Safe operation control method for permanent magnet motor centrifugal pump

Abstract

The invention discloses a safe operation control method for a permanent magnet motor centrifugal pump, and particularly relates to the technical field of operation control of permanent magnet motor driving fluid equipment. Operation state parameters of a permanent magnet motor side and a centrifugal pump side are synchronously collected in the stages of covering starting, steady-state operation and working condition change, a continuous operation state sequence is formed, and a dynamic feature vector is constructed; judging whether an abnormal development stage exists or not, and forming a judgment result when an abnormal development trend exists; constructing and predicting a plurality of control action sequences according to a judgment result, determining an operation interval, and selecting the control action sequences for execution; according to the method, the abnormal development trend is recognized by constructing the dynamic feature vector, the influence of different control actions on the operation state is predicted on the basis, and the control action sequence meeting the operation safety requirement is selected for execution, so that the abnormal recognition accuracy is improved, the control decision risk is reduced, and the operation safety of the permanent magnet motor centrifugal pump is enhanced.

Description

Technical Field

[0001] This invention relates to the field of operation control technology for fluid equipment driven by permanent magnet motors, and more specifically, to a method for safe operation control of a centrifugal pump driven by a permanent magnet motor. Background Technology

[0002] Permanent magnet motor centrifugal pumps are widely used in industrial conveying, chemical processes, energy systems, and water supply and drainage due to their high energy efficiency, wide speed range, and compact structure. With the increasing complexity of application scenarios, permanent magnet motor centrifugal pumps frequently experience stages of start-up, steady-state operation, sudden load changes, and valve adjustments during actual operation. Their operating state is simultaneously influenced by the output characteristics of the permanent magnet motor, the load characteristics of the centrifugal pump, and the fluid operating state, resulting in a clear dynamic coupling characteristic in their overall operating behavior.

[0003] Existing safe operation control methods for permanent magnet motor centrifugal pumps typically use electrical parameters such as current, voltage, and temperature as the main monitoring targets, and judge abnormal operating conditions by setting fixed thresholds. When a parameter exceeds the limit, protective control strategies such as speed limiting, current limiting, or shutdown are directly executed. This type of method can prevent equipment damage under extreme operating conditions to a certain extent, but its judgment logic is mostly based on the static comparison of a single parameter or a small number of parameters, which makes it difficult to reflect the changing trend of the operating status over time, and also makes it difficult to characterize the dynamic coupling relationship between the permanent magnet motor output, the centrifugal pump load, and the fluid operating state.

[0004] Under non-steady-state conditions, especially during the startup phase or when the operating conditions change rapidly, the operating parameters themselves have normal transient fluctuations. Control methods based on static thresholds are prone to misjudgment, leading to frequent triggering of protection actions. Furthermore, during the gradual development of certain anomalies, the parameters may not have exceeded the threshold, but the energy transfer relationship or dynamic response characteristics within the system have already deviated significantly. Existing methods are difficult to identify in a timely manner, thus delaying the intervention opportunity.

[0005] More importantly, existing technologies, upon detecting abnormal operating conditions, typically do not evaluate the overall dynamic response of the system after the control action is implemented; instead, they directly execute a preset protection strategy. Under complex operating conditions, this type of control strategy may alter the dynamic matching relationship between the permanent magnet motor and the centrifugal pump, causing the centrifugal pump to enter a cavitation-sensitive zone, a structural resonance danger zone, or an operating range with insufficient cooling capacity under the control action, thereby inducing secondary failures caused by the control strategy itself. Therefore, this invention proposes a safe operation control method for a permanent magnet motor centrifugal pump to address the aforementioned problems. Summary of the Invention

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for safe operation control of a permanent magnet motor centrifugal pump includes the following steps: During the operation of the permanent magnet motor centrifugal pump, within the operating range covering the start-up, steady-state operation and operating condition change stages, the operating status parameters of the permanent magnet motor side and the centrifugal pump side are collected synchronously, and unified time reference alignment and transient interference suppression processing are performed to form a continuous operating status sequence. Based on the operating state sequence, the rate of change characteristics of the operating state parameters on the permanent magnet motor side and the centrifugal pump side with time, the energy transfer characteristics reflecting the relationship between power input and load response, and the phase consistency characteristics between the operating state parameters on the permanent magnet motor side and the centrifugal pump side are calculated. A dynamic feature vector reflecting the coupling relationship between the permanent magnet motor output, the centrifugal pump load and the fluid operating state is constructed. The operating status of the permanent magnet motor centrifugal pump is analyzed based on dynamic feature vectors to determine whether the operating status is in an abnormal development stage. When an abnormal development trend is determined, the influence of the preset control action on the abnormal development trend under the current operating status is combined to form a judgment result reflecting the severity of the abnormality and the influence of the control action. Based on the judgment results, multiple control action sequences are constructed, and the changing trend of the permanent magnet motor centrifugal pump's operating status after the execution of each control action sequence is predicted and analyzed to determine the operating range corresponding to each control action sequence. Then, the control action sequence that meets the operating safety requirements is selected and executed.

[0007] In a preferred embodiment, when acquiring parameters, the operating status parameters of the permanent magnet motor side acquired include at least current, voltage, speed, electromagnetic torque, and winding temperature. The operating status parameters collected from the centrifugal pump side should include at least the outlet pressure, inlet pressure, flow rate, vibration, and pump body temperature. The operating status parameters of the permanent magnet motor and the centrifugal pump are collected in parallel using the same sampling period, and the collection time is marked. The parallel acquisition results are merged into the same recording frame according to the acquisition time, which serves as the basic unit of the running state sequence.

[0008] In a preferred embodiment, the unified time reference alignment and transient interference suppression processing includes: Establish a unified time reference and map the acquisition times of the permanent magnet motor side acquisition channel and the centrifugal pump side acquisition channel to a unified time axis; Delay compensation is performed based on the fixed channel delay, and data at different sampling frequencies are resampled to ensure that each recorded frame has a consistent time interval. Identify abrupt peaks and sample loss discontinuities, apply sliding median suppression to peaks, restore continuity for sample loss by time axis interpolation, and output an uninterrupted running state sequence according to a unified time axis.

[0009] In a preferred embodiment, the rate of change characteristic refers to: splitting the permanent magnet motor side operating status parameters and the centrifugal pump side operating status parameters into multiple single-parameter time series according to parameter names, calculating the parameter value difference for each single-parameter time series at adjacent sampling times, and dividing the parameter value difference by the sampling interval value to obtain the rate of change characteristic value corresponding to the parameter. Energy transfer characteristics refer to: selecting voltage and current values ​​from the operating status parameters of the permanent magnet motor to calculate the input power value, which is the product of the voltage and current values; selecting outlet pressure, inlet pressure, and flow rate values ​​from the operating status parameters of the centrifugal pump to calculate the hydraulic power value, where the pressure difference is the difference between the outlet pressure and the inlet pressure, and the hydraulic power value is the product of the pressure difference and the flow rate. The energy transfer characteristic values ​​are calculated based on the input power value and the hydraulic power value, including: the energy ratio characteristic value is the sum of the hydraulic power value and the input power value and the zero-prevention constant; the energy difference characteristic value is the sum of the input power value and the hydraulic power value; and the zero-prevention constant is a preset minimum positive number to ensure that the denominator of the division is not zero.

[0010] In a preferred embodiment, the phase consistency feature refers to: Establish a parameter pair set, which includes at least one pair of permanent magnet motor side operating state parameters and centrifugal pump side operating state parameters. For each parameter pair, calculate the normalized cross-correlation value for different time offsets within a fixed time window. The normalized cross-correlation value is the sum of the product of the two sequences at that time offset, divided by the square root of the sum of the squares of the amplitudes of the two sequences. For each parameter pair, select the time offset value corresponding to the maximum normalized cross-correlation value as the time offset characteristic value of that parameter pair.

[0011] When constructing the dynamic feature vector, the main frequency value is calculated for the permanent magnet motor side parameter corresponding to the time offset feature value within the same time window. The main frequency value is the reciprocal of the interval between adjacent zero crossings. The time offset feature value is converted into the phase consistency feature value. The phase consistency feature value is the main frequency value multiplied by the time offset feature value and then multiplied by 360. The rate of change characteristic value, energy transfer characteristic value, and phase consistency characteristic value are combined in a predetermined order and then subjected to amplitude normalization to form a dynamic characteristic vector.

[0012] In a preferred embodiment, the determination result based on the dynamic feature vector includes: The risk index is obtained by subtracting each component of the dynamic feature vector from the safety benchmark vector and taking the absolute value, and then summing them according to preset weights. Calculate the risk index difference between two adjacent time windows, and compare the risk index and the risk index difference with the corresponding thresholds respectively. When the risk index exceeds the risk threshold and the risk index difference exceeds the growth threshold, it is determined that the abnormal development stage has been entered. The severity of the abnormality is the product of the risk index and the risk index difference. For each preset control action, a short-term trial with a fixed amplitude is performed while keeping the other control quantities unchanged. The risk index after the trial is recalculated, and the degree of influence of the control action is taken as the difference between the risk index before and after the trial. The severity of the anomaly is combined with the degree of impact of the control action to output the judgment result.

[0013] In a preferred embodiment, during the process of obtaining the risk index by weighted summation according to preset weights, the parameter types and arrangement order in the dynamic feature vector and the safety benchmark vector remain consistent. For each parameter type, the parameter value in the dynamic feature vector is compared with the corresponding safety benchmark in the safety benchmark vector. When the parameter value is higher than the safety benchmark, the first weight preset for that parameter type is selected; when the parameter value is not higher than the safety benchmark, the second weight preset for that parameter type is selected. The sum of the first weight and the second weight is one. After all parameter types have completed weight selection, the sum of all selected weights is used as the normalization benchmark to normalize each weight, and the deviation of each parameter is weighted and summed based on the normalized weights.

[0014] In a preferred embodiment, determining the operating interval corresponding to each control action sequence refers to: Based on the severity of the anomaly and the degree of influence of the control action in the judgment result, corresponding discrete adjustment ranges and durations are set for the speed setpoint, torque limit value, current limit value and shutdown trigger condition, and they are combined in a predetermined execution order to form a variety of control action sequences. Using the dynamic feature vector corresponding to the end of the current operating state sequence as the initial state, the operating state parameters of the permanent magnet motor and the centrifugal pump are calculated piecewise and recursively using the rate of change feature in the dynamic feature vector to obtain the predicted operating state sequence of each control action sequence under the corresponding execution stage. The risk index is calculated segment by segment based on the predicted operating status sequence, and the risk level interval is divided according to the risk index threshold. The time period in which the risk index is continuously maintained within the same risk level interval is defined as the corresponding operating interval. The boundary of the operating range is checked by using the energy transfer characteristics and phase consistency characteristics corresponding to the predicted operating state sequence. When the energy transfer characteristics or phase consistency characteristics exceed the preset safety range, the corresponding time period is removed from the operating range to obtain the effective operating range of each control action sequence.

[0015] In a preferred embodiment, selecting and executing a sequence of control actions that meet operational safety requirements includes: Within the effective operating range, candidate control action sequences are selected. The selection criteria include that the risk index corresponding to the predicted operating state sequence does not enter the preset prohibited operating zone, and the difference in risk index is decreasing. For the selected candidate control action sequences, the disturbance amplitude index is calculated. The disturbance amplitude index is the normalized absolute value of the change in speed setpoint, the change in torque limit value, and the change in current limit value. The weighting weights are determined by the combination of parameter weights corresponding to speed, electromagnetic torque, and current in the risk index calculation. After normalization according to the weight sum, the weighted sum is calculated, and the control action sequence with the smallest disturbance amplitude index is selected as the execution sequence. Control commands are issued in segments according to the execution order and duration of each control action in the execution sequence, so that the permanent magnet motor centrifugal pump runs segment by segment according to the execution sequence.

[0016] The technical effects and advantages of this invention are as follows: 1. This invention synchronously collects operating status parameters of both the permanent magnet motor and centrifugal pump sides within the operating range covering startup, steady-state operation, and operating condition change stages. The collected operating status parameters are then aligned to a unified time reference and subjected to transient interference suppression to form a continuous operating status sequence. This ensures that the operating status of the permanent magnet motor and centrifugal pump sides maintains consistency and continuity in the time dimension, avoiding distortion caused by asynchronous acquisition, signal jitter, or transient interference. Consequently, subsequent operating status analysis is based on a true and continuous operating status, improving the completeness and reliability of operating status perception. This is especially true during startup and operating condition change stages, where it more accurately reflects the dynamic changes in operating status.

[0017] 2. This invention calculates the rate of change characteristics, energy transfer characteristics, and phase consistency characteristics based on the operating state sequence, and further constructs a dynamic feature vector reflecting the coupling relationship between the permanent magnet motor output, the centrifugal pump load, and the fluid operating state. This makes the operating state analysis no longer limited to a single parameter or static threshold judgment, but comprehensively characterizes the operating state from multiple dimensions such as the parameter change rate, the relationship between power input and load response, and the dynamic response relationship between parameters. This can more comprehensively reflect the dynamic behavior characteristics of the permanent magnet motor centrifugal pump under different operating stages, improve the ability to identify the trend of operating state changes, and provide a more sufficient and stable basis for judging whether the operating state is in an abnormal development stage.

[0018] 3. After determining that the operating state is in an abnormal development stage based on dynamic feature vectors, this invention further combines the influence of preset control actions on the abnormal development trend under the current operating state to form a judgment result. On this basis, multiple control action sequences are constructed, and the change trend of the permanent magnet motor centrifugal pump's operating state after the execution of each control action sequence is predicted and analyzed. Then, the operating range corresponding to each control action sequence is determined, and the control action sequence that meets the operating safety requirements is selected for execution. This transforms the control decision-making process from the traditional single protection response to a multi-scheme comparison and selection process based on the evolution trend of the operating state. Thus, under the premise of ensuring operating safety, it avoids unnecessary and drastic disturbances to the operating state caused by control actions, reduces the possibility of operating risks caused by improper control strategies, and improves the overall operating safety and stability of the permanent magnet motor centrifugal pump under complex operating conditions. Attached Figure Description

[0019] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a schematic diagram of a safe operation control method for a permanent magnet motor centrifugal pump according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0021] Reference Figure 1 The following examples were obtained: Example 1: A method for safe operation control of a permanent magnet motor centrifugal pump, comprising the following steps: During the operation of the permanent magnet motor centrifugal pump, operating status parameters of both the permanent magnet motor and centrifugal pump are simultaneously collected within the operating range covering startup, steady-state operation, and changes in operating conditions. These parameters are then aligned to a unified time reference and processed to suppress transient interference, forming a continuous sequence of operating status parameters. By simultaneously acquiring operating status information from both the permanent magnet motor and centrifugal pump sides throughout the entire operation phase and performing time alignment and interference suppression on data from different sources, the operating status parameters are made comparable and continuous on the same time axis. This avoids misjudgments caused by asynchronous sampling or transient noise, providing a real, stable, and traceable data foundation for subsequent operating status analysis and control decisions.

[0022] Based on the operating state sequence, the rate of change characteristics of the operating state parameters on the permanent magnet motor side and the centrifugal pump side over time, the energy transfer characteristics reflecting the relationship between power input and load response, and the phase consistency characteristics between the operating state parameters on the permanent magnet motor side and the centrifugal pump side are calculated. A dynamic feature vector reflecting the coupling relationship between the permanent magnet motor output, the centrifugal pump load, and the fluid operating state is constructed. By extracting features from the operating state sequence from three dimensions—time variation, energy relationship, and parameter coordination—the original multidimensional operating data is transformed into a dynamic feature vector that can characterize the dynamic behavior of the system. This allows the coupling relationship between the permanent magnet motor output characteristics, the centrifugal pump load characteristics, and the fluid operating state to be centrally reflected, thus providing a unified and structured representation basis for identifying abnormal development trends.

[0023] The operation status of a permanent magnet motor centrifugal pump is analyzed based on dynamic feature vectors to determine whether the operation status is in an abnormal development stage. When an abnormal development trend is determined, the influence of preset control actions on the abnormal development trend under the current operation status is combined to form a judgment result reflecting the severity of the abnormality and the degree of influence of the control actions. By analyzing the changes in dynamic feature vectors, it is not only determined whether the operation status deviates from the safe operation status, but also further distinguishes whether the abnormality is in a continuous evolution stage. At the same time, the influence direction and magnitude of different control actions on the abnormal development are evaluated, thereby linking the operation status assessment with control decisions and avoiding the introduction of new risks by directly taking control measures based on a single abnormality judgment.

[0024] Based on the judgment results, multiple control action sequences are constructed, and the changing trends of the permanent magnet motor centrifugal pump's operating state after the execution of each control action sequence are predicted and analyzed to determine the operating range corresponding to each control action sequence. The control action sequence that meets the operational safety requirements is then selected and executed. By constructing multiple possible control action sequences for the same abnormal state and predicting the changing trends of the operating state after the execution of each control action sequence, the control decision-making process shifts from a single response to a multi-solution comparison and selection. Under the premise of ensuring that the operating state does not enter the dangerous operating range, the control action sequence with the most reasonable impact on system operation is selected for execution, thereby reducing the risk of secondary failures induced by the control strategy and improving the overall safety and reliability of the permanent magnet motor centrifugal pump.

[0025] During the operation of the permanent magnet motor centrifugal pump, within the operating range covering the start-up, steady-state operation, and operating condition change stages, the operating status parameters of both the permanent magnet motor and the centrifugal pump are collected synchronously. Specifically, by using the instruction manual or factory settings, the operating range is divided into three categories: start-up, steady-state operation, and operating condition change. For example, the start-up stage is defined as the speed increasing from 0 to 80% of the rated speed; the steady-state operation stage is defined as the speed fluctuation being less than ±1% of the rated speed and the outlet pressure fluctuation being less than ±2% of the rated pressure for at least 30 seconds; and the operating condition change stage is triggered when the speed setpoint changes by more than ±5% of the rated speed or the outlet pressure changes by more than ±5% of the rated pressure. Based on this stage division, the same sampling period for parallel acquisition is set, for example, 5 milliseconds, to ensure that the rapid increase in speed and current can be captured during the start-up stage, small drifts can be captured during the steady-state operation stage, and step changes in pressure and flow can be captured during the operating condition change stage. The operating status parameters of both the permanent magnet motor and the centrifugal pump are obtained synchronously within each sampling period.

[0026] During parameter acquisition, the permanent magnet motor side operating status parameters collected should at least include current, voltage, speed, electromagnetic torque, and winding temperature; the centrifugal pump side operating status parameters collected should at least include outlet pressure, inlet pressure, flow rate, vibration, and pump body temperature. Specifically, a parameter set and dimension verification rules should be established to ensure that the data in each sampling period meets the availability requirements. Among them, the permanent magnet motor side operating status parameters should at least include current (e.g., 0–120 amperes, resolution 0.1 amperes), voltage (e.g., 0–600 volts, resolution 0.1 volts), speed (e.g., 0–6000 rpm, resolution 1 rpm), electromagnetic torque (e.g., 0–80 Nm, resolution 0.1 Nm), and winding temperature (e.g., -20–180 degrees Celsius, resolution 0. 1 degree Celsius); the operating status parameters of the centrifugal pump side should at least include outlet pressure (e.g., 0–2.5 MPa range, resolution 0.001 MPa), inlet pressure (e.g., -0.1–1.0 MPa range, resolution 0.001 MPa), flow rate (e.g., 0–200 cubic meters per hour range, resolution 0.1 cubic meters per hour), vibration (e.g., 0–50 millimeters per second range, resolution 0.1 millimeters per second), and pump body temperature (e.g., -20–150 degrees Celsius range, resolution 0.1 degrees Celsius). In addition, the parameters can be expanded to include bus voltage, power factor, insulation resistance related quantities, bearing temperature, sealing cavity temperature, outlet temperature, inlet temperature, valve opening, liquid level, and pipeline differential pressure, so that the operating status parameters cover electromagnetic, thermal, structural, and fluid multidimensional changes.

[0027] The operating status parameters of the permanent magnet motor and the centrifugal pump are collected in parallel with the same sampling period, and the collection time is marked. Specifically, the sampling time is used as the unified anchor point for each set of parallel data collection. The sampling of the operating status parameters of the permanent magnet motor and the centrifugal pump are triggered with the same sampling period, and a collection time mark is generated at each trigger, such as an incrementing timestamp in milliseconds: 1000, 1005, 1010, etc. During the startup phase, it can be observed that as the collection time progresses, the rotational speed gradually increases from 0, 300, 800, to 1500 revolutions per minute. As the current increases, it changes from 0, 18, 42, and 65 amperes, and the outlet pressure gradually builds up from 0.02, 0.08, 0.18, and 0.35 MPa. During the change in operating conditions, it can be observed that valve action causes the outlet pressure to jump from 0.85 MPa to 1.05 MPa and the flow rate to decrease from 120 cubic meters per hour to 95 cubic meters per hour. Moreover, these changes correspond one-to-one with the change in electromagnetic torque from 32 N·m to 45 N·m at the same acquisition time, thus ensuring that there will be no hidden danger of "misalignment comparison" when the motor side and pump side are correlated and analyzed in the future.

[0028] The parallel acquisition results are merged into a single recording frame according to the acquisition time, serving as the basic unit of the operating state sequence. Specifically, the operating state parameters of the permanent magnet motor side and the centrifugal pump side at the same acquisition time are aggregated into a single recording frame and arranged in ascending order of acquisition time to form an operating state sequence; for example, the recording frame corresponding to acquisition time 1000 includes current 0 Amperes, voltage 0 V, speed 0 rpm, electromagnetic torque 0 N·m, winding temperature 28.6 degrees Celsius, outlet pressure 0.02 MPa, inlet pressure 0.10 MPa, and flow... The flow rate was 0 cubic meters per hour, the vibration was 0.3 millimeters per second, and the pump body temperature was 27.9 degrees Celsius. The record frame corresponding to the acquisition time 1010 contained the following data: current 42 amps, voltage 320 volts, speed 800 rpm, electromagnetic torque 18 N·m, winding temperature 29.1 degrees Celsius, outlet pressure 0.18 MPa, inlet pressure 0.09 MPa, flow rate 35 cubic meters per hour, vibration 0.9 millimeters per second, and pump body temperature 28.4 degrees Celsius. Through the unified structure of the record frames, the operating status sequence maintained the same fields and the same order in each stage.

[0029] The unified time reference alignment and transient interference suppression processing includes: establishing a unified time reference and mapping the acquisition times of the permanent magnet motor-side acquisition channel and the centrifugal pump-side acquisition channel to a unified time axis. Specifically, during the operation of the permanent magnet motor and centrifugal pump, a zero point and timing unit of the unified time axis are selected, and all acquisition times generated by the permanent magnet motor-side acquisition channel and the centrifugal pump-side acquisition channel are converted into timestamp sequences under the unified time axis. For example, if the unified time axis is in milliseconds, and the start-up instant is set to 0 milliseconds, the permanent magnet motor-side acquisition channel generates acquisition times at 0, 5, 10, and 15 milliseconds, and the centrifugal pump-side acquisition channel generates acquisition times at 1, 6, 11, and 16 milliseconds, respectively, which are mapped to 0, 5, 10, and 15 milliseconds and 1, 6, 11, and 16 milliseconds under the unified time axis. This forms a set of acquisition times that can be compared on the same time axis, providing a unified reference for subsequent delay compensation and resampling, and avoiding feature calculation mismatches caused by inconsistent time bases between different acquisition channels.

[0030] Delay compensation is performed based on the fixed channel delay, and data at different sampling frequencies are resampled to ensure that each recorded frame has a consistent time interval. Specifically, the fixed channel delays of the permanent magnet motor-side acquisition channel and the centrifugal pump-side acquisition channel are first obtained through calibration. The acquisition time of the centrifugal pump-side acquisition channel is then compensated by overall translation based on the fixed delay, so that the compensated acquisition time is aligned with the permanent magnet motor-side acquisition channel on the same time axis. For example, if the fixed delay of the permanent magnet motor-side acquisition channel is 2 milliseconds and the fixed delay of the centrifugal pump-side acquisition channel is 7 milliseconds, then the centrifugal pump-side acquisition channel has a fixed delay of 5 milliseconds relative to the permanent magnet motor-side acquisition channel. The delay compensation is completed by subtracting 5 milliseconds from the acquisition time of the centrifugal pump-side acquisition channel. Subsequently, resampling is performed on data with different sampling frequencies. For example, the sampling period of the permanent magnet motor side acquisition channel is 5 milliseconds, and the sampling period of the centrifugal pump side acquisition channel is 10 milliseconds. Then, 5 milliseconds is used as a uniform time interval, and target sampling points are constructed at 0, 5, 10, 15, and 20 milliseconds. The 5 and 15 millisecond sampling points missing in the centrifugal pump side acquisition channel under the 10 millisecond period are generated by interpolation to generate corresponding data, so that each recorded frame after resampling is completely covered at a uniform time interval, laying a consistent frame structure for subsequent peak identification and sample loss recovery.

[0031] Abrupt peaks and sample loss discontinuities are identified. Slipping median suppression is applied to peaks, and time-axis interpolation is used to restore continuity for sample loss. Specifically, for each type of permanent magnet motor-side operating state parameter and centrifugal pump-side operating state parameter, adjacent differences are calculated on consecutive recording frames. If a single difference exceeds the maximum allowable jump variable obtained statistically during the steady-state operation phase, it is marked as an abrupt peak. For example, the maximum allowable jump variable for adjacent differences in outlet pressure during steady-state operation is 0.03 MPa. If the outlet pressure jumps from 0.80 MPa to 1.10 MPa in a certain recording frame, it is determined to be an abrupt peak. Slipping median suppression is applied to abrupt peaks. For example, a 5-frame window centered on the current frame is selected, and the peak value of the parameter is replaced with the median of the five values ​​within the window, thereby suppressing the amplified effect of single-point anomalies on the rate of change and energy transfer characteristics. The identification of sample loss discontinuities is based on a uniform time interval. If a gap appears in the continuous timestamps, for example, the target sampling point sequence contains 0, 5, 10, 15, and 20 milliseconds, but the actual recorded frame is missing 10 milliseconds, it is determined to be a sample loss discontinuity. Time axis interpolation is used to generate recorded frame data at 10 milliseconds. Linear interpolation is used for parameters with slow changes, or other methods in the prior art can be used. This is not the core inventive content of this invention. For parameters with rapid changes, an interpolation method that retains the previous frame and limits the maximum amount of change is used, thereby restoring continuity while avoiding the introduction of unreasonable transitions.

[0032] Each recorded frame, after processing with delay compensation, resampling, sliding median suppression, and time axis interpolation, is sorted and output in ascending order according to the timestamp of a unified time axis, forming a continuous operating state sequence. For example, the output operating state sequence covers 0 to 2000 milliseconds, with a recorded frame formed every 5 milliseconds, for a total of 401 frames. Each frame contains both the operating state parameters of the permanent magnet motor and the centrifugal pump. The field order of the recorded frames remains fixed, thus ensuring that when calculating the rate of change characteristics, energy transfer characteristics, and phase consistency characteristics based on the operating state sequence, features can be stably extracted within a fixed time window and the dynamic changes of data from different sources can be accurately aligned. This reduces the interference of transient noise and sampling misalignment on the judgment of abnormal development stages, and improves the availability and reliability of the permanent magnet motor centrifugal pump safe operation control method during startup and operating condition change stages.

[0033] Based on the operating state sequence, the rate of change characteristics of the operating state parameters on the permanent magnet motor side and the centrifugal pump side over time, as well as the energy transfer characteristics reflecting the relationship between power input and load response, are calculated. During the operation of the permanent magnet motor and centrifugal pump, a fixed time window length and sliding step size are selected, for example, a time window length of 200 milliseconds and a sliding step size of 50 milliseconds. Within each time window, the corresponding operating state sequence segment is extracted. Subsequently, the operating state parameters on the permanent magnet motor side and the centrifugal pump side are split into multiple single-parameter time series according to parameter name, ensuring that each single-parameter time series has a consistent sampling time arrangement within the same time window. For example, with a 5-millisecond sampling period, a 200-millisecond time window contains 41 sampling points. The current single-parameter time series can be represented as 10.2, 10.6, 11.0… Amperes, and the outlet pressure single-parameter time series can be represented as 0.62, 0.63, 0.65… MPa. This preparation process ensures that the subsequent rate of change characteristic values ​​and energy transfer characteristic values ​​can be aligned, calculated, and compared within the same time window.

[0034] The rate of change characteristic refers to: splitting the operating state parameters of the permanent magnet motor side and the centrifugal pump side into multiple single-parameter time series according to parameter name, calculating the parameter value difference for each single-parameter time series at adjacent sampling times, and dividing the parameter value difference by the sampling interval value to obtain the corresponding rate of change characteristic value. Specifically, within each time window, the parameter value difference at adjacent sampling times is calculated for each single-parameter time series, and normalized by the sampling interval value; for example, if the sampling interval value is 0.005 seconds, and the rotational speed increases from 2900 rpm to 2920 rpm at adjacent sampling times, then the parameter value difference is 20 rpm, and the rate of change characteristic value is 20 divided by 0.005, which gives 4000 rpm; if the outlet pressure decreases from 0.80 MPa to 0.78 MPa, then the parameter value difference is -0.02 MPa, and the rate of change characteristic value is -0.02 divided by 0.005, which gives -4 MPa per second. By applying this calculation method to both the permanent magnet motor side operating status parameters and the centrifugal pump side operating status parameters, the rate of change characteristic value can directly reflect the dynamic change speed of the operating status during startup, steady-state operation and operating condition change stages, and provide a changing background for the subsequent interpretation of energy transfer characteristics.

[0035] The energy transfer characteristics refer to: Calculating the input power value by selecting voltage and current values ​​from the permanent magnet motor's operating parameters, where the input power value is the product of the voltage and current values; and calculating the hydraulic power value by selecting outlet pressure, inlet pressure, and flow rate from the centrifugal pump's operating parameters, where the pressure difference is the difference between the outlet and inlet pressure values, and the hydraulic power value is the product of the pressure difference and flow rate. Specifically, at each sampling moment, the input power value and hydraulic power value are calculated separately to form a corresponding power time series. For example, at a certain sampling moment, if the voltage value is 380 volts and the current value is 25 amperes, then the input power value is 380 multiplied by 25, resulting in 9500. At the same sampling moment, if the outlet pressure value is 0.90 MPa, the inlet pressure value is 0.10 MPa, the pressure difference is 0.80 MPa, and the flow rate is 110 cubic meters per hour, then the hydraulic power value is 0.80 multiplied by 110, resulting in 88. Furthermore, during the changing operating conditions, it can be observed that an increase in current leads to an increase in the input power value, while valve action causes a decrease in flow rate, resulting in a decrease in the hydraulic power value. This synchronous or divergent relationship between the input power value and the hydraulic power value can be used to characterize whether the power input and load response are matched, providing energy-level evidence for judging the abnormal development stage.

[0036] The energy transfer characteristic values ​​are calculated based on the input power and hydraulic power values, including: the energy ratio characteristic value is the sum of the hydraulic power value and the input power value plus the zero-prevention constant; the energy difference characteristic value is the sum of the input power value and the hydraulic power value; and the zero-prevention constant is a preset minimum positive number to ensure that the denominator of the division is not zero. Specifically, the energy ratio characteristic value and the energy difference characteristic value are calculated at each sampling time to form a sequence of energy transfer characteristic values. For example, if the zero-prevention constant is 1, the input power value is 9500, and the hydraulic power value is 88, then the energy ratio characteristic value is 88 divided by 9501, which gives approximately 0.00926, and the energy difference characteristic value is 9500 minus 88, which gives 9412. When the input power value is close to 0 for a short period during the start-up phase, the zero-prevention constant ensures that the energy ratio characteristic value can still be calculated and that the denominator does not become zero, thereby maintaining the continuity of the energy transfer characteristic values ​​during the start-up and operating condition change phases. By outputting the rate of change characteristic value and the energy transfer characteristic value in parallel within the same time window, the operating status of the permanent magnet motor centrifugal pump can be described from the perspectives of both "rate of change" and "energy matching". This provides a sufficient and reproducible numerical basis for the subsequent construction of dynamic feature vectors and the identification of abnormal development stages.

[0037] To demonstrate the phase consistency between the operating state parameters of the permanent magnet motor and the centrifugal pump, a parameter pair set needs to be pre-established. This set includes at least one pair of operating state parameters from both the permanent magnet motor and centrifugal pump sides. Specifically, the parameter pair set is constructed and a time window is set. During the operation of the permanent magnet motor and centrifugal pump, a fixed time window is selected based on the operating state sequence. For example, the time window length is 200 milliseconds and the sliding step size is 50 milliseconds. Within each time window, parameter pairs with coupling significance are selected from the operating state parameters of both the permanent magnet motor and centrifugal pump sides to form the parameter pair set. For example, "electromagnetic torque-outlet pressure" is selected as the first parameter pair, "current-flow rate" as the second, and "speed-vibration" as the third, ensuring that the parameter pair set includes at least one pair of operating state parameters from both the permanent magnet motor and centrifugal pump sides. With this construction method, the subsequent calculation of normalized cross-correlation values ​​can be performed within the same time window for each pair of parameters, thereby avoiding direct mixing caused by different dimensions and dynamic responses between different parameters, and improving the sensitivity of phase consistency characteristics to changes in coupling relationship.

[0038] For each pair of parameters, normalized cross-correlation values ​​are calculated one by one for different time offsets within a fixed time window. The normalized cross-correlation value is the sum of the point-by-point products of the two sequences at that time offset, divided by the square root of the sum of the squares of the amplitudes of the two sequences. Specifically, two sequences are generated for each pair of parameters, and the normalized cross-correlation value is calculated one by one within the time offset range. For example, if the time offset range is -50 ms to +50 ms, with a step of 5 ms, the electromagnetic torque sequence has 41 sampling points within a 200 ms window, and the outlet pressure sequence also has 41 sampling points. When the time offset is +10 ms, the outlet pressure sequence is shifted back by 10 ms and multiplied point-by-point with the electromagnetic torque sequence to obtain the sum of the point-by-point products. At the same time, the sum of the squares of the amplitudes of the electromagnetic torque sequence and the outlet pressure sequence are calculated and their square roots are taken respectively. Then, the normalized cross-correlation value at that time offset is obtained by dividing the sum of the point-by-point products by the square root of the sum of the squares of the amplitudes of the two sequences. By calculating each time offset individually, a set of normalized cross-correlation curves that vary with time offset can be obtained, providing a clear basis for the selection of subsequent time offset characteristic values.

[0039] For each parameter pair, the time offset value corresponding to the maximum normalized cross-correlation value is selected as the time offset characteristic value for that parameter pair. Specifically, for each parameter pair, the maximum value of the normalized cross-correlation value is found within its time offset range, and the time offset value corresponding to the maximum value is recorded as the time offset characteristic value. For example, for the electromagnetic torque-outlet pressure parameter pair, the maximum value of the normalized cross-correlation value in the range of -50 ms to +50 ms appears at +15 ms, so the time offset characteristic value is taken as +15 ms, indicating that the outlet pressure has a 15 ms response lag relative to the electromagnetic torque; for the current-flow parameter pair, the maximum value of the normalized cross-correlation value appears at -5 ms, so the time offset characteristic value is taken as -5 ms, indicating that the flow rate change has a 5 ms advance or a 5 ms lead relationship after sampling alignment relative to the current change. This time offset characteristic value can quantify the dynamic response timing relationship between the permanent magnet motor side operating state parameters and the centrifugal pump side operating state parameters within the same time window, laying the numerical foundation for the conversion of phase consistency characteristic values.

[0040] When constructing the dynamic feature vector, the main frequency value of the permanent magnet motor side parameter corresponding to the time offset feature value is calculated within the same time window. The main frequency value is the reciprocal of the interval between adjacent zero crossings. The time offset feature value is converted into a phase consistency feature value, which is the main frequency value multiplied by the time offset feature value and then multiplied by 360. The rate of change feature value, energy transfer feature value, and phase consistency feature value are combined in a predetermined order and subjected to amplitude normalization to form a dynamic feature vector. Specifically, within the same time window, zero crossings are detected for the permanent magnet motor side parameter corresponding to the time offset feature value. For example, the interval between adjacent zero crossings is counted for the electromagnetic torque sequence. If the interval between adjacent zero crossings is 20 milliseconds, the main frequency value is the reciprocal of 0.02 seconds, which is 50 Hz. When the time offset feature value is +15 milliseconds, the main frequency value of 50 Hz is multiplied by 0.015 seconds and then multiplied by 360 to obtain a phase consistency feature value of 270 degrees. Subsequently, the phase consistency characteristic value is combined with the rate of change characteristic value and energy transfer characteristic value calculated within the same time window in a predetermined order, and the values ​​of different dimensions are normalized, for example, by dividing each type of characteristic value by the maximum allowable amplitude or statistical standard deviation of the corresponding characteristic in the steady-state operation phase, so that the final dynamic characteristic vector remains comparable under different operating conditions. In this way, the dynamic characteristic vector simultaneously carries information on the rate of change, energy matching, and response phase, providing a publicly available and reproducible numerical input for subsequent analysis of the operating status of the permanent magnet motor centrifugal pump based on the dynamic characteristic vector, judging the abnormal development stage, and forming a judgment result.

[0041] The judgment result based on dynamic feature vectors includes: subtracting the components of the dynamic feature vector from the safety benchmark vector and taking the absolute value, summing them according to preset weights to obtain the risk index, calculating the risk index difference between two adjacent time windows, and comparing the risk index and the risk index difference with the corresponding thresholds to form a sequence of risk index changes over time within a continuous time window; for example, if the risk index is 0.42, 0.48, 0.56, 0.65, and 0.78 in five consecutive time windows, the risk index difference between two adjacent time windows is 0.06, 0.08, 0.09, and 0.13 respectively. When the risk index exceeds the preset risk threshold and the risk index difference exceeds the growth threshold, the operating state is judged to have entered the abnormal development stage from a general deviation. The severity of the abnormality is calculated by multiplying the risk index and the risk index difference according to the severity of the abnormality, so that the severity of the abnormality reflects both the deviation magnitude and the development speed, thereby distinguishing between slowly evolving abnormalities and rapidly deteriorating abnormalities, and providing a classification basis for subsequent control action assessment.

[0042] For each preset control action, a short-term trial with a fixed amplitude is performed while keeping other control variables unchanged. The risk index after the trial is recalculated. The degree of influence of the control action is taken as the difference between the risk indices before and after the trial. Specifically, an impact assessment is performed for each preset control action. For example, for three types of preset control actions: reducing the speed setpoint, tightening the torque limit, and reducing the current limit, a short-term trial with a fixed amplitude is applied without changing other control variables. After the trial period ends, the dynamic feature vector of the corresponding time window is re-acquired and the risk index is calculated. The difference between the risk index before and after the trial is taken as the degree of influence of the control action. When the difference is negative and the absolute value is large, it indicates that the control action can effectively suppress the abnormal development trend. When the difference is positive or close to zero, it indicates that the control action is not effective in suppressing the abnormal development trend and may even exacerbate the risk. This method avoids directly selecting control actions based solely on experience or static rules.

[0043] The judgment result is output by combining the severity of the anomaly with the impact of the control action. Specifically, the severity value of the anomaly and the impact value of the control action for each control action are combined to form a judgment result item. For example, when the anomaly severity is 0.073, the impact of the control action corresponding to the speed setpoint reduction is -0.18, the impact of the control action corresponding to the torque limit tightening is -0.09, and the impact of the control action corresponding to the current limit reduction is -0.04. The judgment result can clearly reflect the difference in the ability of different control actions to suppress the current stage of anomaly development. Based on this judgment result, subsequent steps can prioritize the sequence of control actions that have a significant effect on suppressing the anomaly development trend and have a clear direction of risk evolution, while ensuring operational safety. This effectively connects the operational status analysis with control decision-making, reduces the probability of secondary failures induced by control strategies, and improves the safe operation level of permanent magnet motor centrifugal pumps under complex operating conditions.

[0044] The determination of the risk threshold and growth threshold is based on a comprehensive characterization of the long-term statistical characteristics and dynamic evolution of the safe operating state of the permanent magnet motor centrifugal pump. The risk threshold characterizes the maximum permissible deviation of the operating state from the safety baseline vector, while the growth threshold characterizes the maximum permissible rate of change of this deviation over time. Specifically, the risk threshold is obtained through statistical analysis of the dynamic characteristic vectors collected during the normal startup, steady-state operation, and controllable operating condition changes of the permanent magnet motor centrifugal pump, ensuring that the risk index distribution corresponding to the vast majority of safe operating states falls below the risk threshold. The growth threshold is determined based on the statistical results of the risk index changes in adjacent time windows, ensuring that the rate of change of the risk index under normal operation and controllable disturbance conditions does not trigger anomaly development judgment. By corresponding the risk threshold and growth threshold to the static deviation upper limit and dynamic evolution upper limit of the operating state, respectively, the judgment of the abnormal development stage simultaneously satisfies both the "exceeding the limit of deviation" and the "abnormal development rate," thereby avoiding misjudgments caused by transient disturbances or slow drifts and ensuring the reliability and engineering applicability of anomaly identification.

[0045] Before calculating the risk index, it is necessary to ensure that the parameter types and arrangement order in the dynamic feature vector and the safety benchmark vector are consistent. Specifically, this involves establishing a calculation entry point and alignment constraints. During the operation of the permanent magnet motor centrifugal pump, the dynamic feature vector output at a fixed time window is used as the input for the current window. At the same time, a safety benchmark vector matching the operating conditions of that time window is selected as the comparison input. The parameter types and arrangement order of the two are verified. Only when the change rate feature, energy transfer feature, and phase consistency feature are arranged in the same position in the two vectors will the weight selection process begin. For example, the dynamic feature vector is arranged in a predetermined order as "change rate feature value group - energy ratio feature value - energy difference feature value - phase consistency feature value group". The safety benchmark vector adopts the same arrangement, so that each component has a unique corresponding safety benchmark. This avoids the problem of mismatch between weight selection and deviation calculation caused by misalignment of the order from the source, and provides a reproducible structural basis for subsequent comparison and weighting according to parameter type.

[0046] For each parameter type, the parameter value in the dynamic feature vector is compared with the corresponding safety benchmark in the safety benchmark vector. When the parameter value is higher than the safety benchmark, the first weight preset for that parameter type is selected; when the parameter value is not higher than the safety benchmark, the second weight preset for that parameter type is selected. The sum of the first and second weights is one. Specifically, the deviation direction of each parameter type is introduced into the weight selection process, so that the contribution of the same parameter type to the risk index when it is higher or lower can be reflected with different weights. For example, for the parameter type "energy difference feature value", the first weight is preset to 0.7 and the second weight to 0.3. When the energy difference feature value is higher than the corresponding safety benchmark, 0.7 is selected, and when the energy difference feature value is not higher than the corresponding safety benchmark, 0.3 is selected. For the parameter type "phase consistency feature value", the first weight is preset to 0.6 and the second weight to 0.4. When the phase consistency feature value is higher than the corresponding safety benchmark, 0.6 is selected, and when the phase consistency feature value is not higher than the corresponding safety benchmark, 0.4 is selected. By using this method of selecting weights based on the direction of deviation, the risk index has a stronger ability to distinguish different abnormal development trends, avoiding the problem of insensitivity to the direction of deviation caused by using only fixed weights.

[0047] After all parameter types have had their weights selected, the sum of all selected weights is used as the normalization benchmark to normalize each weight. Specifically, the weights selected for different parameter types in the current window are uniformly mapped to a directly comparable set of weights, ensuring that the sum of the weight sets meets the normalization requirements. For example, if five parameter types have their weights selected in the current time window, namely 0.7, 0.6, 0.5, 0.4, and 0.3, the sum of all selected weights is 2.5. Dividing each weight by 2.5 yields normalized weights of 0.28, 0.24, 0.20, 0.16, and 0.12, ensuring that the risk index under different windows and operating conditions has consistent dimensions and comparability during subsequent weighted summation. Furthermore, during the startup phase, some parameter types may not have valid values. In this case, only the parameter types with valid values ​​are selected, and their sum is used as the normalization benchmark, thus ensuring that the continuous output of the risk index is not interrupted due to the absence of individual parameters.

[0048] The deviations of each parameter are weighted and summed based on normalized weights. Specifically, the difference between the parameter value of each parameter type in the dynamic feature vector and the corresponding safety benchmark in the safety benchmark vector is converted into a parameter deviation. Then, the deviations are weighted and summed according to the normalized weights to obtain the risk index. For example, if the parameter deviations of five parameter types in a certain time window are 0.10, 0.06, 0.20, 0.05, and 0.08, and the corresponding normalized weights are 0.28, 0.24, 0.20, 0.16, and 0.12, the weighted sum is 0.10 multiplied by 0.28 plus 0.06 multiplied by 0.24 plus 0.20 multiplied by 0.20 plus 0.05 multiplied by 0.16 plus 0.08 multiplied by 0.12, resulting in 0.028 plus 0.0144 plus 0.040 plus 0.008 plus 0.0096, totaling 0.100, which forms the risk index for that time window. Through a continuous process of "consistent parameter type order - selection of direction trigger weight - weight normalization - weighted summation of deviations", the risk index can reflect the difference in deviation direction while maintaining numerical reproducibility, so that the judgment of subsequent abnormal development stages and the impact assessment of preset control actions have a stable and interpretable quantitative basis.

[0049] Determining the operating range corresponding to each control action sequence refers to: after the permanent magnet motor centrifugal pump's operating state is determined to have entered an abnormal development stage, using the severity of the abnormality in the determination result as the basis for selecting the action amplitude level, and using the degree of influence of the control action as the basis for selecting the action priority and combination order; for example, if the abnormality severity is 0.10 and the control action influence level shows a stronger suppression effect on the speed setpoint reduction, then the discrete adjustment amplitude for the speed setpoint is set to three levels: -3%, -6%, and -9%, and the duration is set to three levels: 1 second, 3 seconds, and 5 seconds; the discrete adjustment amplitude for the torque limit value is set to two levels: -5% and -10%, and the duration is set to two levels: 2 seconds and 4 seconds; and the discrete adjustment amplitude for the current limit value is set to three levels: -5% and -10%, and the duration is set to two levels: 2 seconds and 4 seconds. The adjustment range is set to -5% and -8%, and the duration is set to 2 seconds and 4 seconds. Trigger threshold and delay duration are set for shutdown trigger conditions. Then, the various levels are combined into multiple control action sequences according to a predetermined execution order. For example, the control action sequence is "speed setpoint reduced by -6% and held for 3 seconds → torque limit value reduced by -5% and held for 2 seconds → current limit value reduced by -5% and held for 2 seconds", and the control action sequence is "torque limit value reduced by -10% and held for 2 seconds → speed setpoint reduced by -3% and held for 3 seconds → shutdown trigger condition enters high sensitivity mode". Different action combinations cover different paths from mild intervention to strong protection, providing a complete candidate set for subsequent prediction of operating state sequences.

[0050] Using the dynamic feature vector corresponding to the end of the current operating state sequence as the initial state, the operating state parameters of the permanent magnet motor and the centrifugal pump are recursively calculated piecewise using the rate of change feature in the dynamic feature vector. This yields the predicted operating state sequence for each control action sequence at the corresponding execution stage. Specifically, the dynamic feature vector at the end of the current operating state sequence is used as the starting point. The rate of change feature is considered as the basis for the direction and speed of parameter change within a short prediction interval, and recursion is performed according to the piecewise structure of the control action sequence. For example, at the current moment, the current is 40 amps, the voltage is 380 volts, the speed is 2900 rpm, the electromagnetic torque is 28 N·m, the winding temperature is 92 degrees Celsius, the outlet pressure is 0.85 MPa, the inlet pressure is 0.12 MPa, and the flow rate is... With a flow rate of 105 cubic meters per hour, a vibration of 3.2 millimeters per second, and a pump body temperature of 78 degrees Celsius, and considering the rate of change characteristics showing a decreasing trend in rotational speed per second, an increasing trend in outlet pressure per second, and a decreasing trend in flow rate per second, when the first segment of the control action sequence is a -6% reduction in the rotational speed setpoint held for 3 seconds, the operating state parameters of the permanent magnet motor and the centrifugal pump are recursively calculated point by point within the 3-second prediction interval using sampling intervals to obtain the first segment of the predicted operating state sequence. Subsequently, when the second segment, a torque limit value, is reduced and held for 2 seconds, the predicted value at the end of the first segment is used as the starting point of the second segment to continue the recursion until the entire control action sequence is predicted. This ensures that the segmented structure of the predicted operating state sequence and the control action sequence corresponds one-to-one, facilitating subsequent segment-by-segment calculation of the risk index and division of the operating interval.

[0051] Based on the predicted operating state sequence, a risk index is calculated segment by segment, and risk level intervals are divided according to the risk index threshold. The time period during which the risk index remains continuously within the same risk level interval is defined as the corresponding operating interval. Specifically, within each execution stage of each control action sequence, the risk index is calculated for the predicted operating state sequence using a fixed time window, and the risk index is compared with the risk index threshold to determine its corresponding risk level interval. For example, if the risk level interval is divided into three levels, a risk index less than 0.40 is the first risk level interval, a risk index between 0.40 and 0.70 is the second risk level interval, and a risk index greater than 0.70 is the third risk level interval. If the risk index of the control action sequence decreases from 0.78 to 0.62 within the first 3 seconds and remains in the second risk level range, and further decreases to 0.38 within the second 2 seconds and remains in the first risk level range, and stabilizes at 0.35 within the third 2 seconds, then the period during which the risk index in the first segment remains continuously in the second risk level range is defined as one operating interval of the control action sequence. The period during which the risk index in the second and third segments remains continuously in the first risk level range is defined as another operating interval. This ensures that each operating interval reflects the sustainable range of risk level after the execution of the control action sequence and provides a clear time period for boundary verification.

[0052] The boundary of the operating interval is checked using the energy transfer characteristics and phase consistency characteristics corresponding to the predicted operating state sequence. When the energy transfer characteristics or phase consistency characteristics exceed the preset safety range, the corresponding time period is removed from the operating interval, and the effective operating interval of each control action sequence is obtained. Specifically, within each operating interval, the energy transfer characteristics and phase consistency characteristics are checked simultaneously to see if they fall within the preset safety range, so as to avoid hidden risks caused by relying solely on the risk index. For example, the preset safety range stipulates that the energy ratio characteristic value should not be less than 0.005 and not greater than 0.200, the energy difference characteristic value should not exceed 12000, and the phase consistency characteristic value should not exceed 300 degrees. When the energy ratio characteristic value drops to 0.003 or the phase consistency characteristic value rises to 330 degrees in the first risk level operating interval of a certain control action sequence, the corresponding time period is removed from the operating interval, and the remaining time period is taken as the effective operating interval. Through this continuous process of "forming the operating range from the risk level range to verifying and eliminating energy transfer characteristics and phase consistency characteristics," the effective operating range not only meets the safety requirements at the risk index level, but also meets the requirements of matching power input with load response and stable response phase. This provides a more reliable range basis for selecting and executing control action sequences that meet the operating safety requirements, reducing the probability that the control action sequence will lead the permanent magnet motor centrifugal pump into the cavitation sensitive area, resonance risk area, or insufficient cooling capacity area.

[0053] The selection and execution of control action sequences that meet operational safety requirements includes: screening candidate control action sequences within the effective operating range. Screening criteria include that the risk index corresponding to the predicted operating state sequence does not enter the preset prohibited operating zone, and the risk index difference is decreasing. Specifically, the first round of screening of candidate control action sequences is conducted. After obtaining the effective operating range of each control action sequence in the previous stage, the predicted operating state sequence corresponding to each control action sequence within the effective operating range is extracted segment by segment, and the corresponding risk index and risk index difference are read segment by segment. For example, the preset prohibited operating zone is defined as a risk index greater than 0.80. If the risk index of a certain control action sequence increases from 0.76 to 0.82 within the effective operating range, it is determined that the risk index has entered the preset prohibited operating zone and the control action sequence is removed. If the risk index of another control action sequence decreases segment by segment from 0.74, 0.68, and 0.61 within the effective operating range, with risk index differences of -0.06 and -0.07 respectively, then the risk index difference is decreasing and it is retained as a candidate control action sequence. This screening process will eliminate control action sequences that may further worsen the risk or touch the prohibited boundaries in advance, so that subsequent disturbance magnitude index calculations will only be performed on candidate control action sequences with clear safety trends, thereby reducing invalid comparisons and improving the reliability of control decisions.

[0054] For the selected candidate control action sequences, a disturbance amplitude index is calculated. The disturbance amplitude index is the normalized absolute value of the changes in the speed setpoint, torque limit, and current limit. The weighting weights are determined by the combination of parameter weights corresponding to speed, electromagnetic torque, and current in the risk index calculation, and are normalized and then weighted and summed. Specifically, the adjustment range of the candidate control action sequence to the speed setpoint, torque limit, and current limit at each execution stage is converted into a comparable dimensionless disturbance. For example, if the speed setpoint change is -6%, the torque limit change is -10%, and the current limit change is -5%, and the maximum allowable change ranges are 10%, 20%, and 10% respectively as the normalization benchmarks, then the normalized absolute values ​​are 0.6, 0.5, and 0.5 respectively. Further, parameter weight combinations corresponding to speed, electromagnetic torque, and current are extracted from the risk index calculation. For example, if the weight combination selected in the current window is 0.50, 0.30, and 0.20, after normalization by a total weight of 1.00, it remains 0.50, 0.30, and 0.20. Therefore, the disturbance amplitude index is 0.6 multiplied by 0.50 plus 0.5 multiplied by 0.30 plus 0.5 multiplied by 0.20, resulting in 0.30 plus 0.15 plus 0.10, totaling 0.55. By keeping the weighted values ​​of the disturbance amplitude index consistent with the corresponding parameter weight combinations in the risk index calculation, the evaluation of the "cost of modification" of the control action sequence and the evaluation of the "risk contribution" of the risk index are aligned in terms of parameter importance. This avoids the possibility of risk-sensitive parameters being significantly adjusted, which could trigger new instabilities.

[0055] The control action sequence with the smallest disturbance amplitude index is selected as the execution sequence. Specifically, under the premise of meeting the screening conditions in the previous step, the candidate control action sequences are optimized for minimum disturbance. For example, if the disturbance amplitude index of candidate control action sequence A is 0.55, the disturbance amplitude index of candidate control action sequence B is 0.42, and the disturbance amplitude index of candidate control action sequence C is 0.47, then candidate control action sequence B is selected as the execution sequence. The significance of this optimization method is that, under the condition that the safety trend constraint of the risk index not entering the preset prohibited operation zone and the risk index difference showing a decreasing direction has been met, the control action sequence with smaller changes to the speed setpoint, torque limit value, and current limit value is further selected. This reduces the disturbance to the original operating balance of the permanent magnet motor centrifugal pump while suppressing the abnormal development trend, and reduces the secondary risks caused by excessive control actions.

[0056] Control commands are issued in segments according to the execution order and duration of each control action in the execution sequence, so that the permanent magnet motor centrifugal pump runs segment by segment according to the execution sequence. Specifically, the execution sequence is divided into multiple consecutive execution segments, and a corresponding control command is issued at the beginning of each execution segment and maintained for the corresponding duration. For example, if the execution sequence is "speed setpoint reduced by -6% and held for 3 seconds → torque limit value reduced by -5% and held for 2 seconds → current limit value reduced by -5% and held for 2 seconds", then the control command to reduce the speed setpoint by -6% is issued first and held for 3 seconds, then the control command to reduce the torque limit value by -5% is issued and held for 2 seconds, and then the control command to reduce the current limit value by -5% is issued and held for 2 seconds, so that the permanent magnet motor centrifugal pump runs segment by segment according to the execution sequence. By using this segmented distribution method, the amplitude and duration of the control actions are strictly consistent with the control action sequence used in the aforementioned predictive analysis, thereby ensuring the consistency of the execution process and the prediction process in terms of temporal structure. This allows the selected results to stably reproduce the predicted risk reduction trend in actual operation and improves the feasibility and engineering controllability of the permanent magnet motor centrifugal pump safe operation control method during the start-up and operating condition change stages.

[0057] It should be noted that the candidate control action sequence is not a pre-set fixed setting, but rather a combination of multiple candidate sequences dynamically formed during operation based on preset control action elements according to the severity of the anomaly and the degree of influence of the control action. In this invention, the predicted operating status and risk constraints are combined for screening and optimization, ultimately ensuring the safe operation of the permanent magnet motor centrifugal pump.

[0058] The above algorithms or formulas are all dimensionless and numerical calculations, and the results are obtained by software simulation based on a large amount of collected data to obtain the most recent real-world results. The preset parameters are set by those skilled in the art according to the actual situation.

[0059] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0060] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0061] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0062] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology 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. A method for safe operation control of a permanent magnet motor centrifugal pump, characterized in that, Includes the following steps: During the operation of the permanent magnet motor centrifugal pump, within the operating range covering the start-up, steady-state operation and operating condition change stages, the operating status parameters of the permanent magnet motor side and the centrifugal pump side are collected synchronously, and unified time reference alignment and transient interference suppression processing are performed to form a continuous operating status sequence. Based on the operating state sequence, the rate of change characteristics of the operating state parameters on the permanent magnet motor side and the centrifugal pump side with time, the energy transfer characteristics reflecting the relationship between power input and load response, and the phase consistency characteristics between the operating state parameters on the permanent magnet motor side and the centrifugal pump side are calculated. A dynamic feature vector reflecting the coupling relationship between the permanent magnet motor output, the centrifugal pump load and the fluid operating state is constructed. The operating status of the permanent magnet motor centrifugal pump is analyzed based on dynamic feature vectors to determine whether the operating status is in an abnormal development stage. When an abnormal development trend is determined, the influence of the preset control action on the abnormal development trend under the current operating status is combined to form a judgment result reflecting the severity of the abnormality and the influence of the control action. Based on the judgment results, multiple control action sequences are constructed, and the changing trend of the permanent magnet motor centrifugal pump's operating status after the execution of each control action sequence is predicted and analyzed to determine the operating range corresponding to each control action sequence. Then, the control action sequence that meets the operating safety requirements is selected and executed.

2. The method for safe operation control of a permanent magnet motor centrifugal pump according to claim 1, characterized in that, When collecting parameters, the operating status parameters of the permanent magnet motor should include at least the current, voltage, speed, electromagnetic torque, and winding temperature. The operating status parameters collected from the centrifugal pump side should include at least the outlet pressure, inlet pressure, flow rate, vibration, and pump body temperature. The operating status parameters of the permanent magnet motor and the centrifugal pump are collected in parallel using the same sampling period, and the collection time is marked. The parallel acquisition results are merged into the same recording frame according to the acquisition time, which serves as the basic unit of the running state sequence.

3. The method for safe operation control of a permanent magnet motor centrifugal pump according to claim 2, characterized in that, Unified time reference alignment and transient interference suppression processing include: Establish a unified time reference and map the acquisition times of the permanent magnet motor side acquisition channel and the centrifugal pump side acquisition channel to a unified time axis; Delay compensation is performed based on the fixed channel delay, and data at different sampling frequencies are resampled to ensure that each recorded frame has a consistent time interval. Identify abrupt peaks and sample loss discontinuities, apply sliding median suppression to peaks, restore continuity for sample loss by time axis interpolation, and output an uninterrupted running state sequence according to a unified time axis.

4. The method for safe operation control of a permanent magnet motor centrifugal pump according to claim 1, characterized in that, The rate of change characteristic refers to: splitting the operating state parameters of the permanent magnet motor side and the operating state parameters of the centrifugal pump side into multiple single-parameter time series according to the parameter name, calculating the parameter value difference for each single-parameter time series at adjacent sampling times, and dividing the parameter value difference by the sampling interval value to obtain the rate of change characteristic value corresponding to the parameter; The energy transfer characteristic refers to: selecting voltage and current values ​​from the operating state parameters of the permanent magnet motor to calculate the input power value, and the input power value is the product of the voltage and current values; The hydraulic power value is calculated by selecting the outlet pressure value, inlet pressure value and flow rate value from the centrifugal pump side operating status parameters. The pressure difference value is the outlet pressure value minus the inlet pressure value, and the hydraulic power value is the product of the pressure difference value and the flow rate value. The energy transfer characteristic values ​​are calculated based on the input power value and the hydraulic power value, including: the energy ratio characteristic value is the sum of the hydraulic power value and the input power value and the zero-prevention constant; the energy difference characteristic value is the input power value minus the hydraulic power value.

5. The method for safe operation control of a permanent magnet motor centrifugal pump according to claim 4, characterized in that, Phase consistency characteristics refer to: Establish a parameter pair set, which includes at least one pair of permanent magnet motor side operating state parameters and centrifugal pump side operating state parameters. For each parameter pair, calculate the normalized cross-correlation value for different time offsets within a fixed time window. The normalized cross-correlation value is the sum of the product of the two sequences at that time offset, divided by the square root of the sum of the squares of the amplitudes of the two sequences. For each parameter pair, select the time offset value corresponding to the maximum normalized cross-correlation value as the time offset characteristic value of that parameter pair.

6. The method for safe operation control of a permanent magnet motor centrifugal pump according to claim 5, characterized in that, When constructing the dynamic feature vector, the main frequency value is calculated for the permanent magnet motor side parameter corresponding to the time offset feature value within the same time window. The main frequency value is the reciprocal of the interval between adjacent zero crossings. The time offset feature value is converted into the phase consistency feature value. The phase consistency feature value is the main frequency value multiplied by the time offset feature value and then multiplied by 360. The rate of change characteristic value, energy transfer characteristic value, and phase consistency characteristic value are combined in a predetermined order and then subjected to amplitude normalization to form a dynamic characteristic vector.

7. The method for safe operation control of a permanent magnet motor centrifugal pump according to claim 6, characterized in that, The determination results based on dynamic feature vectors include: The risk index is obtained by subtracting each component of the dynamic feature vector from the safety benchmark vector and taking the absolute value, and then summing them according to preset weights. Calculate the risk index difference between two adjacent time windows, and compare the risk index and the risk index difference with the corresponding thresholds respectively. When the risk index exceeds the risk threshold and the risk index difference exceeds the growth threshold, it is determined that the abnormal development stage has been entered. The severity of the abnormality is the product of the risk index and the risk index difference. For each preset control action, a short-term trial with a fixed amplitude is performed while keeping the other control quantities unchanged. The risk index after the trial is recalculated, and the degree of influence of the control action is taken as the difference between the risk index before and after the trial. The severity of the anomaly is combined with the degree of impact of the control action to output the judgment result.

8. The method for safe operation control of a permanent magnet motor centrifugal pump according to claim 7, characterized in that, In the process of obtaining the risk index by weighted summation according to preset weights, the parameter types and arrangement order in the dynamic feature vector and the safety benchmark vector are kept consistent. For each parameter type, the parameter value in the dynamic feature vector is compared with the corresponding safety benchmark in the safety benchmark vector. When the parameter value is higher than the safety benchmark, the first weight preset for that parameter type is selected. When the parameter value is not higher than the safety benchmark, the second weight preset for that parameter type is selected. The sum of the first weight and the second weight is one. After all parameter types have completed the weight selection, the sum of all selected weights is used as the normalization benchmark to normalize each weight, and the deviation of each parameter is weighted and summed based on the normalized weights.

9. A method for safe operation control of a permanent magnet motor centrifugal pump according to claim 8, characterized in that, Determining the operating range corresponding to each control action sequence refers to: Based on the severity of the anomaly and the degree of influence of the control action in the judgment result, corresponding discrete adjustment ranges and durations are set for the speed setpoint, torque limit value, current limit value and shutdown trigger condition, and they are combined in a predetermined execution order to form a variety of control action sequences. Using the dynamic feature vector corresponding to the end of the current operating state sequence as the initial state, the operating state parameters of the permanent magnet motor and the centrifugal pump are calculated piecewise and recursively using the rate of change feature in the dynamic feature vector to obtain the predicted operating state sequence of each control action sequence under the corresponding execution stage. The risk index is calculated segment by segment based on the predicted operating status sequence, and the risk level interval is divided according to the risk index threshold. The time period in which the risk index is continuously maintained within the same risk level interval is defined as the corresponding operating interval. The boundary of the operating range is checked by using the energy transfer characteristics and phase consistency characteristics corresponding to the predicted operating state sequence. When the energy transfer characteristics or phase consistency characteristics exceed the preset safety range, the corresponding time period is removed from the operating range to obtain the effective operating range of each control action sequence.

10. A method for safe operation control of a permanent magnet motor centrifugal pump according to claim 9, characterized in that, Select and execute a sequence of control actions that meet operational safety requirements, including: Within the effective operating range, candidate control action sequences are selected. The selection criteria include that the risk index corresponding to the predicted operating state sequence does not enter the preset prohibited operating zone, and the difference in risk index is decreasing. For the selected candidate control action sequences, the disturbance amplitude index is calculated. The disturbance amplitude index is the normalized absolute value of the change in speed setpoint, the change in torque limit value, and the change in current limit value. The weighting weights are determined by the combination of parameter weights corresponding to speed, electromagnetic torque, and current in the risk index calculation. After normalization according to the weight sum, the weighted sum is calculated, and the control action sequence with the smallest disturbance amplitude index is selected as the execution sequence. Control commands are issued in segments according to the execution order and duration of each control action in the execution sequence, so that the permanent magnet motor centrifugal pump runs segment by segment according to the execution sequence.

Images