at four points: 60℃, 80℃, 100℃, and 120℃, with the reference water temperature as the reference. The temperature is set to 60℃, consistent with the first point of the calibration table. The equivalent coefficient of thermal expansion β is set to 0.0015 per degree Celsius. (Base form) The interpolation table and the benchmark formula are consistent point-by-point within the calibration range, and they mutually verify each other. As water temperature increases, the balance gap increases due to thermal expansion, leading to increased leakage flow, decreased pressure in the balance chamber, and reduced balance force. A value less than 1 indicates this relative change. The physical path of this mechanism is as follows: Increased water temperature causes thermal expansion of the metal components of the balance disc and balance drum, increasing the axial dimension of the balance gap, decreasing the gap flow resistance, increasing the leakage flow through the balance gap, and consequently decreasing the pressure in the balance chamber, thus reducing the balancing force acting on the effective area of ​​the balance chamber. The magnitude of the increase in leakage flow depends on the gap size and pressure difference. The relative change in the gap is between 1e-4 and 1e-3 degrees Celsius, corresponding to an equal magnitude change in leakage flow. Therefore, the compensation coefficient monotonically decreases at a rate of approximately 0.03 per 20°C within the 60°C to 120°C range. The thermal expansion effect of the balance gap is more pronounced in the high-temperature range. The cumulative compensation at 120°C relative to 60°C is approximately 9%. Without correction, the residual load will be underestimated by approximately 9%, a magnitude sufficient to affect the safety margin of the thrust bearing. Therefore, water temperature compensation is a necessary step in hot water operation. The correction amount should not exceed a certain percentage of the reference balancing force. If the correction amount exceeds the limit, a compensation anomaly flag is set. The correction limit percentage is taken as 10% of the reference balancing force, i.e. The value range of 0.9 to 1.1 corresponds to the upper and lower limits of the correction amount. When the measured water temperature exceeds the limit, the endpoint value of 0.91 is taken, and the correction amount is 9%, which is still within the limit ratio. The compensation continues to operate according to the endpoint coefficient. If the coefficient of the loaded compensation table exceeds the range of 0.9 to 1.1 or the interpolation calculation jumps, the correction amount exceeds the limit. The compensation abnormality flag is set and recorded by the protection step. The adjustment step operates at a downgraded level according to the uncompensated reference balance force to avoid abnormal correction values ​​from participating in closed-loop regulation.

[0055] For example, continuing with the aforementioned interface state. The temperature is 95.3℃, the current throttle valve opening θ is 45%, and the calibration reference balance force table gives the following values ​​for this opening and a reference water temperature of 60℃. The pressure in the balance chamber is approximately 5500N. With the effective working area of ​​the balance cavity for The product of, that is . Falling into the 80℃ to 100℃ range, Corrected equilibrium forces Counterexample: If the actual measured water temperature The value is 130℃, which exceeds the upper limit of 120℃ on the calibration table. The interpolator takes the endpoint value. If the value is equal to 0.91, set the water temperature exceedance flag and record the event via the protection mechanism. The compensation process continues under limited conditions without interrupting the adjustment flow. During the boundary crossing period, the compensation coefficient remains at the endpoint value and does not continue to decrease in the outward direction to avoid excessive correction of the balancing force, which could lead to distortion of the residual load estimation. After the water temperature boundary crossing flag is set, the adjustment steps still use the balancing force corrected by the endpoint coefficient in the closed loop. At the same time, the protection steps record the water temperature boundary crossing event for operation and maintenance to handle. Once the water temperature returns to the calibrated range and stabilizes for several consecutive cycles, the boundary crossing flag is automatically cleared, and the compensation resumes interpolation calculation. The output status includes the corrected balancing force. 5209N, compensation coefficient The value is 0.9471, indicating the compensation flag is normal. The compensation flag distinguishes between three levels: normal, out-of-range, and abnormal. The normal level indicates that the water temperature is within the calibrated range and the correction amount is not exceeded; the out-of-range level indicates that the water temperature exceeds the calibrated range and the endpoint value is used; the abnormal level indicates that the consistency check of the compensation table has failed or the correction amount has exceeded the limit. The adjustment execution module 12 determines the method of adopting the compensation result according to the compensation flag. The out-of-range level continues to use the endpoint coefficient, and the abnormal level is degraded to the uncompensated benchmark value.

[0056] S500: Determine the residual load of the thrust bearing based on the corrected total axial force and the corrected balancing force, and adjust the opening of the throttle valve of the balancing chamber to minimize the residual load.

[0057] Adjustment execution module 12 reads the total axial force after correction. With the corrected balance force Calculate the signed balance error A value greater than 0 indicates insufficient balancing force, while a value less than 0 indicates excessive balancing force. (This refers to the residual load on the thrust bearing.) The opening increment is calculated according to the signed error proportional-integral control law, and the throttle valve opening command θ is updated. After the actuator actuates, it waits for the next sampling cycle. This cycle repeats until the residual load converges to the minimum allowable range. Figure 5 As shown, the adjustment process is as follows: residual load calculation, signed error proportional-integral calculation, opening limit, actuator command output and valve position feedback detection.

[0058] In a signed error proportional-integral control law, the opening increment in the k-th cycle... Opening command And limited to 0% to 100%. Proportional gain. With integral gain Calibration is based on the actuator stroke and the pump axial force level. The value ranges from 0.001 to 0.005 per Newton; in this embodiment, 0.002 per Newton is used. The value range is 0.0002 to 0.0008 per Newton-second; in this embodiment, 0.0004 per Newton-second is used, and the sampling period is... It takes 0.1 seconds. and The calibration is based on the following: the axial force change is on the order of several thousand Newtons, corresponding to the full range of the opening adjustment. The proportional gain is calculated by back-calculating the change in balance force corresponding to each 1% change in opening, i.e., the opening force slope m, to ensure that the single-cycle adjustment does not exceed the rate of change limit and the convergence speed meets the response requirements. During calibration, a typical working condition step is simulated on the test bench, with the tuning target being no overshoot in the opening response and a residual load convergence period not exceeding 30 sampling periods. The adjustments are made sequentially. and Until the goal is achieved.

[0059] When the opening degree increases, the pressure relief in the balance chamber reduces the balancing force. The opening degree and the balancing force are monotonically negatively correlated, and the slope m of the opening force is determined by calibration. The relationship between the opening degree and the balancing force is described by the actuator model. In this embodiment, a linear approximation is used. The balancing force is at its maximum value when the throttle valve is fully closed. For every 1% increase in opening degree, the balancing force decreases linearly according to the opening force slope m. m is greater than 0 to ensure correct control polarity determination. m is obtained by fitting the calibration steps after changing the opening degree point by point under a fixed water temperature and measuring the pressure in the balance chamber. During calibration, the flow rate and rotation speed are kept constant, and the corresponding points of opening degree and balance chamber pressure are recorded. m and the maximum balancing force when fully closed are determined by linear fitting. When the fitting residual exceeds the limit, a piecewise linear model is used, with the piecewise nodes being 25%, 50%, and 75% of the opening degree. The dynamic process of pressure relief in the balance chamber has a time constant. The opening degree change rate limit suppresses the pressure relief impact and avoids oscillation of the balancing force.

[0060] When e is greater than 0, the opening increment is negative, and the throttle valve is closed to allow the balancing force to recover. When e is less than 0, the opening increment is automatically positive, and the throttle valve is opened to allow the balancing force to fall. The control polarity automatically flips with the equilibrium point to ensure bidirectional convergence. The opening change rate is limited to no more than 2% per cycle to suppress impact. When the opening reaches the limit, the integral accumulation is frozen to prevent integral saturation. The absolute value of the residual load does not exceed the convergence threshold. That is, convergence is determined at 200N, and the current opening is maintained. Convergence threshold. The acceptable upper limit for the residual load of the thrust bearing is determined based on the following: the safe percentage of the rated dynamic load of the thrust bearing is approximately 40,000 N, and 200 N is far less than the residual load under normal operating conditions, which is in the thousands of Newtons. Values ​​below this do not affect bearing life, while avoiding excessively small thresholds that could lead to frequent opening movements. When bearing specifications or operating conditions fluctuate, the value can be recalibrated according to the safe percentage of the rated dynamic load. The integral anti-saturation freezing mechanism works as follows: when the opening command reaches 0% or 100% limit, the integral accumulation term maintains its current value and does not continue to accumulate with the error. Integration resumes after the opening exceeds the limit, preventing overshoot when the opening reverses due to continuous increase in integral during the limit period. During the limit period, the proportional term is still calculated based on the error, providing the ability to quickly break free from the limit. Convergence determination uses a hysteresis method. When the absolute value of the residual load does not exceed 200 N for several consecutive cycles, convergence is determined and the opening is locked. After locking, comparisons are still performed every cycle. When the residual load exceeds the limit again, the lock is released and adjustment resumes, preventing frequent switching of critical states.

[0061] The convergence threshold and opening force parameters of the throttle valve adjustment step are loaded by the calibration step. In alternative implementations, a proportional-integral-derivative (PID) control law with an added derivative term can be used to suppress oscillations, or a feedforward plus feedback composite control can be used. The PID control in this embodiment already meets the convergence requirements. The PID control law adds a derivative term to the proportional-integral control law. The derivative term predicts the adjustment direction according to the error rate of change, and decelerates in advance near the equilibrium point. It is suitable for scenarios where the dynamic time constant of the pressure relief of the balance chamber is large and overshoot oscillations are prone to occur. The feedforward plus feedback composite control uses the balance force model corresponding to the current opening as the feedforward reference. The feedback term only compensates for the model error, which can speed up the response and reduce the steady-state error. It is suitable for scenarios where the operating conditions change frequently. The parameters and limiting rules of the alternative control law are consistent with those of the PID control, and are subject to opening limit, rate of change limit, and integral anti-saturation constraints.

[0062] For example, continuing the aforementioned interface state, It is 5794N. It is 5209N. It is 0.002 per cow. The value is 0.0004 N / s, and the current opening θ is 45%. The first period's e is 585 N, and the integral accumulation term is... Newton-second, opening increment In the second cycle, the opening is reduced, causing the pressure in the balance chamber to rise. The balancing force increases to 5257 N, e becomes 537 N, the integral accumulation term is 112.2 N / s, Δθ becomes -1.12%, and θ is updated to 42.69%. In the third cycle, the balancing force increases to 5302 N, e becomes 492 N, the integral accumulation term is 161.4 N / s, Δθ becomes -1.05%, and θ is updated to 41.64%. The residual load decreases cycle by cycle: 585 N, 537 N, 492 N. Figure 9 As shown, the residual load converges with the iteration cycle, and the iteration continues until it is no greater than 200N to determine convergence; when it approaches the equilibrium point, the error crosses zero, and the sign of the opening increment in subsequent cycles is automatically flipped. In steady state, the residual load converges within the threshold.

[0063] The mathematical characteristics of the convergence process are as follows: the opening increment in the k-th cycle changes in the same direction as the error; the opening adjustment causes the balancing force to move along the direction of decreasing error; the error amplitude decreases approximately proportionally with each iteration; the attenuation ratio is determined by the product of the proportional gain and the slope of the opening force; in this embodiment, the product is... This corresponds to a single-cycle error decay of approximately 8%, with the initial error of 585N decaying to below 200N after approximately 13 cycles. Figure 9 The convergence trend shown is consistent; the integral term provides residual error elimination capability in steady state, ensuring that the opening command stabilizes near the equilibrium point rather than merely approximating it. Counterexample 1: If the axial force suddenly drops to 4800N in the next cycle while the balancing force is temporarily 5302N, e is -502N, resulting in excess balancing force. The proportional term automatically becomes positive, increasing the opening, decreasing the balancing force, and causing the error to recover and approach zero, preventing further closing in the incorrect direction. Counterexample 2: If the valve position feedback shows an actual opening deviation greater than 5% from the command, an actuator malfunction is determined, adjustment is stopped, a fault flag is set, and the protection procedure is recorded to prevent open-loop misadjustment.

[0064] The valve position feedback closed loop performs a command and feedback comparison once in each sampling cycle. The feedback deviation threshold of 5% is calibrated based on the actuator stroke measurement accuracy and dead zone. A deviation less than this value is considered complete execution; a deviation greater than this value and persisting for several cycles is considered a fault, avoiding false alarms due to momentary jitter in a single cycle. After an actuator fault is determined, the integral accumulation value is reset to zero to prevent residual integral values ​​from causing opening jumps after fault recovery. During an actuator fault, the balance chamber throttle valve maintains its current mechanical position, and the adjustment steps no longer output new opening commands. The protection module enters a pre-warning state according to a conservative strategy, awaiting manual inspection and repositioning. Output states include θ being 41.64%, The value is 492N and the convergence indicator is in the process of convergence.

[0065] like Figure 5 As shown, the adjustment process consists of residual load calculation, signed error proportional-integral calculation, opening limit, actuator command output, and valve position feedback detection. Each step is completed sequentially within the same sampling period. At the end of the period, the opening command is output and sent through the actuator channel. After the convergence flag is set, the adjustment steps still maintain periodic comparison. When the residual load fluctuates within the convergence threshold, the current opening is maintained. Once the operating conditions change and the residual load exceeds the limit again, the convergence lock is immediately released and adjustment is resumed. The switching between the convergence state and the adjustment state is achieved by hysteresis judgment to avoid frequent switching near the threshold. The output state is updated with the sampling period for the protection steps to read and perform threshold comparison.

[0066] S600 performs graded early warning and linkage protection based on the corrected total axial force and residual load.

[0067] Protection module 13 reads the total axial force after correction. Residual load Along with a control threshold table, which is loaded during the calibration process, including axial force warning thresholds. Axial force protection threshold Residual load early warning threshold With residual load protection threshold .Compare and , ,Compare and , When any threshold is reached, an early warning signal is output; when any threshold is reached, a linkage protection action is executed, and an event log containing timestamps, operating condition snapshots, and actions is recorded.

[0068] The hierarchical protection logic is as follows: Not less than or Not less than Execute a coordinated shutdown when necessary; otherwise, if Not less than or Not less than The system outputs a warning and reduces speed when the axial force falls below 90% of the warning threshold, otherwise it is considered to be in normal condition. Warning cancellation is performed according to the hysteresis interval. The warning cancellation threshold is 90% of the warning threshold, meaning the warning is cancelled after the axial force falls below 90% of the warning threshold and remains below it for a set duration. The hysteresis interval is used to prevent frequent fluctuations near the threshold. The hysteresis interval is implemented as follows: the warning trigger edge is the warning threshold itself, and the warning cancellation edge is 90% of the warning threshold, forming a 10% hysteresis interval between the two thresholds. When the axial force falls within the hysteresis interval, the current state is maintained without reversal. The warning is only cancelled after the force crosses the cancellation edge and remains below it for a set duration. The set duration is accumulated by a hysteresis timer, which increments when the axial force is below the cancellation threshold and resets to zero when it is above the cancellation threshold. After 60 seconds, the system is considered stable and the warning is cancelled.

[0069] The hysteresis mechanism is also applied to the residual load warning release, with the residual load release threshold set at 90% of the residual load warning threshold. Protection actions are executed in stages: when the residual load exceeds the warning threshold, the speed is first reduced to a safe speed; when the axial force exceeds the protection threshold, a shutdown is initiated. These two types of judgments are performed in parallel. The execution path for speed reduction and shutdown is as follows: speed reduction is initiated by sending a speed command via the inverter communication interface, gradually reducing the speed at a ramp rate of 5% of the rated speed per second to avoid water hammer and axial force impact caused by rapid speed reduction; shutdown is initiated by disconnecting the motor power supply via a contactor and simultaneously closing the throttle valve to the fully closed position. Simultaneously, the controller state is reset after the event log is completed to prevent data loss during shutdown. During the execution of protection actions, the soft measurement and verification steps stop outputting new estimated values; the protection module completes the judgment based on the final estimated value at the time of execution.

[0070] The graded protection state machine manages the transition between three states: normal, warning, and protection. When the normal state reaches the warning threshold, it transitions to the warning state. When the warning state reaches the protection threshold or continues to exceed the warning timeout, it transitions to the protection state. The protection state returns to the normal state after manual reset and confirmation that the fault has been cleared. The state machine executes transition judgments cycle by cycle according to the sampling period. State transitions are accompanied by an action sequence: when transitioning from normal to warning, a warning signal is output and a timeout timer is started; when transitioning from warning to protection, a stop command is output and the reset interface is locked; when transitioning from protection to normal, both a manual reset signal and fault clearance confirmation must be met simultaneously. If either condition is missing, the protection state remains in effect. The timeout timer triggers an upgrade after 60 seconds of warning status and is reset to zero when the warning is lifted. The thresholds loaded in the calibration steps include an axial force warning threshold of 28000N and an axial force protection threshold of 32000N, as well as a residual load warning threshold of 8000N and a residual load protection threshold of 10000N, all determined by the thrust bearing's rated load and safety factor.

[0071] The threshold determination rules are as follows: The rated dynamic load of the thrust bearing is given in the bearing specification sheet. The protection threshold is 80% of the rated dynamic load, and the warning threshold is 87.5% of the protection threshold. That is, the axial force protection threshold of 32000N and the warning threshold of 28000N satisfy this ratio. The residual load threshold is determined according to the bearing static load and life requirements. The residual load protection threshold of 10000N corresponds to the limit of the rated static load, and the warning threshold is 80% of the protection threshold. When loading the threshold, the warning threshold is checked to be less than the protection threshold, and both are not greater than the bearing rated load. If the check fails, a conservative value is loaded and an alarm is triggered. An example of a warning timeout is 60 seconds without relief, in which case it is upgraded to the protection state. When the soft measurement degradation flag or actuator fault flag is set, the strategy is downgraded to conservative and directly enters the warning state. The event log uses a circular buffer with a capacity of 100 entries, each containing a timestamp. , Throttle valve opening θ, water temperature The event log is written according to the following rules: logs are only written during state transitions, warning triggering and deactivation, protection action execution and reset, and fault flag setting and clearing. Logs are not written during normal cycles to prevent the circular buffer from being filled with regular data. Once the circular buffer is full, the oldest entry is overwritten to ensure that the latest events are always available. Alarm output methods include local audible and visual alarms and host computer communication messages. Audible and visual alarms are used to alert on-site personnel, while host computer messages carry the event type, timestamp, and operating condition snapshot for remote monitoring and maintenance analysis.

[0072] For example, continuing with the aforementioned interface state. It is 5794N. The value is 492N. In normal scenarios, 5794N is less than 28000N and 492N is less than 8000N, indicating a normal state with no action and no new events recorded in the event log. In warning scenarios, if the operating conditions deteriorate... If the speed increases to 28500N, but is not less than 28000N and less than 32000N, a warning state is determined, a warning signal is output, the frequency converter reduces its speed to 85% of the rated speed, the soft measurement and verification steps continue to estimate, the axial force decreases with the speed reduction, and after decreasing to 25200N (less than 90% of 28000N) and remaining there for 60 seconds, the warning is lifted and full-speed operation resumes. The basis for taking 85% of the rated speed as the speed reduction target value is: the axial force changes approximately with the square of the speed, and the theoretical reduction in axial force when the speed is reduced to 85% is about 28%, which is sufficient to pull the axial force under the warning condition back within the threshold, while maintaining the pump's head capacity to meet the minimum downhole water supply requirements; the speed reduction is completed in steps according to the frequency converter ramp time to avoid hydraulic shock caused by sudden speed changes. The design intent of holding the hysteresis interval for 60 seconds is that after the speed reduction, the operating conditions need to be confirmed to be stable through multiple sampling cycles. 60 seconds is more than 10 complete adjustment cycles, which can filter out short-term fluctuations without delaying the restoration of normal operation. The holding time is loaded by the calibration steps and can be adjusted according to the well conditions and load fluctuation characteristics.

[0073] In the protected scenario, if If the current rises to 33000N and is not less than 32000N, determine the protection status, execute the linkage shutdown, and record the event {timestamp, ...} 33000N, θ is the current opening degree. The system is in a standby state, awaiting manual reset (for water temperature and verification results). The shutdown process includes disconnecting the motor power, de-energizing the throttle valve actuator to maintain its current position, and outputting a shutdown message. After shutdown, the soft measurement and verification steps enter standby mode, awaiting manual reset and confirmation that the fault has been eliminated before re-initialization. The output status is normal, the event log shows no new entries, and the system is operating in a closed loop. The output status includes the current state machine status, warning signal level, and protection action flag, for display on the host computer monitoring interface and for maintenance personnel to review. Under normal conditions, the warning signal remains low, the protection action flag is inactive, and the system continuously operates at the sampling period. It operates in a closed loop until a shutdown command or protection trigger is received.

[0074] Based on the same inventive concept, embodiments of this application also provide an adaptive balancing system for axial force conditions of a deep well hot water submersible pump. For example... Figure 6 As shown, the system includes a parameter library module 7, a signal acquisition module 8, a soft measurement engine 9, a calibration module 10, a compensation module 11, an adjustment execution module 12, and a protection module 13. Each module can be implemented by the processor executing instructions from the memory. Figure 10 As shown, the system also includes a deep well hot water submersible pump 1, a balance chamber 2, a throttle valve 3, a thrust bearing 4, a controller 5, and a ground sensor group 6. The ground sensor group 6 is connected to the controller 5 via a cable, and the controller 5 is connected to the actuator of the throttle valve 3.

[0075] Parameter library module 7 stores pump nameplate parameters, discrete points of factory characteristic curves, water temperature compensation coefficient tables, and control threshold tables, providing calibration data for each module. Signal acquisition module 8 connects to ground sensor group 6, acquiring outlet pressure, flow rate, motor input power, speed, and water temperature signals at a fixed sampling period. After performing sliding window filtering and amplitude limiting detection, it outputs a synchronization signal. Soft measurement engine 9 reads the flow rate and speed output from signal acquisition module 8, calls the characteristic curve function calibrated by parameter library module 7, and estimates the total axial force according to the impeller stage axial force model. Verification module 10 reads the total head and total axial force output from soft measurement engine 9, verifies them using the measured motor input power value, and cross-checks them with the outlet pressure, outputting the corrected total axial force. Compensation module 11 reads the reference balance force corresponding to the measured water temperature and current opening, corrects the balance force according to the water temperature compensation coefficient, and outputs the corrected balance force. Adjustment execution module 12 determines the residual load based on the corrected total axial force and the corrected balance force, outputs the throttle valve opening command according to the signed error proportional-integral control law, and the actuator of throttle valve 3 operates according to the command. After comparison and correction, the total axial force and residual load of the protection module 13 are the same as the warning threshold and protection threshold. The module outputs a warning signal and linkage protection action, and manages the state transition through a hierarchical protection state machine.

[0076] The data flow between modules is as follows: Parameter library module 7 provides calibration data to soft measurement engine 9, compensation module 11, and protection module 13; signal acquisition module 8 provides preprocessed signals to soft measurement engine 9, calibration module 10, and compensation module 11; soft measurement engine 9 provides estimation results to calibration module 10 and protection module 13; calibration module 10 provides correction results to adjustment execution module 12 and protection module 13; compensation module 11 provides corrected balancing force to adjustment execution module 12; adjustment execution module 12 provides opening degree and residual load to protection module 13. Each module is executed periodically by the processor within controller 5, with the scheduler operating on a sampling period. To maintain a consistent pace, each module task is woken up sequentially. Modules exchange data via shared memory. Data slots are organized in a way that allows only writes by the producer and reads by the consumer to avoid write conflicts. After the system powers on, the parameter library module 7 is loaded and calibrated first. After loading is complete and the self-test passes, the signal acquisition module 8 is started. Subsequent modules are started step by step as valid data becomes available, forming a complete closed-loop cycle.

[0077] During system operation, soft sensor degradation flags, actuator fault flags, or water temperature exceeding limits flags are written to shared memory by the corresponding modules for protection module 13 to read and trigger actions. Data is transferred between modules via the shared memory area, which is divided into independent data slots for each module. Each data slot uses a double-buffered structure. The writing module updates the backup buffer first, then switches the active buffer pointer after completion. The reading module always reads from the active buffer to avoid reading partially updated data. The flag handshake protocol stipulates that the writing module sets the update flag after data update, and the reading module clears the update flag after consuming data. The writing and reading sides interact according to the flag bit timing to prevent unconsumed data from being overwritten. The execution cycle and sampling cycle of each module are also specified. Alignment is achieved by executing parameters sequentially within a single cycle, including loading, acquisition, estimation, verification, compensation, adjustment, and protection. The output of the preceding module is consumed by the subsequent module within the cycle, forming a cycle-level closed-loop pipeline. When the opening of throttle valve 3 increases, the pressure in the balance chamber 2 is released, reducing the balancing force. The adjustment execution module 12 selects to close or open throttle valve 3 according to the polarity of the signed balance error, causing the residual load on the thrust bearing 4 to converge to the minimum allowable range.

[0078] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit described above can be implemented in hardware.

[0079] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An adaptive balancing method for axial force conditions of a deep well hot water submersible pump, characterized in that, include: Obtain the pump characteristic curve, impeller stage parameters, and water temperature compensation coefficient table, and calibrate the characteristic curve function; Collect pump outlet pressure, flow rate, motor input power, speed and water temperature operating parameters, and preprocess the operating parameters; Based on the aforementioned characteristic curves, the total axial force is estimated according to the impeller stage axial force model. The axial force of each stage is a function of the pressure difference between the impeller inlet and outlet and the projected area of ​​the hub, and is accumulated stage by stage. The total axial force is verified using the motor input power and cross-checked using the outlet pressure. The total axial force is then corrected based on the verification results. Determine the balancing force of the balancing chamber throttle valve at the current opening, and correct the balancing force according to the water temperature compensation coefficient; The residual load of the thrust bearing is determined based on the corrected total axial force and the corrected balancing force. The opening of the throttle valve in the balancing chamber is adjusted to minimize the residual load.

2. The method according to claim 1, characterized in that, The preprocessing includes: performing sliding window filtering on each operating parameter to remove outliers that exceed the limit and retaining the most recent valid value; when the number of continuously removed outliers exceeds the window length, setting a signal quality degradation flag and reducing the verification weight.

3. The method according to claim 1, characterized in that, The estimation of total axial force includes: converting the characteristic curve to the current rotational speed according to the similarity law, and determining the total head by fitting a function.

4. The method according to claim 3, characterized in that, The estimated total axial force further includes: distributing the total head to the inlet and outlet pressure differences of each impeller according to the stage group distribution coefficient, and ensuring that the inlet and outlet pressure differences of each stage satisfy the energy conservation constraint.

5. The method according to claim 4, characterized in that, The estimation of the total axial force further includes: determining the axial force of each stage based on the inlet and outlet pressure difference and the hub projection area of ​​each stage impeller, and accumulating the axial forces of each stage to the total axial force.

6. The method according to claim 1, characterized in that, The step of verifying the total axial force using the motor input power includes: determining the total head based on the flow rate, the characteristic curve, and the current rotational speed; determining the hydraulic power based on the flow rate and the total head; converting the hydraulic power into an estimated value of the motor input power using an efficiency chain; verifying the total axial force according to the relative deviation between the estimated value of the motor input power and the actual motor input power, and correcting it according to the deviation limit; and reducing the verification weight when the outlet pressure exceeds the cross-verification limit.

7. The method according to claim 1, characterized in that, The step of correcting the balancing force according to the water temperature compensation coefficient includes: determining the water temperature compensation coefficient according to the measured water temperature using an interpolation table or a reference formula; when the measured water temperature exceeds the applicable range of the interpolation table or the reference formula, taking the endpoint value as the water temperature compensation coefficient and outputting an out-of-bounds alarm.

8. The method according to claim 1, characterized in that, The adjustment of the opening of the balance chamber throttle valve includes: determining the signed balance error based on the corrected total axial force and the corrected balance force; determining the opening increment according to the signed error proportional integral control law; applying opening limit and rate of change limit to the opening increment, and freezing the integral accumulation during the limit; and determining convergence and maintaining the current opening when the absolute value of the signed balance error is not greater than the convergence threshold.

9. The method according to claim 1, characterized in that, The method further includes: When the corrected total axial force is not less than the warning threshold or the residual load is not less than the residual load warning threshold, a warning signal is output or speed reduction is executed. When the corrected total axial force is not less than the protection threshold or the residual load is not less than the residual load protection threshold, a coordinated shutdown is executed; the warning is lifted according to the hysteresis interval, and the transfer between the warning state and the protection state is managed by the hierarchical protection state machine.

10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of any one of claims 1 to 9.

Images