Intelligent frequency conversion control system of primary energy efficiency three-phase asynchronous motor

Abstract

This invention relates to the field of motor drive and frequency conversion control technology, specifically to an intelligent frequency conversion control system for a first-level energy-efficient three-phase asynchronous motor. The system includes: a data acquisition module for acquiring three-phase voltage data, three-phase current data, and speed feedback data output from the frequency converter; an electromechanical feature identification module for performing harmonic analysis based on the three-phase current data and identifying mechanical inertia data, transmission clearance data, and the system's current resonant frequency data; a fatigue risk quantification module for calculating mechanical fatigue penalty integral data in conjunction with subsequent speed and voltage adjustment commands; a joint optimization control module for optimizing a comprehensive objective function including energy consumption and mechanical fatigue indicators; an active damping execution module for generating active damping compensation current data; and an output feedback module for outputting a flexible frequency conversion drive pulse width modulation signal and generating an electromechanical system health assessment report. This invention achieves a joint balance between energy consumption, mechanical fatigue, and operational stability.

Description

Technical Field

[0001] This invention relates to the field of motor drive and frequency conversion control technology, specifically to an intelligent frequency conversion control system for a first-class energy-efficient three-phase asynchronous motor. Background Technology

[0002] In industrial scenarios such as petrochemicals, water treatment, and continuous manufacturing, first-level energy-efficient three-phase asynchronous motors are usually combined with frequency converters and load-side mechanical transmission systems to form a speed-regulating drive link. For frequency conversion control of such motor systems, vector control, direct torque control, or energy-saving optimization control based on electrical parameters are generally the main methods to achieve speed regulation, reduce power consumption, and improve operating efficiency. Currently, most control methods support motor operation status adjustment based on three-phase voltage, three-phase current and speed feedback. If the operation safety of the mechanical transmission system on the load side needs to be taken into account, vibration, strain or condition monitoring devices are usually configured to independently monitor components such as couplings, gearboxes and bearings in order to obtain mechanical condition assessment results. However, when controlling the motor system in the above manner, electrical energy-saving control and mechanical health management are usually handled in isolation. This results in a situation where, although a single speed and pressure regulation action can meet the flow, pressure, or energy consumption targets, it cannot simultaneously constrain the excitation effects of electromagnetic torque pulsation on mechanical inertia, transmission clearance, and resonant frequency band, making it difficult to balance energy efficiency and mechanical lifespan. At the same time, this approach often only monitors or alarms the mechanical fatigue risk on the load side after the fact, failing to quantify mechanical stress and fatigue damage and incorporate them into a unified optimization process before control execution, resulting in insufficient feedforward assessment of mechanical risks. Furthermore, it requires reliance on additional sensors, discrete monitoring links, or manual experience to set strategies, making it difficult to utilize existing current and speed information to achieve online identification of electromechanical coupling parameters, active damping compensation, and joint early warning of remaining lifespan, resulting in low system integration and control synergy. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent frequency conversion control system for a first-class energy-efficient three-phase asynchronous motor, solving the following technical problems: This approach avoids the problem of traditional control methods that rely solely on electrical parameters for energy saving while neglecting the lifespan of mechanical equipment, thereby achieving a combined balance between energy consumption, mechanical fatigue, and operational stability.

[0004] The objective of this invention can be achieved through the following technical solutions: An intelligent frequency conversion control system for a Class I energy-efficient three-phase asynchronous motor is applied in a motor drive circuit comprising a three-phase asynchronous motor, a frequency converter, and a load-side mechanical transmission system, including: The data acquisition module is used to acquire the three-phase voltage data, three-phase current data, and speed feedback data output by the frequency converter; The electromechanical feature identification module is used to perform harmonic analysis based on the three-phase current data, extract current fluctuation features, and identify the electromechanical coupling dynamic parameters of the load-side mechanical transmission system based on the current fluctuation features. The electromechanical coupling dynamic parameters include mechanical inertia data, transmission clearance data, and the current resonant frequency data of the system. The fatigue risk quantification module is used to receive the next speed and pressure adjustment command sent by the host computer system, input the next speed and pressure adjustment command and the mechanical inertia data and transmission clearance data in the electromechanical coupling dynamics parameters into the electromagnetic torque pulsation and mechanical stress mapping model, and calculate the mechanical fatigue penalty integral data corresponding to the next speed and pressure adjustment command. The joint optimization control module is used to construct a comprehensive objective function, wherein the comprehensive objective function includes an energy consumption index calculated based on the three-phase voltage data and the three-phase current data in the current control cycle, as well as a mechanical fatigue index. Within a preset prediction time window, the comprehensive objective function is optimized and solved based on a preset energy efficiency optimization step size and the mechanical fatigue penalty integral data, and a reference torque command and a reference flux linkage command are output. The active damping execution module is used to extract the corresponding resonant oscillation signal based on the current resonant frequency data of the system, generate active damping compensation current data with the opposite phase to the resonant oscillation signal, inject the active damping compensation current data into the underlying vector control algorithm, and generate a flexible frequency conversion drive pulse width modulation signal according to the reference torque command and the reference flux command. The output feedback module is used to output the flexible frequency conversion drive pulse width modulation signal to the inverter side of the frequency converter to drive the three-phase asynchronous motor to run, and generate an electromechanical system health assessment report based on the mechanical fatigue penalty integral data.

[0005] In one possible implementation, the electromechanical feature identification module is specifically used for: Perform a fast Fourier transform on the three-phase current data to separate the fundamental current data and the high-frequency harmonic current data; Based on the high-frequency harmonic current data and the rotational speed feedback data, a nonlinear mechanical observer is constructed on the load side; The nonlinear mechanical observer can be used to calculate the mechanical inertia data, the transmission clearance data, and the current resonant frequency data of the system in real time.

[0006] In one possible implementation, the fatigue risk quantification module is specifically used for: Calculate the expected electromagnetic torque pulsation data based on the next speed and voltage regulation command; The expected electromagnetic torque pulsation data, the mechanical inertia data, and the transmission clearance data are input into the electromagnetic torque pulsation and mechanical stress mapping model to calculate the expected mechanical stress data. Rainflow counting analysis is performed on the expected mechanical stress data within a preset time period to extract the stress amplitude and cycle number, and the mechanical fatigue penalty integral data is calculated by accumulating the data based on the linear fatigue damage accumulation theory.

[0007] In one possible implementation, the joint optimization control module is specifically used for: The current energy consumption is calculated by integrating the product of the three-phase voltage data and the three-phase current data within the current control cycle. Multiply the mechanical fatigue penalty integral data by a preset fatigue weight coefficient to obtain the current mechanical fatigue quantification value; The current power consumption is added to the current mechanical fatigue quantification value to construct the comprehensive objective function; Within the preset prediction time window, the preset energy efficiency optimization step size is used to minimize the comprehensive objective function based on the model predictive control algorithm, generating the reference torque command and the reference flux command that minimize the value of the comprehensive objective function.

[0008] In one possible implementation, the system further includes a policy adjustment module, which is used to: The mechanical fatigue penalty integral data is compared with a preset fatigue risk threshold. If the mechanical fatigue penalty integral data is greater than or equal to the preset fatigue risk threshold, a health priority strategy is triggered, the preset energy efficiency optimization step size in the joint optimization control module is reduced to a preset conservative optimization step size, and the amplitude of the active damping compensation current data is multiplied by a preset compensation gain coefficient greater than 1. If the mechanical fatigue penalty integral data is less than the preset fatigue risk threshold, an energy efficiency priority strategy is triggered to maintain the preset energy efficiency optimization step size and the current amplitude of the active damping compensation current data.

[0009] In one possible implementation, the active damping execution module is specifically used for: Extract the resonant frequency band corresponding to the current resonant frequency data of the system and the resonant oscillation signal within the resonant frequency band; generate a compensation signal with the opposite phase to the resonant oscillation signal as the active damping compensation current data; The active damping compensation current data is superimposed on the torque current setpoint of the underlying vector control algorithm to cancel out the frequency components that cause mechanical resonance, forming a superimposed torque current setpoint signal.

[0010] In one possible implementation, the underlying vector control algorithm is either a field-oriented control algorithm or a direct torque control algorithm; The field-oriented control algorithm is used to generate the flexible frequency conversion drive pulse width modulation signal based on the reference torque command, the reference flux linkage command, and the superimposed torque and current given signal, through coordinate transformation and voltage space vector modulation.

[0011] In one possible implementation, the output feedback module is specifically used for: Based on the mechanical fatigue penalty integral data and the preset mechanical component fatigue limit life data, the remaining service life data of the load-side mechanical transmission system is calculated. The remaining service life data and the electromechanical system health assessment report are packaged into non-electrical variable monitoring and early warning data; The non-electric variable monitoring and early warning data are sent to the host computer system to achieve joint monitoring of process energy efficiency and mechanical status.

[0012] The beneficial effects of this invention are that it uses an electromechanical feature identification module to perform harmonic analysis on three-phase current and combines it with a nonlinear mechanical observer to calculate mechanical inertia, transmission clearance and resonant frequency in real time. This design can accurately distinguish between normal load fluctuations and abnormal oscillations of mechanical transmission chains without the need to install additional vibration sensors on the mechanical side, thus reducing the complexity of hardware deployment and maintenance difficulty. Attached Figure Description

[0013] The invention will now be further described with reference to the accompanying drawings.

[0014] Figure 1 This is a schematic diagram of a module of an intelligent frequency conversion control system for a three-phase asynchronous motor with first-level energy efficiency, provided in an embodiment of this application. Detailed Implementation

[0015] 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Please see Figure 1 An intelligent frequency conversion control system for a three-phase asynchronous motor with first-class energy efficiency is applied in a motor drive circuit that includes a three-phase asynchronous motor, a frequency converter, and a load-side mechanical transmission system. The system includes: a data acquisition module for acquiring three-phase voltage data, three-phase current data, and speed feedback data output by the frequency converter. The electromechanical feature identification module is used to perform harmonic analysis based on three-phase current data, extract current fluctuation characteristics, and identify the electromechanical coupling dynamic parameters of the load-side mechanical transmission system based on the current fluctuation characteristics. The electromechanical coupling dynamic parameters include mechanical inertia data, transmission clearance data, and the current resonant frequency data of the system. The fatigue risk quantification module is used to receive the next speed and pressure adjustment command sent by the host computer system, input the next speed and pressure adjustment command and the mechanical inertia data and transmission clearance data in the electromechanical coupling dynamics parameters into the electromagnetic torque pulsation and mechanical stress mapping model, and calculate the mechanical fatigue penalty integral data corresponding to the next speed and pressure adjustment command. The joint optimization control module is used to construct a comprehensive objective function, which includes energy consumption indicators and mechanical fatigue indicators calculated based on the three-phase voltage and three-phase current data in the current control cycle. Within a preset prediction time window, the comprehensive objective function is optimized and solved based on the preset energy efficiency optimization step size and mechanical fatigue penalty integral data, and the reference torque command and reference flux command are output. The active damping execution module is used to extract the corresponding resonant oscillation signal based on the current resonant frequency data of the system, and generate active damping compensation current data that is opposite in phase to the resonant oscillation signal. The active damping compensation current data is injected into the underlying vector control algorithm, and a flexible frequency conversion drive pulse width modulation signal is generated according to the reference torque command and the reference flux command. The output feedback module is used to output the flexible frequency conversion drive pulse width modulation signal to the inverter side of the frequency converter to drive the three-phase asynchronous motor and generate an electromechanical system health assessment report based on the mechanical fatigue penalty integral data.

[0017] This embodiment provides an intelligent frequency conversion control mechanism for a first-level energy-efficient three-phase asynchronous motor. Specifically, this mechanism is deployed in the drive link of a high-power centrifugal pump in the circulating cooling water station of a petrochemical park. The drive link consists of a first-level energy-efficient three-phase asynchronous motor, a frequency converter, a flexible coupling, a speed-increasing gearbox, and a multi-stage centrifugal pump. The pump set needs to operate continuously for a long time. During the day, the speed is frequently adjusted due to changes in heat exchange load, and at night, it needs to maintain pressure at low flow rate. Therefore, there is a need for power optimization, as well as fatigue accumulation problems in couplings, gear meshing pairs and bearings. The entire control system revolves around the main line of focusing on both the electrical operating status and the mechanical life change trend of the same speed adjustment action. Specifically, the data acquisition module collects three-phase voltage data, three-phase current data, and speed feedback data in control cycles. For ease of explanation, assume that a microscopic data window is collected at a certain moment, with the effective values ​​of the three-phase voltages being approximately 381V, 379V, and 382V, the effective values ​​of the three-phase currents being approximately 92A, 95A, and 91A, and the speed feedback being 1486rpm. The system does not make decisions based on single-point data, but processes time windows consisting of several consecutive sampling points, such as obtaining 100 sample values ​​within a 10ms window; this can distinguish between normal load fluctuations and continuous oscillation components with mechanical significance. The electromechanical feature identification module performs harmonic analysis on the current data within the aforementioned time window, extracts current fluctuation characteristics, and further identifies mechanical inertia, transmission clearance, and the current resonant frequency. To further illustrate the data flow process described above with an example: Assume that after analysis, the fundamental amplitude is 94A, the low-order sideband harmonic amplitude is 3A, and the amplitude of a certain high-frequency harmonic is 1.2A; at the same time, there is a low-frequency oscillation with an amplitude of about 24.8Hz in the speed feedback that is lower than the preset steady-state deviation threshold; after the system combines the sideband harmonic amplitude exceeding the preset reference threshold with the stable speed oscillation frequency, it determines that the load side is not a pure electromagnetic disturbance, but rather that the mechanical link has torsional flexibility and a small hysteresis. Furthermore, the identification module can output a set of current electromechanical coupling dynamic parameters, such as the equivalent mechanical inertia of 12 kg·m², the transmission clearance equivalent angle of 0.18°, and the current resonant frequency of the system of 26 Hz. Real-time online identification is used here, so these parameters are not fixed calibration values ​​during installation, but are updated as the operating conditions change. After obtaining the above electromechanical parameters, the fatigue risk quantification module receives the next speed and voltage adjustment command from the host computer. Assume that in order to meet the heat exchange load, the host computer requires the pump speed to be increased from 1486 rpm to 1510 rpm within the next 2 seconds, and the reference flux command is increased by a preset increment step to enhance the dynamic response. This module does not execute the instruction directly, but first sends the action to be executed into the electromagnetic torque pulsation and mechanical stress mapping model; The calculation logic of the mapping model specifically includes: if the speed adjustment brings a torque slope greater than the preset rate of change threshold, and the current mechanical inertia is greater than the first preset parameter threshold, the transmission clearance is greater than the second preset clearance threshold, and the resonant frequency is close to the adjustment excitation frequency band, then the corresponding mechanical stress rises faster. Specifically, the electromagnetic torque pulsation and mechanical stress mapping model has a built-in preset empirical mapping relationship lookup table. This lookup table takes mechanical inertia data and transmission clearance data as input dimensions and maps out the corresponding stress amplification factor. By multiplying the expected electromagnetic torque pulsation data by this stress amplification factor, the corresponding expected mechanical stress data can be calculated. The model outputs the equivalent stress amplitude sequence within the future short window as 45, 52, and 48 MPa, with a corresponding mechanical fatigue penalty integral of 0.37. The higher this integral value, the more significant the impact on the life of the transmission chain, even though the adjustment may improve the operating condition. The joint optimization control module combines the energy consumption index and the mechanical fatigue quantification index into a single objective function and solves it within the prediction time window. For ease of explanation, the prediction window can be set to the next 5 control steps, and the energy efficiency optimization step size can be set to increase or decrease the torque by 2% and the flux linkage by 1% per step. If only electrical energy-related indicators are considered, the system may tend to rapidly increase the rotational speed and pull the flux towards a more efficient curve; However, after adding fatigue penalty, the system will compare multiple candidate schemes. For example, scheme A is to accelerate quickly, with an energy index of 9.6 and a fatigue index of 4.1. Option B is a smooth acceleration, with an energy index of 9.9 and a fatigue index of 1.8; Option C is an acceleration that is delayed by half a window, with an energy index of 10.3 and a fatigue index of 1.2. After minimizing the overall objective function, option B can be selected, which generates corresponding reference torque and reference flux commands. For example, the reference torque is slowly adjusted from the current 620 N·m to 645 N·m, and the reference flux is kept at 0.97 times the rated value instead of being directly increased. This transforms a speed regulation action from simply pursuing instantaneous efficiency into a regulation process that takes into account both energy efficiency and lifespan constraints. After receiving the current resonant frequency, the active damping execution module generates a compensation current to cancel out the resonant oscillation and injects it into the underlying vector control algorithm. For example, when a significant mechanical resonance tendency is detected around 26Hz, the module constructs a set of phase-opposite compensation components and superimposes them onto the torque current channel. If the original torque current has a fluctuation of 26Hz and amplitude of 1.5A, the compensation component can be a suppression term with opposite phase and amplitude of 1.1A. After superposition, the residual oscillation is significantly reduced. The system combines the reference torque command and the reference flux command to generate a flexible frequency conversion drive pulse width modulation signal, which is output to the inverter side of the frequency converter to drive the three-phase asynchronous motor to run smoothly. After the drive is executed, the output feedback module packages the mechanical fatigue penalty integral calculated this time, the current health status, and the strategy adopted by the controller to form an electromechanical system health assessment report. For example, the report can record that the speed adjustment has been changed from rapid acceleration to smooth acceleration; the current fatigue risk of the coupling is moderate; it is recommended to check the gearbox backlash change in the next maintenance window; this report can be uploaded to the host computer periodically for use by production scheduling and equipment maintenance. As a fault-tolerant control mode, if there is a phase loss, obvious jump or timestamp misalignment in the three-phase voltage or current sampling, the system first freezes the current optimization result and uses the reference torque command and reference flux command of the previous effective cycle to maintain operation, so as to avoid making distorted decisions when the data quality is insufficient. If the speed feedback is temporarily lost, the identification module will switch to degraded mode, estimate the resonance trend based only on the harmonics on the current side, and add a safety margin to the fatigue penalty integral, for example, by 10% on the original basis, to make the control action more conservative; if the host computer does not issue new speed and voltage adjustment commands, the controller will continue to perform local energy efficiency-health joint optimization around the current load point, without making unnecessary large step adjustments. During the afternoon peak operating conditions of the circulating cooling water station, the heat exchanger group suddenly increased its cooling demand, and the host computer originally hoped that the pump group would quickly increase its speed to maintain the flow rate. The system first identified a slight increase in gearbox hysteresis from the current ripple, and the resonance trend near 26Hz was enhanced. Therefore, although the pump hydraulic conditions allowed for continued rapid acceleration, the controller switched to output a smoother reference torque and flux trajectory, and simultaneously injected active damping compensation current, which ultimately ensured that the flow rate met the standard while preventing further amplification of the coupling amplitude. The purpose of this step is to expand the traditional energy-saving control based solely on electrical parameters to a system-level control that synchronously constrains the mechanical transmission state, thereby achieving a joint balance between energy consumption, mechanical fatigue, and operational stability. To further clarify, the 100 sampled values ​​obtained within the aforementioned 10ms window are mainly used for underlying voltage and current transient sampling and current loop control refresh; while the frequency analysis used to identify the mechanical oscillation characteristics of a class of 24Hz to 30Hz is not completed directly with a single 10ms original window, but rather by splicing or sliding the feature values ​​of multiple consecutive control cycles to form a longer analysis window before trend extraction. Several 10ms raw windows can be combined into an identification window within the range of 0.5s to 2s to observe the stable changes in low-frequency torsional vibration, rotational speed oscillation, and related sideband energy. Thus, the results mentioned above, such as the 24.8Hz oscillation and the 26Hz resonant frequency, correspond to continuous estimates within a longer identification window, rather than a one-time judgment within a single 10ms window. To further clarify, the aforementioned sideband components around 24Hz to 30Hz are mechanically related frequency features extracted from the modulation sideband near the fundamental frequency of the three-phase current, the current residual signal in the synchronous rotating coordinate system, or the envelope spectrum after demodulation of the fundamental current. Their physical meaning is the low-frequency modulation of electromagnetic quantities caused by mechanical torsional vibration, rather than the existence of a dominant power supply frequency component that is independent of the fundamental frequency in the original spectrum of the phase current. With this formulation, the data link between the current harmonic characteristics and the mechanical resonant frequency becomes clearer, and it is also easier to maintain consistency with the subsequent active damping frequency band extraction process.

[0018] In a preferred embodiment of the present invention, the electromechanical feature identification module is specifically used to: perform a fast Fourier transform on the three-phase current data to separate the fundamental current data and the high-frequency harmonic current data; construct a nonlinear mechanical observer on the load side based on the high-frequency harmonic current data and the speed feedback data; and calculate the mechanical inertia data, transmission clearance data and the current resonant frequency data of the system in real time through the nonlinear mechanical observer.

[0019] This embodiment provides a specific step for electromechanical feature identification; specifically, based on the continuous operation of the above-mentioned circulating cooling water station pump group, the mechanical inertia, transmission clearance and resonant frequency on the load side are identified online only by relying on the existing three-phase current data and speed feedback data on the inverter side, without the need to install vibration sensors at the coupling or gearbox. Furthermore, if the load condition is estimated based solely on the effective value of the current or the average speed, although the overall load size can be reflected, it is impossible to distinguish between two different sources of fluctuation: the increase in hydraulic load and the torsional oscillation of the transmission chain. For example, if the current increases by 2A, one possibility is that the pump outlet pressure has increased, and another possibility is that the harmonics have been enhanced due to the meshing impact of the gearbox. In order to solve this problem, this embodiment first performs a fast Fourier transform on the three-phase current to separate the fundamental component from the high-frequency harmonic components. To illustrate the transformation results, we can assume that in the A-phase current spectrum within a certain analysis window, the fundamental amplitude at 50Hz is 93A, the amplitudes at 100Hz and 150Hz are lower than the preset lower limit of amplitude, while the sideband components near 24Hz to 30Hz are significantly raised, with an equivalent amplitude of approximately 1.4A. Phase B and Phase C also exhibit similar characteristics; the system extracts these high-frequency harmonic components related to rotational speed oscillation as observation inputs, instead of sending the entire spectrum into the subsequent model, thereby reducing the interference of irrelevant frequencies on the identification of mechanical parameters. Furthermore, a nonlinear mechanical observer on the load side is constructed based on high-frequency harmonic current data and speed feedback data; the nonlinearity here is reflected in the fact that the transmission chain is not an ideal rigid body, the transmission gap will cause the response to be sluggish when the reverse torque is small, and the resonant frequency will also drift due to changes in operating conditions. The nonlinear mechanical observer uses the torque current component and speed deviation as state variables, and performs iterative prediction through a preset state space equation to update the estimated values ​​of electromechanical coupling dynamic parameters at the current moment in a way that minimizes the observation error. The preset state space equations are established based on the motor motion equations and the torque balance equations of the mechanical transmission chain dual mass blocks. The torque current and speed deviation are used as the quantities of the system state. The state matrix is ​​iteratively updated and calculated through extended Kalman filtering or Luneburger observer algorithm. For ease of explanation, the observer can be simplified as three recursive channels: the first channel updates the mechanical inertia estimate based on the rate of change of rotational speed caused by torque changes; The second channel updates the transmission clearance estimate based on the degree of current harmonic enhancement but speed response lag. The third channel updates the current resonant frequency estimate based on the location of the current-side peak. Assuming that the actual speed increments are 5, 4, and 5 rpm in three consecutive control windows, while the ideal rigid model expects 6, 6, and 6 rpm under the same torque change, this indicates that the system inertia is too large or there is elastic dissipation in the link. Furthermore, considering that the sideband peak in the 24Hz to 30Hz frequency band has shifted from 25.2Hz to 26.0Hz, the observer can correct the current resonant frequency to around 26Hz. A solution process is presented in the form of simulation: at the initial moment, the system uses the installation and debugging values ​​as priors, the equivalent inertia is set to 10, the transmission clearance is set to 0.10°, and the resonant frequency is set to 24Hz; after one analysis window, the inertia is updated to 11.2, the clearance is 0.14°, and the frequency is 25.1Hz based on the high-frequency harmonics and rotational speed deviation. After the second analysis window, the values ​​are updated to inertia 11.8, gap 0.17°, and frequency 25.8Hz; after the third analysis window, the values ​​tend to stabilize at inertia 12.0, gap 0.18°, and frequency 26.0Hz; the control system uses these subsequent stable values ​​as real-time electromechanical coupling dynamic parameters input to the subsequent modules. As a fault-tolerant control mode, if the amplitude of the high-frequency harmonic component in the fast Fourier transform result is lower than the preset feature extraction lower limit, and the speed feedback change rate is lower than the preset identification excitation threshold, the system maintains the mechanical parameter estimation of the previous cycle and raises the parameter refresh threshold to prevent repeated oscillations when the amount of data information is insufficient; if the high-frequency harmonic suddenly amplifies abnormally at a certain moment, but the speed feedback does not show a corresponding change, it is preferentially judged as sampling noise, inverter switching interference or grid-side disturbance, and is not immediately identified as an increase in mechanical clearance; At this point, it is required that the mechanical parameters be updated only after multiple consecutive windows meet the consistency condition; if the speed encoder jitters for a short time, the observer can use the sliding median method to pre-filter the speed feedback to avoid a sudden jump in inertia estimation caused by a single glitch. During the transition phase of the above pump set, which experienced low flow pressure maintenance at night and increased flow during the day, the system found that the pump body hydraulic load change rate was within the preset safety tolerance range, but the current band near 26Hz continued to increase, and the speed increment was always slightly lower than the model prediction value when the speed was increased in small steps. After continuous updates by the nonlinear mechanical observer, the system corrected the transmission backlash from 0.10° to 0.18° and the resonant frequency from 24Hz to 26Hz; this result was used to limit the excitation frequency band of subsequent speed regulation actions. The purpose of this step is to convert the current harmonics and speed deviation information into mechanical parameters that can be directly used for control, thereby enabling online analysis of the load coupling dynamics hidden behind the electrical signals. To further clarify, the frequency position in the obvious rise of the sideband components around 24Hz to 30Hz is preferably understood as the mechanical modulation frequency component obtained by synchronous demodulation and envelope analysis of the fundamental current, or in the current residual signal in the rotating coordinate system. If we analyze the original phase current spectrum, it usually appears as paired sidebands around the fundamental frequency and its vicinity, rather than directly taking 24Hz to 30Hz as the original main frequency band of the phase current; this preserves the technical main line of extracting mechanical information through current, and also avoids confusing the mechanical characteristic frequency with the fundamental frequency of the motor power supply. To further explain, the execution level of the Fast Fourier Transform can adopt a short window sampling and long window identification method; the lower level control still collects current samples with a shorter control window, while the electromechanical parameter identification slides and splices the spectral characteristics, envelope energy or residual signals of multiple consecutive short windows, and then completes peak tracking and observer update within a longer identification window. A single short window can be used to refresh the current loop in real time, while the determination of the mechanical resonant frequency must combine the statistical results of at least multiple consecutive short windows to ensure sufficient time resolution for low-frequency torsional vibration around 25Hz. Thus, the process of correcting from 24Hz to 26Hz should be understood as the convergence result after iteration of multiple consecutive analysis windows, rather than the direct frequency calculation result within a single extremely short sampling window. Furthermore, the high-frequency harmonic current data in this embodiment is non-fundamental disturbance characteristic data relative to the fundamental steady-state current component. It can be directly derived from the non-fundamental spectrum component after the fast Fourier transform, or it can be derived from the mechanical modulation sideband characteristics obtained after synchronous demodulation, envelope analysis, or rotation coordinate system residual extraction of the fundamental current. Therefore, when the examples in the text mention the sideband around 24Hz to 30Hz, it refers to the performance of the mechanically correlated modulation frequency in the characteristic domain, rather than directly equating 24Hz to 30Hz with the independent power supply harmonics above the fundamental wave of 50Hz in the original spectrum of the phase current. Therefore, the sideband components of the high-frequency harmonic current data and the mechanical modulation components in the residual signal in this embodiment belong to different processing levels under the same identification link, and all serve the input construction of the nonlinear mechanical observer, rather than representing multiple sets of data sources that are separate from each other. To further clarify, phase A, phase B, and phase C in this article are only used to distinguish the three-phase current sampling channels, where phase A represents the first phase current channel, phase B represents the second phase current channel, and phase C represents the third phase current channel; these three are phase identifiers on the stator side of the three-phase motor, rather than independent algorithm variables. Correspondingly, the descriptions in the text such as the A-phase current spectrum and the B-phase and C-phase having similar characteristics should all be understood as explanations of the channels from which the three-phase sampling data originates.

[0020] In a preferred embodiment of the present invention, the fatigue risk quantification module is specifically used to: calculate the expected electromagnetic torque pulsation data according to the next speed and pressure adjustment command; input the expected electromagnetic torque pulsation data, mechanical inertia data, and transmission clearance data into the electromagnetic torque pulsation and mechanical stress mapping model to calculate the expected mechanical stress data; perform rainflow counting analysis on the expected mechanical stress data within a preset time period, extract the stress amplitude and cycle number, and calculate the mechanical fatigue penalty integral data based on the linear fatigue damage accumulation theory.

[0021] This embodiment provides a fatigue risk quantification step; specifically, after the mechanical inertia, transmission clearance and resonant frequency have been obtained through the aforementioned online identification, if the control action is still selected based solely on whether the energy consumption reduction requirement is met after the next speed adjustment, the cumulative effect of mechanical stress is easily overlooked. A single speed adjustment may only generate a small amount of additional stress, but in scenarios like a circulating cooling water station that operates continuously throughout the year, repeated small stresses hundreds of thousands of times can lead to significant fatigue damage. Therefore, in this embodiment, before performing speed and pressure regulation, the action to be performed is first converted into a mechanical fatigue penalty integral. Furthermore, the system calculates the expected electromagnetic torque pulsation data based on the next speed and voltage adjustment command. Assuming the host computer requires the speed to be increased by 20 rpm and the flux linkage to be increased by 3% within 0.5s, the controller can first obtain an expected torque change trajectory. For example, the average electromagnetic torque corresponding to this trajectory in the next 5 short steps is 630, 642, 648, 646, and 640 N·m, respectively, while the superimposed high-frequency pulsation amplitudes are 18, 24, 21, 15, and 12 N·m, respectively. Here, what is actually used for fatigue calculation is not the average torque itself, but the pulsating part that will excite the vibration of the mechanical link. The expected electromagnetic torque pulsation data, mechanical inertia, and transmission clearance are input into the electromagnetic torque pulsation and mechanical stress mapping model. This mapping model can be established by combining table lookup and simplified dynamic equations. The purpose is to convert the torque jitter on the electromagnetic side into equivalent stress on the coupling, gear tooth root, or shaft segment. For ease of understanding, we can assume that the current equivalent mechanical inertia is 12 and the transmission clearance is 0.18°. Under the action of pulsating amplitudes of 18, 24, 21, 15, and 12 N·m, the expected mechanical stress peak-valley sequence corresponding to the model output is 30, 46, 41, 28, and 22 MPa. If the transmission clearance is larger or the inertia is higher, the mapped stress amplitude will be larger under the same pulsating input. Next, rainflow counting analysis is performed on the expected mechanical stress data within the preset time period; since mechanical fatigue is related to the number of stress cycles, rather than just the maximum peak value, the system needs to decompose the stress time series into several closed loops; To demonstrate this process, an exemplary stress sequence can be used: 22, 46, 25, 41, 24, 30, 23 MPa; after rainflow counting, two main cycles can be identified, with a stress amplitude of approximately 12 MPa in each cycle, counted once. Another cycle has a stress amplitude of approximately 8 MPa, counted once; several residual half-cycles are converted to 0.5 cycles according to preset rules; based on the linear fatigue damage accumulation theory, the single damage corresponding to each type of stress amplitude is accumulated; for example, an amplitude of 12 MPa corresponds to a single damage of 0.18, an amplitude of 8 MPa corresponds to a single damage of 0.07, and the half-cycle converted damage is 0.04, so the total mechanical fatigue penalty integral is 0.18 + 0.07 + 0.04 = 0.29; the higher this integral, the greater the wear and tear on the transmission chain life caused by this speed adjustment; Furthermore, this integral can be used not only for single-step decision-making but also for comparisons within a multi-step prediction window. For example, if the first candidate solution has lower energy consumption but an integral of 0.40, and the second candidate solution has higher energy consumption but an integral of only 0.16, then the second candidate solution is the better solution in the global evaluation index. Thus, the fatigue penalty integral becomes an intermediate bridge connecting mechanical life and control optimization. As a fault-tolerant control mode, if the stress sequence within the preset time period is too short to form stable rain flow statistics, the system adopts a conservative strategy, giving higher weight to unclosed semi-cycles in fatigue damage to avoid underestimating the risk. If there are abnormal jumps in the mechanical inertia or transmission clearance received by the mapping model, such as changes exceeding a set proportion compared to the previous cycle, fatigue calculation will not be directly entered. Instead, smoothed estimates from the most recent cycles will be called to prevent fatigue integral distortion caused by fluctuations. If the stress sequence is lower than the material fatigue sensitivity threshold, the integral can be compressed to near zero, but not set to absolute zero, thus preserving the ability to accumulate long-term micro-damage. During the afternoon high-load phase of the circulating cooling water station, the host computer continuously issued speed-up commands to compensate for the heat exchange load. The system found that although the third speed-up command only increased the speed by 5 rpm compared to the previous one, the transmission clearance had slightly increased and the mechanical resonance frequency band was closer to the control excitation frequency. As a result, the expected stress amplitude obtained after mapping increased significantly, and the mechanical fatigue penalty integral reached 0.37, which was higher than the 0.12 and 0.18 of the previous two commands. Therefore, the subsequent controller no longer adopted the same optimization rhythm. The purpose of this step is to quantify the mechanical fatigue risk, which is originally difficult to directly enter into the control loop, into a comparable and cumulative integral index, thereby realizing a feedforward assessment of the impact of speed regulation behavior on mechanical life. In a preferred embodiment of the present invention, the joint optimization control module is specifically used to: calculate the current energy consumption based on the integral of the product of three-phase voltage data and three-phase current data within the current control cycle; Multiply the mechanical fatigue penalty integral data by a preset fatigue weight coefficient to obtain the current mechanical fatigue quantification value; add the current power consumption to the current mechanical fatigue quantification value to construct a comprehensive objective function; Within a preset prediction time window, a preset energy efficiency optimization step size is used to minimize the comprehensive objective function based on the model predictive control algorithm, generating reference torque command and reference flux command that minimize the comprehensive objective function value.

[0022] This embodiment provides a joint optimization control step; specifically, if the fatigue integral can quantify the risk of mechanical damage, and if it is still only used as monitoring information and not actually used as the control target, then the controller may still prioritize pursuing the shortest power consumption. Therefore, in this embodiment, the power consumption index and the mechanical fatigue index are directly incorporated into the same comprehensive objective function, and the optimal reference torque and reference flux are solved within the future window through model predictive control. Furthermore, the current energy consumption can be obtained by integrating the product of the three-phase voltage and the three-phase current within the current control cycle; to illustrate this process, we can assume that the instantaneous three-phase power within a certain control cycle is integrated and converted into the energy consumption of this cycle as 9.8 units. Meanwhile, the mechanical fatigue penalty integral output by the fatigue risk quantification module is 0.29; the system then presets the fatigue weight coefficient according to the importance of the process, for example, taking 6, so the current mechanical fatigue quantification value is 0.29×6=1.74 units; after adding the two, the comprehensive objective function value is 11.54. Within the prediction time window, the system will enumerate or perform rolling calculations on multiple candidate control sequences; assuming the prediction window is for the next 4 steps, the energy efficiency optimization step size specifies that the reference torque adjustment range for each step shall not exceed 10 N·m, and the reference flux adjustment range shall not exceed 1% of the rated value; Several candidate sequences can be formed. The first candidate sequence adopts rapid acceleration and strong excitation, with a cumulative energy index of 36.5 and a cumulative mechanical fatigue index of 9.2 in 4 steps, and a total target value of 45.7. The second candidate sequence adopts medium acceleration and slightly reduced flux linkage, with a cumulative energy index of 37.0 and a cumulative mechanical fatigue index of 4.8, and a total target value of 41.8. The third candidate sequence uses slow acceleration and weak excitation, with a cumulative energy index of 38.3, a cumulative mechanical fatigue index of 3.1, and a total target value of 41.4. In this example, although the energy index of the third candidate sequence is higher, it significantly reduces the fatigue index, resulting in the smallest overall objective function. Therefore, the generated reference torque command and reference flux command come from the third candidate sequence. The model predictive control here can be executed in a rolling optimization manner, that is, only the first step of the control is implemented each time, and the fatigue index is resampled, reidentified, recalculated and optimized in the next cycle; this can avoid the failure of one-time planning due to sudden load changes or parameter drift. For example, if the first step of the third candidate sequence gives a reference torque of 632 N·m and a reference flux of 0.96 times the rated value, then after the system implements this step, the next cycle will re-evaluate whether to continue along the original path. Furthermore, the fatigue weighting coefficient is not static and can be adjusted according to equipment status, production plan, or maintenance cycle. Before a major overhaul, in order to ensure safe operation, the coefficient can be set to a larger value, making the controller more stable. During the new installation or short-term emergency production increase phase, the coefficient can be reduced, allowing the system to allow a certain range of lifespan consumption in exchange for process response capability. As a fault-tolerant control mode, if there is a lack of voltage or current data, which makes it impossible to accurately integrate the current energy consumption index, the system can use a combination of the effective index of the previous cycle and the current average power estimate as a substitute, and simultaneously reduce the energy efficiency optimization step size to avoid making aggressive behavior when the objective function is incomplete. If the fatigue weight coefficient is abnormally large, causing the controller to hardly respond to speed regulation requirements, a lower limit constraint can be set to ensure that process indicators such as flow rate and pressure can still meet the minimum operating requirements. If the model predictive control does not converge within the time limit, the controller outputs a set of smooth reference values ​​near the previous optimization result, instead of reverting to the unconstrained fast control mode. When the circulating cooling water station enters the evening load reduction phase, the process side hopes that the pump set will reduce its speed appropriately to reduce power consumption; if only power indicators are considered, the controller may significantly reduce torque and demagnetize at once. However, after combining the fatigue integral, the system found that rapid deceleration would cause reverse impact on the coupling under the current transmission clearance conditions, resulting in a significant increase in fatigue index. Therefore, model predictive control ultimately provides a control trajectory that first slightly reduces torque and then slowly reduces flux linkage, so that the pump set can reduce reverse mechanical shock while meeting flow regulation requirements. The purpose of this step is to establish a unified cost comparison framework so that the trend of electrical energy change and mechanical life consumption can be measured simultaneously in the same optimization loop, thereby achieving control outputs that are more in line with the long-term operation goals of industry.

[0023] In a preferred embodiment of the present invention, the system further includes a strategy adjustment module, which is used to: compare the mechanical fatigue penalty integral data with a preset fatigue risk threshold; If the mechanical fatigue penalty integral data is greater than or equal to the preset fatigue risk threshold, the health priority strategy is triggered, the preset energy efficiency optimization step size in the joint optimization control module is reduced to the preset conservative optimization step size, and the amplitude of the active damping compensation current data is multiplied by a preset compensation gain coefficient greater than 1. Among them, the amplitude of the amplified active damping compensation current data is strictly limited by the transient current safety limiting parameter of the inverter hardware to prevent overcurrent from triggering hardware protection. If the mechanical fatigue penalty integral data is less than the preset fatigue risk threshold, the energy efficiency priority strategy is triggered to maintain the preset energy efficiency optimization step size and maintain the current amplitude of the active damping compensation current data.

[0024] This embodiment provides a strategy adjustment mechanism; specifically, although the aforementioned joint optimization has incorporated fatigue into the objective function, under certain extreme conditions, relying solely on continuous optimization may still not be intuitive enough, nor is it conducive to the equipment management layer understanding the current bias of the controller; Therefore, this embodiment further introduces a strategy adjustment module to divide the control system into two operating states: energy efficiency priority and health priority, and automatically switches between them based on whether the mechanical fatigue penalty integral exceeds the threshold. Furthermore, the system first compares the mechanical fatigue penalty integral with the preset fatigue risk threshold; assuming the threshold is set to 0.30; when the current integral is 0.18, it indicates that the life loss caused by this adjustment is within an acceptable range, the system maintains the energy efficiency priority strategy, does not change the original energy efficiency optimization step size, and does not additionally increase the active damping compensation amplitude. If the current score rises to 0.37, it indicates that if energy efficiency optimization continues at the existing step size, mechanical stress may be pushed to a higher level, triggering the health priority strategy. Under the health-first strategy, the system will perform two types of actions: First, reduce the energy efficiency optimization step size; to illustrate with an example, if the original reference torque is allowed to change by 10 N·m per step and the reference flux is allowed to change by 1% per step, then after switching, it can be reduced to 4 N·m and 0.4% per step; Second, increase the amplitude of the active damping compensation current. Assuming the original compensation current amplitude is 1.1A and the preset compensation gain coefficient is 1.4, the compensation amplitude will increase to 1.54A after switching, which is used to more effectively suppress resonant frequency components. When the two actions occur simultaneously, on the one hand, new excitation input is reduced, and on the other hand, the ability to cancel existing oscillations is enhanced. Correspondingly, when the integral is lower than the threshold, the energy efficiency priority strategy is triggered. At this time, the controller maintains the original energy efficiency optimization step size, and the active damping compensation amplitude also remains at the current value, instead of completely turning off the compensation. The mechanism of the above configuration is that even if the mechanical system is in a low-risk state, there may still be a basic torsional vibration component. Retaining the current amplitude helps to maintain stability and prevent vibration rebound due to sudden cancellation of compensation. To facilitate understanding, a simulation process with two consecutive cycles can be constructed; the fatigue penalty integral in the first cycle is 0.28, which is less than 0.30, so the system maintains energy efficiency priority, the optimization step size is still 10 N·m, and the compensation current amplitude is still 1.1 A; In the second cycle, due to a sudden increase in flow rate on the process side, the integral rises to 0.36, exceeding the threshold. The system immediately switches to health priority, the step size is reduced to 4 N·m, and the compensation amplitude is increased to 1.54 A. In the third cycle, if the integral falls back to 0.27, in order to avoid the strategy from fluctuating around the threshold, it can be required to return to energy efficiency priority after two consecutive cycles below the threshold, or a dual threshold mechanism with different entry and exit thresholds can be set. As a fault-tolerant control mode, if the fatigue penalty integral is exactly equal to the threshold, in order to avoid being too aggressive in the critical state, this embodiment is treated as high risk and prioritizes the health priority strategy. If the integral continuously exceeds the threshold but the process system has a minimum flow rigidity requirement, the controller cannot infinitely reduce the optimization step size. In this case, a conservative lower limit for the optimization step size is set to ensure that the pump unit can still reach the minimum process target within the specified time. If the compensation gain after amplification is greater than or equal to the difference between the inverter current limit and the preset safety margin, the compensation amplitude will be limited to not exceed the safety boundary, and the insufficient vibration suppression capability will be compensated by further reducing the optimization step size; if the sensor data is unreliable for a short time, the strategy adjustment module can temporarily maintain the previous state and will not switch the strategy on a single abnormal data. Under the high-temperature conditions of the circulating cooling water station in summer, the heat exchange load continued to rise, and the host computer repeatedly requested a rapid increase in flow rate; the first two system assessments were both at low risk, and energy efficiency was prioritized. The third assessment found that the mechanical fatigue penalty integral caused by gearbox backlash reached 0.35, exceeding the threshold. The controller then narrowed the optimization step size and increased the active damping compensation current amplitude. Although the acceleration process was slightly slower, the coupling amplitude was controlled, avoiding abnormal vibration alarms during the night shift. The purpose of this mechanism is to add a clear state management logic on top of continuous optimization, so as to prioritize the protection of machine health under high-risk operating conditions and make full use of energy efficiency potential under low-risk operating conditions.

[0025] In a preferred embodiment of the present invention, the active damping execution module is specifically used to: extract the resonant frequency band corresponding to the current resonant frequency data of the system and the resonant oscillation signal within the resonant frequency band; generate a compensation signal with the opposite phase to the resonant oscillation signal as active damping compensation current data; and superimpose the active damping compensation current data onto the torque current setpoint of the underlying vector control algorithm to cancel the frequency components that cause mechanical resonance, thereby forming the superimposed torque current setpoint signal.

[0026] This embodiment provides an active damping execution step; specifically, after the aforementioned strategy switching has been able to determine whether to be more conservative, if there is no specific vibration suppression execution method, the system can still only passively avoid resonance by slowing down the speed adjustment. Therefore, this embodiment further generates a compensation current with opposite phase for the identified resonant frequency, directly weakening the frequency components that cause mechanical resonance at the current control level; Furthermore, the system first extracts the resonant frequency band from the current resonant frequency data; assuming that the center resonant frequency given by the identification result is 26Hz, considering the operating condition drift, the resonant frequency band can be set to 24Hz to 28Hz; Extract the resonant component within the frequency band from the current or speed oscillation signal; for ease of explanation, assume that a 26Hz oscillation term is separated from the torque current channel, which is equivalent to an amplitude of 1.5A and an initial phase of 30°. The active damping module does not need to change other frequency bands; it only constructs a compensation signal with opposite phase for the components within this resonant frequency band, such as generating a compensation current with an amplitude of 1.2A and a phase of 210°. The compensation current is superimposed on the torque current setpoint of the underlying vector control algorithm; if the original torque current setpoint is 18A at a certain moment, and a positive 1.5A 26Hz oscillation component is superimposed, the equivalent oscillation after compensation can be reduced to about 0.3A. The controller then generates a voltage modulation signal based on the processed given value, which ultimately makes the electromagnetic torque output by the motor smoother within the resonant frequency band. This is different from simply reducing the overall torque. The former is to specifically weaken the oscillation of a specific frequency band, and therefore has less impact on the process response speed. Furthermore, the amplitude of the compensation signal can be linked with the aforementioned strategy adjustment results; when the system is in a health-first strategy, the compensation gain is increased, for example, from 1.0 to 1.4; when in an energy-first strategy, the base gain is maintained; in this way, the active damping module can independently complete the band cancellation and also form a consistent action with the upper-level strategy. To better illustrate the data flow, a more detailed simulation can be provided: In a certain control cycle, the energy of the 24-28Hz frequency band is detected to be 0.85; after compensation, the energy of the same frequency band in the next cycle drops to 0.42; and in the cycle after that, it drops to 0.31. The system uses the decrease in frequency band energy as a closed-loop feedback for the damping effect, and appropriately fine-tunes the compensation amplitude to avoid overcompensation or undercompensation. If the compensation is too strong, it may introduce new fluctuations in adjacent frequency bands. Therefore, the modules are usually added or removed gradually, rather than changed drastically all at once. As a fault-tolerant control mode, if the resonant frequency drifts rapidly in a short time, such as a sudden change from 26Hz to 29Hz, the system first maintains a small compensation at the previous center frequency, while rapidly updating the frequency band range. After the new frequency stabilizes for several consecutive cycles, it fully switches to the new compensation center. If the phase of the extracted resonant component is unstable, it indicates that the confidence level of the current feature extraction is insufficient. Therefore, the amplitude of the active damping compensation current data is multiplied by an attenuation coefficient of 0.5 to prevent the phase estimation error from causing the reverse compensation to become the same direction excitation. If the current limit is close after superimposed compensation, the main torque control is given priority, and the compensation signal is clipped to ensure that the motor does not overcurrent. During a nighttime speed reduction process at the circulating cooling water station, the pump set experienced a rapid drop in flow rate, which intensified the torsional oscillation around 26Hz. After the system extracted the increased energy in this frequency band, it actively generated a compensation current with opposite phase and injected it into the torque current setpoint. After several control cycles, the equivalent vibration characteristics on the gearbox side were significantly reduced, while the pump outlet pressure remained within the allowable range of the process. The purpose of this step is to implement the suppression of mechanical resonance at the current actuation layer, thereby achieving active vibration reduction in specific dangerous frequency bands, rather than passively avoiding them by simply reducing the overall control strength. Furthermore, the 24Hz to 28Hz resonant frequency band in this embodiment is preferably extracted from the torque current residual signal in the synchronous rotating coordinate system, the envelope signal after fundamental wave demodulation, or the speed swing signal, rather than simply understood as directly extracting an absolute frequency band parallel to the main power frequency component from the original three-phase phase current spectrum. In other words, active damping targets the modulation component corresponding to mechanical torsional vibration in the control coordinate system. Therefore, the compensation current can be reasonably superimposed on the torque current setpoint and transformed into targeted suppression of the mechanical resonance frequency band through the vector control link. Furthermore, the phase acquisition of the above-mentioned compensation signal can be achieved by combining the results of the resonant component zero crossover, phase-locked tracking, or bandpass filtering analytical signal in the most recent control cycles. Its core is not to require absolutely accurate phase estimation within a single sampling, but to obtain a stable relative phase relationship in continuous cycles. This avoids phase misjudgment caused by excessively short single-cycle samples, and also ensures that the phase-opposite compensation is consistent with the resonant frequency data obtained by the aforementioned long-window identification.

[0027] In a preferred embodiment of the present invention, the underlying vector control algorithm is a field-oriented control algorithm or a direct torque control algorithm; the field-oriented control algorithm is used to generate a flexible frequency conversion drive pulse width modulation signal based on the reference torque command, the reference flux command, and the superimposed torque current given signal, through coordinate transformation and voltage space vector modulation.

[0028] This embodiment provides a specific execution method for low-level vector control; specifically, the above-mentioned active damping compensation signal has been superimposed at the torque current given end, but if its landing path in the low-level control is not clear, the compensation signal and the reference torque and reference flux output by the upper layer are easily separated from each other. Therefore, this embodiment further illustrates that the underlying vector control can adopt a field-oriented control algorithm or a direct torque control algorithm, and preferably, under the field-oriented control framework, the three are uniformly converted into a flexible frequency conversion drive pulse width modulation signal; Furthermore, under the field-oriented control mode, the system first calculates the basic torque current command based on the reference torque command, and then superimposes it with the active damping compensation current to form the final torque current command. At the same time, it forms the excitation current command based on the reference flux command. A simplified example can be used to illustrate this: Assume that the reference torque output from the upper layer corresponds to a base torque current of 16A, and the active damping compensation current is -1.2A at the current moment. Then, the superimposed torque current is 14.8A; the reference flux corresponds to an excitation current of 8.5A; the controller compares these two current settings with the sampled current, obtains the voltage command through the current regulator, and then generates the inverter switching pulse through coordinate transformation and space vector modulation. The key to this process is that the compensation signal does not bypass the original controller and act independently, but enters the unified modulation link as part of the torque current command; this ensures that the compensation action and the main drive action are synchronized in time and coordinated in terms of voltage utilization and current constraints. For example, if the main control wants to increase the torque at a certain moment, while the compensation signal wants to weaken the 26Hz oscillation, the final output will be a comprehensive command that increases the average torque but suppresses jitter in a specific frequency band, rather than two conflicting commands. For direct torque control, a similar idea can be adopted, that is, the compensation information is converted into the torque error or voltage vector selection logic, so that the system can reduce the excitation of the mechanical resonance frequency band while maintaining a fast dynamic response. Considering that the circulating cooling water station pump set emphasizes stable and long-term operation, this embodiment preferably adopts a magnetic field directional control method, whose current loop structure is more conducive to achieving fine-grained compensation injection; To illustrate the signal transformation relationship, we can assume that the voltage command obtained by coordinate transformation at a certain moment is 120V on the d-axis and 168V on the q-axis, which, after space vector modulation, is converted into the duty cycle sequence of the three-phase bridge arms: 0.62, 0.48, and 0.55. In the next control cycle, since the compensation signal changes to -0.8A, the q-axis voltage demand decreases slightly, and the duty cycle changes accordingly to 0.60, 0.49, and 0.56. Through this continuous adjustment of the amplitude change rate being less than the preset step size threshold, flexible suppression of mechanical oscillations is achieved. As a fault-tolerant control mode, if the rotor flux orientation deviation is too large when using field-oriented control, such as a sudden change in the estimated flux angle, the system can temporarily reduce the weight of the compensation signal to prioritize orientation stability, and then gradually restore the compensation injection after the flux estimation is restored. If the voltage vector switching is too frequent under direct torque control, resulting in a significant increase in switching losses, a minimum hold time constraint can be added to the vector selection logic. If the superimposed torque current exceeds the upper limit of the current loop, the amplitude is first limited and the compensation component is compressed proportionally to prevent current saturation caused by compensation. During the aforementioned nighttime deceleration phase, the controller operates using a field-oriented control method; the reference torque provided by the upper layer decreases slightly, but the active damping module simultaneously detects an increase in 26Hz oscillation, so it adds a reverse compensation amount to the torque current setpoint. After coordinate transformation and space vector modulation, the inverter outputs a smoother pulse width modulation signal, which enables the motor to maintain sufficient water pressure during deceleration and does not trigger the gearbox vibration alarm. The purpose of this step is to clarify the coupling path between the upper-level joint optimization results and the lower-level execution signals, so as to achieve consistent output of reference torque, reference flux and active damping compensation in the same control framework. Furthermore, it should be noted that the output signal of the flexible frequency conversion drive pulse width modulation signal voltage command after modulation and the inverter switching pulse in this embodiment are corresponding descriptions of the same underlying execution output at different description levels: Among them, the current regulator and coordinate transformation obtain the voltage command before modulation, and the space vector modulation forms the pulse width modulation duty cycle or inverter switching pulse. The three correspond to each other step by step along the same control link, and finally fall together on the flexible frequency conversion drive pulse width modulation signal output. Therefore, there are not multiple independent drive output signals in this paper, but the same output result is explained from the control algorithm layer, modulation layer and execution layer. To further clarify, the d-axis and q-axis in the text represent the excitation axis and torque axis in the synchronous rotating coordinate system used for field-oriented control, respectively: the d-axis is used to characterize the axial component related to flux linkage establishment, and the q-axis is used to characterize the axial component related to electromagnetic torque adjustment. Therefore, d and q in the above d-axis 120V and q-axis 168V are only coordinate axis names used to distinguish the two types of voltage command components, and are not independent mathematical variables without explanation.

[0029] In a preferred embodiment of the present invention, the output feedback module is specifically used to: calculate the remaining service life data of the load-side mechanical transmission system based on the mechanical fatigue penalty integral data and the preset mechanical component fatigue limit life data; The remaining service life data and the electromechanical system health assessment report are packaged into non-electric variable monitoring and early warning data; the non-electric variable monitoring and early warning data are sent to the host computer system to achieve joint monitoring of process energy efficiency and mechanical status.

[0030] This embodiment provides an output feedback and life warning step; specifically, after the aforementioned control loop can constrain fatigue risk in real time, if the relevant results only remain inside the controller, then maintenance personnel will find it difficult to arrange maintenance based on this, and process schedulers will also be unable to understand the reasons behind some seemingly less aggressive speed regulation strategies. Therefore, this embodiment further converts the mechanical fatigue penalty integral into remaining service life data and sends it to the host computer in the form of non-electrical variable monitoring and early warning data to achieve joint monitoring of process operation and equipment health; Furthermore, the system calculates the remaining service life based on the mechanical fatigue penalty integral and the preset fatigue limit life data of the mechanical components; For ease of explanation, we can assume that the fatigue limit life of the coupling's elastic element is converted to 100 damage units, the current historical cumulative damage has reached 62.4 units, and the newly generated mechanical fatigue penalty integral is converted to 0.37 damage units. Therefore, the updated cumulative damage is 62.77 units, and the remaining life is approximately 37.23%. If we combine the current average daily speed adjustment frequency, recent risk trends, and operating intensity, we can further convert it into the estimated remaining number of days of operation, for example, 128 days. The life calculation here is not a precise fracture prediction in the sense of mechanical design, but an estimate of remaining availability for operation and maintenance decisions. The system packages the remaining service life data and health assessment report into non-electric variable monitoring and early warning data; this data package may include at least the current fatigue level, cumulative damage, remaining service life percentage, remaining operational days, current resonant frequency, the status of the most recent strategy switch, and recommended maintenance actions; To maintain clarity of the structure, a simplified recording method can be used, such as: Equipment number P-201 drive chain; current fatigue risk is medium to high; remaining lifespan is 37.23%; current resonant frequency is 26Hz; health priority has been triggered 3 times in the last 12 hours; it is recommended to review the coupling and gearbox hysteresis within 72 hours. After receiving the data, the host computer can display the power consumption trend and mechanical health trend side by side in the process monitoring screen; for example, in the interface with the same time axis, one curve displays the power consumption per unit flow, and the other curve displays the fatigue penalty integral or the rate of decline of remaining life. In this way, the production scheduling end can intuitively monitor that although a certain speed adjustment operation corresponds to a low instantaneous power consumption, it is accompanied by a significant increase in fatigue risk. Conversely, a smooth adjustment that slightly sacrifices instantaneous efficiency corresponds to a relatively stable mechanical lifespan; thus, the control result no longer manifests as an inexplicable control output, but becomes a basis that can be directly used for operation and maintenance analysis. To demonstrate the warning logic, a three-level warning threshold can be set: remaining lifespan above 50% is normal; 20% to 50% is of concern; below 20% is a warning; if the system's current calculation is 37.23%, the host computer interface will mark it as a concern and prompt you to check in the planned shutdown window. If high-risk speed adjustments continue for several days, causing the remaining lifespan to drop to 18%, the system can not only upload early warning data, but also request the host computer to limit the future speed adjustment slope or arrange for the backup pump to switch, in order to avoid the main pump failing during critical production periods. As a fault-tolerant control mode, if the fatigue limit life database lacks the model parameters of a specific mechanical component, the system can call the conservative default values ​​of similar components for estimation and mark the life model using default parameters in the report; If the communication with the host computer is interrupted, the output feedback module can cache the monitoring and early warning data of non-electric variables from the most recent few periods locally, and retransmit them in batches after the link is restored. If the remaining life calculation is incomplete due to missing historical cumulative data, the system will at least upload the current fatigue trend and risk level, and will not force the system to give an accurate life figure with serious deviations. After two months of continuous operation of the circulating cooling water station, the system found that due to frequent process switching, the fatigue damage growth rate of the P-201 pump set coupling exceeded the preset damage growth rate threshold. Based on this, the output feedback module calculates that its remaining lifespan has decreased from 100% at the initial installation stage to 37.23%, and generates non-electrical variable monitoring and early warning data, which is then sent to the host computer. The on-duty engineer saw on the monitoring interface that although the unit flow power consumption of the pump set was still at a low level, the mechanical health trend had deteriorated significantly. Therefore, the pump set was included in the next maintenance window in advance and the nighttime pump switching strategy was adjusted. The purpose of this step is to transform the fatigue risk assessment formed within the control loop into executable information for operation and maintenance and scheduling, thereby achieving joint visualization, joint early warning and joint decision-making of process energy efficiency and mechanical status.

[0031] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. An intelligent frequency conversion control system for a Class I energy-efficient three-phase asynchronous motor, applied in a motor drive circuit comprising a three-phase asynchronous motor, a frequency converter, and a load-side mechanical transmission system, characterized in that, include: The data acquisition module is used to acquire the three-phase voltage data, three-phase current data, and speed feedback data output by the frequency converter; The electromechanical feature identification module is used to perform harmonic analysis based on the three-phase current data, extract current fluctuation features, and identify the electromechanical coupling dynamic parameters of the load-side mechanical transmission system based on the current fluctuation features. The electromechanical coupling dynamic parameters include mechanical inertia data, transmission clearance data, and the current resonant frequency data of the system. The fatigue risk quantification module is used to receive the next speed and pressure adjustment command sent by the host computer system, input the next speed and pressure adjustment command and the mechanical inertia data and transmission clearance data in the electromechanical coupling dynamics parameters into the electromagnetic torque pulsation and mechanical stress mapping model, and calculate the mechanical fatigue penalty integral data corresponding to the next speed and pressure adjustment command. The joint optimization control module is used to construct a comprehensive objective function, wherein the comprehensive objective function includes an energy consumption index calculated based on the three-phase voltage data and the three-phase current data in the current control cycle, as well as a mechanical fatigue index. Within a preset prediction time window, the comprehensive objective function is optimized and solved based on a preset energy efficiency optimization step size and the mechanical fatigue penalty integral data, and a reference torque command and a reference flux linkage command are output. The active damping execution module is used to generate active damping compensation current data based on the current resonant frequency data of the system, inject the active damping compensation current data into the underlying vector control algorithm, and generate a flexible frequency conversion drive pulse width modulation signal according to the reference torque command and the reference flux linkage command. The output feedback module is used to output the flexible frequency conversion drive pulse width modulation signal to the inverter side of the frequency converter to drive the three-phase asynchronous motor to run, and generate an electromechanical system health assessment report based on the mechanical fatigue penalty integral data.

2. The intelligent frequency conversion control system for a first-class energy-efficiency three-phase asynchronous motor according to claim 1, characterized in that, The electromechanical feature identification module is specifically used for: Perform a fast Fourier transform on the three-phase current data to separate the fundamental current data and the high-frequency harmonic current data; Based on the high-frequency harmonic current data and the rotational speed feedback data, a nonlinear mechanical observer is constructed on the load side; The nonlinear mechanical observer can be used to calculate the mechanical inertia data, the transmission clearance data, and the current resonant frequency data of the system in real time.

3. The intelligent frequency conversion control system for a first-class energy-efficiency three-phase asynchronous motor according to claim 1, characterized in that, The fatigue risk quantification module is specifically used for: Calculate the expected electromagnetic torque pulsation data based on the next speed and voltage regulation command; The expected electromagnetic torque pulsation data, the mechanical inertia data, and the transmission clearance data are input into the electromagnetic torque pulsation and mechanical stress mapping model to calculate the expected mechanical stress data. Rainflow counting analysis is performed on the expected mechanical stress data within a preset time period to extract the stress amplitude and cycle number, and the mechanical fatigue penalty integral data is calculated by accumulating the data based on the linear fatigue damage accumulation theory.

4. The intelligent frequency conversion control system for a first-class energy-efficient three-phase asynchronous motor according to claim 1, characterized in that, The joint optimization control module is specifically used for: The current energy consumption is calculated by integrating the product of the three-phase voltage data and the three-phase current data within the current control cycle. Multiply the mechanical fatigue penalty integral data by a preset fatigue weight coefficient to obtain the current mechanical fatigue quantification value; The current power consumption is added to the current mechanical fatigue quantification value to construct the comprehensive objective function; Within the preset prediction time window, the preset energy efficiency optimization step size is used to minimize the comprehensive objective function based on the model predictive control algorithm, generating the reference torque command and the reference flux command that minimize the value of the comprehensive objective function.

5. The intelligent frequency conversion control system for a first-class energy-efficiency three-phase asynchronous motor according to claim 1, characterized in that, The system further includes a strategy adjustment module, which is used for: The mechanical fatigue penalty integral data is compared with a preset fatigue risk threshold. If the mechanical fatigue penalty integral data is greater than or equal to the preset fatigue risk threshold, a health priority strategy is triggered, the preset energy efficiency optimization step size in the joint optimization control module is reduced to a preset conservative optimization step size, and the amplitude of the active damping compensation current data is multiplied by a preset compensation gain coefficient greater than 1. If the mechanical fatigue penalty integral data is less than the preset fatigue risk threshold, an energy efficiency priority strategy is triggered to maintain the preset energy efficiency optimization step size and the current amplitude of the active damping compensation current data.

6. The intelligent frequency conversion control system for a first-level energy-efficiency three-phase asynchronous motor according to claim 1, characterized in that, The active damping execution module is specifically used for: Extract the resonant frequency band corresponding to the current resonant frequency data of the system and the resonant oscillation signal within the resonant frequency band; generate a compensation signal with the opposite phase to the resonant oscillation signal as the active damping compensation current data; The active damping compensation current data is superimposed on the torque current setpoint of the underlying vector control algorithm to cancel out the frequency components that cause mechanical resonance, forming a superimposed torque current setpoint signal.

7. The intelligent frequency conversion control system for a first-class energy-efficiency three-phase asynchronous motor according to claim 6, characterized in that, The underlying vector control algorithm is either a field-oriented control algorithm or a direct torque control algorithm; The field-oriented control algorithm is used to generate the flexible frequency conversion drive pulse width modulation signal based on the reference torque command, the reference flux linkage command, and the superimposed torque and current given signal, through coordinate transformation and voltage space vector modulation.

8. The intelligent frequency conversion control system for a first-class energy-efficient three-phase asynchronous motor according to claim 1, characterized in that, The output feedback module is specifically used for: Based on the mechanical fatigue penalty integral data and the preset mechanical component fatigue limit life data, the remaining service life data of the load-side mechanical transmission system is calculated. The remaining service life data and the electromechanical system health assessment report are packaged into non-electrical variable monitoring and early warning data; The non-electric variable monitoring and early warning data are sent to the host computer system to achieve joint monitoring of process energy efficiency and mechanical status.

Images