A Reverse Osmosis High-Pressure Pump Motor Control Method Based on Digital Twin and Multi-Objective Optimization

The reverse osmosis high-pressure pump motor control method, which utilizes digital twins and multi-objective optimization, solves the problems of low efficiency, high energy consumption, and slow response in existing technologies. It achieves high efficiency, low energy consumption, and stable operation under all working conditions, and extends the motor's lifespan.

CN123122341APending Publication Date: 2026-11-17TANGSHAN SANYOU BLUE OCEAN TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611548124.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-08
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing reverse osmosis high-pressure pump motor control technology cannot optimize efficiency in real time when the load fluctuates or the speed changes, resulting in increased energy consumption. Furthermore, it lacks predictive capabilities and a long-term self-learning optimization mechanism, making it impossible to achieve optimal energy saving and stable operation under all operating conditions.

Method used

A control method based on digital twins and multi-objective optimization is adopted. By combining the digital twin model with the multi-objective optimization algorithm, the load trend is predicted and the optimal control parameters are generated. A real-time efficiency map is constructed to achieve high efficiency, low energy consumption and stable operation under all working conditions.

Benefits of technology

It achieves near-optimal efficiency operation under different load, speed and temperature conditions, reduces energy consumption, optimizes dynamic response to balance energy consumption, temperature rise, torque fluctuation and noise, extends motor life, and continuously optimizes control performance through online learning.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123122341A_ABST
    Figure CN123122341A_ABST
Patent Text Reader

Abstract

This invention relates to the field of motor control technology, specifically a reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization. By real-time acquisition of state parameters such as voltage, current, speed, torque, temperature, load rate, vibration, and noise during the operation of the reverse osmosis high-pressure pump motor, a digital twin model corresponding to the actual operating state of the reverse osmosis high-pressure pump motor is established. Combined with the reverse osmosis high-pressure pump motor efficiency map and multi-objective optimization functions, the control parameters of the reverse osmosis high-pressure pump motor are dynamically adjusted, enabling the reverse osmosis high-pressure pump motor to approach or reach its optimal operating state under different operating conditions. This achieves the technical effects of reducing energy consumption, suppressing temperature rise, reducing torque pulsation, reducing noise, and extending the service life of the reverse osmosis high-pressure pump motor.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of motor control technology, specifically a reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization. Background Technology

[0002] Existing reverse osmosis high-pressure pump motor control technology is widely used in permanent magnet synchronous reverse osmosis high-pressure pump motors, brushless DC reverse osmosis high-pressure pump motors, and induction reverse osmosis high-pressure pump motors, often employing fixed parameter control or single-objective optimization control methods. Fixed parameter control cannot optimize efficiency in real time when the load fluctuates or the speed changes, leading to increased energy consumption; single-objective optimization usually only optimizes efficiency, ignoring torque stability, temperature rise, and noise constraints, leaving reverse osmosis high-pressure pump motors with problems of low efficiency, short lifespan, and high vibration and noise under complex operating conditions.

[0003] Furthermore, most existing control methods lack predictive capabilities, relying solely on feedback adjustments based on the current state, resulting in delayed responses and an inability to proactively adjust control parameters to cope with dynamic loads. They also lack long-term self-learning optimization mechanisms, failing to leverage historical data to continuously improve energy efficiency. Existing efficiency maps or empirical curves often only cover fixed speeds and load ranges, failing to achieve optimal energy savings across all operating conditions.

[0004] Therefore, existing technologies have significant shortcomings in terms of energy saving under all operating conditions, multi-objective optimization, dynamic load response, and long-term optimization. There is a need for an adaptive energy-saving control method for reverse osmosis high-pressure pump motors that can combine digital twin models, all-operating-condition efficiency prediction, multi-objective optimization algorithms, and self-learning mechanisms to achieve the goals of high efficiency, low energy consumption, stable operation, and extended lifespan. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization.

[0006] The technical solution adopted by this invention to solve its technical problem is:

[0007] A reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization includes the following steps:

[0008] S1. Collect real-time status parameters of the reverse osmosis high-pressure pump motor during operation and control parameters of the previous control cycle;

[0009] S2. Perform time synchronization, outlier removal, filtering and normalization on the collected parameters to obtain the state vector and control vector;

[0010] S3. Input the obtained state vector and control vector into the digital twin model of the reverse osmosis high-pressure pump motor for state synchronization, so that the current state of the digital twin model is synchronized with the actual reverse osmosis high-pressure pump motor, and predict the state vector at the next moment based on the discrete state update mapping of the digital twin model.

[0011] S4. Based on historical load data, current load torque, and load change rate, predict the load trend within future control cycles;

[0012] S5. Based on the d-axis current, q-axis current, rotational speed, and winding temperature under the current state vector, calculate the total loss value under the current state using the digital twin model, and predict the winding temperature rise in future control cycles based on the total loss value.

[0013] S6. Based on the digital twin model of the current state, the predicted load trend and the predicted winding temperature rise, generate multiple candidate control parameter combinations, and use the digital twin model to calculate the predicted efficiency, predicted temperature rise, predicted torque pulsation and predicted vibration noise corresponding to each candidate parameter combination.

[0014] S7. Using the d-axis current setpoint, q-axis current setpoint, flux linkage setpoint, and power device switching frequency in the candidate control parameter combination as independent variables and the predicted efficiency corresponding to the digital twin model as the dependent variable, construct a real-time efficiency map, and classify the candidate control parameter combination with a predicted efficiency not lower than a set threshold into the high efficiency region under the premise of meeting safety constraints.

[0015] S8. Construct a comprehensive objective function, which includes efficiency optimization index, temperature rise constraint index, torque ripple index, vibration and noise index, and control smoothness index. At the same time, dynamically adjust the weight of each index according to the current operating conditions. Solve for the optimal control parameter sequence that minimizes the comprehensive objective function from the candidate control parameter combinations in the high-efficiency region, and generate control commands based on the optimal control parameter sequence.

[0016] S9. Send the control command to the power drive module of the reverse osmosis high-pressure pump motor, thereby driving the reverse osmosis high-pressure pump motor to run according to the control command;

[0017] S10. Collect the current running actual state vector and calculate the actual running efficiency;

[0018] S11. Based on the difference between the actual operating efficiency and the predicted efficiency corresponding to the control command, the corresponding efficiency value in the efficiency map is corrected. At the same time, based on the error between the actual state vector and the corresponding predicted state vector, the parameters of the digital twin model are corrected.

[0019] As a preferred embodiment, a further technical solution of the present invention is:

[0020] Preferably, the digital twin model includes an electromagnetic sub-model, a mechanical sub-model, a loss sub-model, a thermal sub-model, and an efficiency sub-model; wherein,

[0021] The electromagnet model is used to describe the relationship between d-axis current, q-axis current, d-axis voltage, q-axis voltage, rotational speed, magnetic flux, and inductance. The d-axis and q-axis voltage equations are used to predict current changes, respectively.

[0022] The mechanical sub-model is used to describe the relationship between electromagnetic torque, load torque, frictional resistance torque, and changes in mechanical speed;

[0023] The loss sub-model is used to calculate copper loss, iron loss, mechanical loss, and additional loss, and then to obtain the total loss;

[0024] The thermal model is used to describe the winding temperature change, which depends on the total losses of the reverse osmosis high-pressure pump motor, ambient temperature, equivalent thermal resistance, and equivalent heat capacity.

[0025] The efficiency sub-model is used to calculate the motor operating efficiency based on the output power and total losses.

[0026] Preferably, in S4, a sliding window weighted method is used to predict the load trend:

[0027] ;

[0028] in, This represents the load torque predicted at the (k+1)th sampling time. This represents the actual load torque at the kr-th sampling time. Indicates the corresponding weight coefficient; n represents the prediction window size;

[0029] In application scenarios where the load change rate exceeds the set conditions, the prediction window size is shortened by a set ratio, and the weight coefficients are allocated according to the principle that the weight coefficients are higher the closer to the k+1 sampling time. In application scenarios where the load change rate is lower than the set conditions, the prediction window size is extended by a set ratio.

[0030] Preferably, the comprehensive objective function in S8 is expressed as:

[0031] ;

[0032] in, Indicates the length of the prediction time domain; These are the weights of each indicator; Indicates efficiency optimization metrics; Indicates the temperature rise constraint index; This indicates the torque ripple index; Indicates vibration and noise levels; This indicates the smoothness control index.

[0033] Preferably, the efficiency optimization index is used to evaluate the prediction efficiency corresponding to the candidate control parameter combination, and the calculation principle is as follows:

[0034] ;

[0035] in, This represents the efficiency optimization index for the k-th control cycle. Indicates the predictive efficiency of the digital twin model;

[0036] The temperature rise constraint index is used to evaluate the extent to which the maximum allowable temperature of the winding is predicted. The calculation principle is as follows:

[0037] ;

[0038] in, This indicates the temperature rise constraint index. This represents the winding temperature predicted by the digital twin model for the k-th control cycle. Indicates the safety winding temperature threshold. Indicates the maximum permissible temperature;

[0039] The torque ripple index is used to evaluate the variation in predicted electromagnetic torque between adjacent control cycles. The calculation principle is as follows:

[0040] ;

[0041] in, Indicates the torque ripple value. This represents the electromagnetic torque predicted in the k-th control cycle. This represents the electromagnetic torque predicted over k-1 control cycles. Indicates the torque reference value;

[0042] Vibration and noise levels are used to evaluate the vibration and noise levels of reverse osmosis high-pressure pump motors. The calculation principle is as follows:

[0043] ;

[0044] in, Indicates vibration and noise index, This represents the normalized vibration acceleration. This represents the normalized noise equivalent value;

[0045] The control smoothness index is used to evaluate the magnitude of change in the control parameter vector between two adjacent control cycles in the current candidate control parameter combination. The calculation principle is as follows:

[0046] ;

[0047] in, This indicates the smoothness control index. This represents the control parameter vector for the k-th control cycle. This represents the control parameter vector for k-1 control cycles. This represents the weight matrix.

[0048] Preferably, the process of dynamically adjusting the weights of each indicator in S8 according to the current operating conditions includes:

[0049] All weights satisfy the normalization constraint: ;

[0050] Based on the current operating conditions, determine the target set S that needs to be prioritized for enhancement: when the predicted winding temperature is below the safety threshold, include efficiency optimization indicators in the target set; when the predicted winding temperature is close to the maximum allowable temperature, include temperature rise constraint indicators in the target set; when the load change rate exceeds the set threshold, include torque ripple indicators and control smoothness indicators in the target set; when vibration or noise exceeds the set threshold, include vibration and noise indicators in the target set.

[0051] Calculate the total weight of enhancements in the target set. and the total weight of non-enhancing terms outside the target set. ;

[0052] Determine the amplification factor based on the degree of deviation from the current operating conditions. The weights within the target set are then amplified and updated. ;

[0053] Calculate the overall reduction factor for the non-enhancing terms: The weights of non-enhancing terms are reduced using the same reduction factor: ;

[0054] Perform boundary constraint checks and set a minimum value for each weight. If the adjusted weight of a certain item is lower than the set minimum value, then the weight will be locked. .

[0055] Preferably, the safety constraints in S7 include stator current constraints, terminal voltage constraints, winding temperature constraints, electromagnetic torque constraints, and mechanical speed constraints.

[0056] This invention also discloses a system for implementing the above-mentioned reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization, comprising:

[0057] The data acquisition and processing module is used to acquire real-time status parameters and control parameters from the previous control cycle during the operation of the reverse osmosis high-pressure pump motor; it performs time synchronization, outlier removal, filtering, and normalization on the acquired parameters to obtain the status vector and control vector.

[0058] The digital twin module is used to input the obtained state vector and control vector into the digital twin model of the reverse osmosis high-pressure pump motor for state synchronization, so that the current state of the digital twin model is synchronized with the actual reverse osmosis high-pressure pump motor, and the state vector of the next moment is predicted based on the discrete state update mapping of the digital twin model.

[0059] The load trend prediction module is used to predict the load trend in future control cycles based on historical load data, current load torque, and load change rate.

[0060] The winding temperature rise prediction module is used to calculate the total loss value under the current state based on the d-axis current, q-axis current, rotational speed and winding temperature under the current state vector using a digital twin model, and to predict the winding temperature rise in future control cycles based on the total loss value.

[0061] The parameter generation module is used to generate multiple candidate control parameter combinations based on the current state digital twin model, the predicted load trend, and the predicted winding temperature rise. The digital twin model is used to calculate the predicted efficiency, predicted temperature rise, predicted torque pulsation, and predicted vibration noise corresponding to each candidate parameter combination.

[0062] The efficiency map module is used to construct a real-time efficiency map with the d-axis current setpoint, q-axis current setpoint, flux linkage setpoint and power device switching frequency in the candidate control parameter combination as independent variables and the predicted efficiency corresponding to the digital twin model as the dependent variable. Under the premise of meeting safety constraints, candidate control parameter combinations with predicted efficiency not lower than a set threshold are classified into the high-efficiency region.

[0063] The multi-objective optimization module is used to construct a comprehensive objective function, which includes efficiency optimization index, temperature rise constraint index, torque ripple index, vibration and noise index, and control smoothness index. At the same time, the weights of each index are dynamically adjusted according to the current operating conditions. The module finds the optimal control parameter sequence that minimizes the comprehensive objective function from the candidate control parameter combinations in the high-efficiency region, and generates control commands based on the optimal control parameter sequence.

[0064] The control correction module is used to send control commands to the power drive module of the reverse osmosis high-pressure pump motor, thereby driving the reverse osmosis high-pressure pump motor to operate according to the control commands; it collects the actual operating state vector and calculates the actual operating efficiency; based on the difference between the actual operating efficiency and the predicted efficiency corresponding to the control command, it corrects the corresponding efficiency value in the efficiency map, and at the same time, based on the error between the actual state vector and the corresponding predicted state vector, it corrects the parameters of the digital twin model.

[0065] The present invention, which adopts the above technical solution, has the following prominent features compared with the prior art:

[0066] Energy saving under all operating conditions: It maintains near-optimal operating efficiency under different loads, speeds and temperatures, effectively reducing the energy consumption of the reverse osmosis high-pressure pump motor.

[0067] Dynamic response optimization: The load prediction mechanism enables advance adjustment of control parameters, avoiding the problem of traditional feedback lag.

[0068] Overall performance improvement: Multi-objective optimization takes into account energy consumption, temperature rise, torque fluctuation and noise, so as to achieve stable operation and extended life.

[0069] Long-term adaptive optimization: The online learning module continuously corrects the efficiency map and digital twin model, enabling the control system to maintain optimal energy-saving performance over the long term.

[0070] Wide applicability: This method is applicable to a variety of application scenarios such as industrial fans, water pumps, and servo systems. Attached Figure Description

[0071] Figure 1 This is a schematic diagram illustrating the principle flow of the reverse osmosis high-pressure pump motor control method in an embodiment of the present invention;

[0072] Figure 2 This is a schematic diagram illustrating the principle of parameter acquisition in an embodiment of the present invention;

[0073] Figure 3 This is a schematic diagram illustrating the principle of data processing in an embodiment of the present invention;

[0074] Figure 4 This is a schematic diagram illustrating the structural principle of the digital twin model in this embodiment of the invention;

[0075] Figure 5 This is a schematic diagram illustrating the load trend prediction principle in an embodiment of the present invention;

[0076] Figure 6 This is a schematic diagram illustrating the principle of efficiency map construction in an embodiment of the present invention;

[0077] Figure 7 This is a schematic diagram of the efficiency map in an embodiment of the present invention. Detailed Implementation

[0078] The present invention will be further illustrated below with reference to specific embodiments. The purpose of this illustration is solely to provide a better understanding of the invention. Therefore, the examples given do not limit the scope of protection of the present invention.

[0079] like Figure 1 As shown in the figure, this embodiment presents a reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization, including the following steps:

[0080] S1. Collect real-time status parameters of the reverse osmosis high-pressure pump motor during operation and control parameters of the previous control cycle.

[0081] During implementation, see Figure 2 The basic state parameters collected may include: three-phase current of the reverse osmosis high-pressure pump motor; phase voltage; DC bus voltage; DC bus current; reverse osmosis high-pressure pump motor terminal voltage; rotor position; mechanical speed; winding temperature; reverse osmosis high-pressure pump motor housing temperature; ambient temperature; load torque; bearing temperature; vibration acceleration; noise equivalent value; cooling medium temperature; cooling medium flow rate; and power switching device temperature.

[0082] The control parameters include at least the d-axis current setpoint, q-axis current setpoint, flux linkage setpoint, and power device switching frequency. Depending on the specific controller architecture, the PWM duty cycle, voltage setpoint, or modulation ratio can also be incorporated into the control vector.

[0083] In applications where torque sensors are not installed, the load torque can be indirectly estimated based on the motor's mechanical motion equations, using electromagnetic torque, mechanical angular acceleration, moment of inertia, and frictional resistance torque.

[0084] S2. Perform time synchronization, outlier removal, filtering, and normalization on the collected parameters to obtain the state vector and control vector.

[0085] The operating data of the reverse osmosis high-pressure pump motor is easily affected by power device switching, electromagnetic interference, mechanical vibration, and sensor accuracy. If the raw data is used directly to update the digital twin model, it may cause frequent fluctuations in the model output and reduce control stability.

[0086] Therefore, see Figure 3 First, time synchronization of data from different sensors is required. For signals with different sampling frequencies, interpolation or resampling can be used to convert them to a unified time axis. Next, outlier identification is performed. When a sampled value exceeds a preset reasonable range, or its rate of change significantly exceeds the allowable range under normal operating conditions, it is marked as an outlier.

[0087] For outliers, interpolation between previous and subsequent sampling points, historical averages, or state observer estimates can be used. For signals that change rapidly, such as current, voltage, and rotational speed, low-pass filtering, Kalman filtering, or extended Kalman filtering can be used. For signals that change relatively slowly, such as temperature, weighted moving average filtering can be used.

[0088] Taking winding temperature as an example, the filtered temperature is calculated by weighting the current temperature and several historical sampled temperatures. The sum of the weights corresponding to each historical sampled value is 1.

[0089] , ;

[0090] represents the winding temperature after filtering at the k-th sampling time; represents the original winding temperature at the kr-th sampling time; and represents the weight of the r-th historical sample value.

[0091] Because temperature, torque, speed, vibration, and noise use different units, they cannot be directly input into the same evaluation function for calculation. Therefore, it is necessary to normalize the data for each type. The normalized parameters are generally between 0 and 1. Taking winding temperature as an example, its normalization result is obtained by subtracting the lowest temperature from the current temperature and then dividing by the difference between the highest and lowest temperatures.

[0092] Finally, for ease of subsequent calculations, the operating states of the reverse osmosis high-pressure pump motor are organized into a state vector. The state vector includes at least the d-axis current, q-axis current, mechanical angular velocity, electromagnetic torque, winding temperature, load torque, equivalent flux linkage, DC bus voltage, vibration equivalent value, and noise equivalent value, as specifically represented below:

[0093] ;

[0094] This represents the d-axis current at the k-th sampling time, which affects copper losses and excitation state; The q-axis current at the k-th sampling moment directly determines the electromagnetic torque and is the core control variable for speed regulation and energy saving. The mechanical angular velocity at the k-th sampling time reflects the rotational speed of the reverse osmosis high-pressure pump motor and is used for power calculation. This is also a key state for efficiency and control optimization; The electromagnetic torque at the k-th sampling moment, together with the d-axis / q-axis current and magnetic flux, determines the output torque of the reverse osmosis high-pressure pump motor; This represents the winding temperature at the k-th sampling time, which affects the winding resistance, copper loss, and temperature rise constraint. This represents the load torque at the k-th sampling time, reflecting the actual mechanical load and affecting the control objective and efficiency calculation; The equivalent flux linkage at the k-th sampling time reflects the magnetic flux generated by the permanent magnet or excitation winding, which affects torque and losses. The DC bus voltage at the k-th sampling time provides a driving voltage reference, affecting the voltage constraint and power output of the reverse osmosis high-pressure pump motor; The actual terminal voltage of the reverse osmosis high-pressure pump motor at the kth sampling time is used for power calculation and closed-loop control feedback. The equivalent value of vibration or noise at the k-th sampling time is used as one of the constraints in multi-objective optimization to ensure that the control process does not cause excessive vibration or noise.

[0095] The control vector includes at least the d-axis current setpoint, the q-axis current setpoint, the flux linkage setpoint, and the switching frequency of the power device, as shown below:

[0096] ;

[0097] This represents the target d-axis current generated based on efficiency and temperature rise constraints after the digital twin model predicts the given d-axis current. Its control function is to adjust the flux linkage distribution and excitation state, thereby affecting copper losses and torque. The q-axis current setpoint is used to generate electromagnetic torque and is a direct speed and torque control quantity. By adjusting the q-axis current, energy saving and load tracking can be achieved. This represents the setpoint of the magnetic flux, which can be used as a reference magnetic flux value for excitation adjustment in permanent magnet synchronous reverse osmosis high-pressure pump motors. This indicates the switching frequency of the power device (PWM switching frequency), which controls the switching rate of the PWM modulation and affects the voltage waveform, losses, and electromagnetic noise of the reverse osmosis high-pressure pump motor.

[0098] S3. Input the obtained state vector and control vector into the digital twin model of the reverse osmosis high-pressure pump motor for state synchronization, so that the current state of the digital twin model is synchronized with the actual reverse osmosis high-pressure pump motor, and predict the state vector at the next moment based on the discrete state update mapping of the digital twin model.

[0099] During implementation, see Figure 4 Digital twin models include electromagnetic sub-models, mechanical sub-models, loss sub-models, thermal sub-models, and efficiency sub-models; among them,

[0100] The electromagnet model is used to describe the relationship between d-axis current, q-axis current, d-axis voltage, q-axis voltage, rotational speed, magnetic flux linkage, and inductance. Current changes are predicted using the d-axis and q-axis voltage equations, respectively. The specific voltage equations are expressed as follows:

[0101] ;

[0102] ;

[0103] Represents the d-axis voltage; Indicates the phase resistance of the stator winding; Represents the d-axis current; This represents the voltage drop across the resistor along the d-axis. Indicates the d-axis inductance; Indicates electric angular velocity; Indicates the q-axis inductance; This represents the cross-coupling voltage generated by the q-axis current on the d-axis; Represents the q-axis voltage; Represents the q-axis current; This represents the voltage drop across the winding resistance in the q-axis direction; This indicates the magnetic flux linkage of a permanent magnet.

[0104] The electromagnetic torque variation can then be obtained based on the predicted d-axis and q-axis currents. Specifically, the electromagnetic torque is determined by the number of pole pairs, permanent magnet flux linkage, d-axis current, q-axis current, and d-axis and q-axis inductance.

[0105] ;

[0106] represents the electromagnetic torque of the reverse osmosis high-pressure pump motor; p represents the number of pole pairs of the reverse osmosis high-pressure pump motor.

[0107] The mechanical sub-model describes the relationship between electromagnetic torque, load torque, frictional resistance torque, and changes in mechanical speed. In practice, part of the electromagnetic torque is used to overcome the load torque, part to overcome mechanical friction, and the remainder is used to change the speed, thus obtaining the speed change, expressed as:

[0108] ;

[0109] It is the moment of inertia, which reflects the inertia of the rotor and rotating parts of the reverse osmosis high-pressure pump motor when accelerating or decelerating. The load torque is the resistance applied to the reverse osmosis high-pressure pump motor shaft by the mechanical load. Mechanical damping, proportional to rotational speed, represents energy loss caused by friction, wind resistance, or other mechanical resistance.

[0110] The loss sub-model is used to calculate copper loss, iron loss, mechanical loss, and additional losses, thereby deriving the total loss. In practice, the total loss of the reverse osmosis high-pressure pump motor includes copper loss, iron loss, mechanical loss, and additional losses, as shown below:

[0111] ;

[0112] This indicates the total loss of the reverse osmosis high-pressure pump motor; Indicates copper loss; Indicates iron loss; Indicates mechanical loss; This indicates additional losses.

[0113] For a permanent magnet synchronous reverse osmosis high-pressure pump motor that uses d-axis and q-axis current control, the copper loss is related to the winding resistance and the sum of the squares of the d-axis and q-axis currents:

[0114] .

[0115] Winding resistance increases with increasing temperature. Higher temperatures result in greater winding resistance and, for the same current, higher copper losses. The winding resistance is expressed as follows:

[0116] ;

[0117] This represents the stator winding resistance at the current winding temperature; This represents the stator winding resistance at the reference temperature; This indicates the temperature coefficient of resistance of the winding material.

[0118] Therefore, the same rotational speed and torque do not necessarily correspond to the same optimal control parameters at different temperatures. This is also an important reason why this invention introduces a digital twin model and a real-time efficiency map.

[0119] Iron loss mainly includes hysteresis loss and eddy current loss. Iron loss is related to the electrical frequency, magnetic flux density, and core material properties, as shown below:

[0120] ;

[0121] Indicates the hysteresis loss coefficient; This indicates the electrical frequency of the reverse osmosis high-pressure pump motor; α represents the amplitude or peak value of the magnetic flux density in the iron core; 'a' represents the hysteresis loss index. This represents the eddy current loss coefficient.

[0122] Mechanical losses mainly include bearing friction, viscous resistance, and wind resistance losses. Mechanical losses increase with increasing machine speed, as shown below:

[0123] ;

[0124] This represents the mechanical loss coefficient for the first-order term; This represents the mechanical loss coefficient of the quadratic term; This represents the mechanical loss coefficient for the cubic term.

[0125] The thermal model is used to describe the winding temperature change, which depends on the total losses of the reverse osmosis high-pressure pump motor, ambient temperature, equivalent thermal resistance, and equivalent heat capacity; it is expressed as follows:

[0126] ;

[0127] Indicates ambient temperature; This indicates the total loss of the reverse osmosis high-pressure pump motor; Indicates the equivalent thermal resistance; Indicates equivalent heat capacity; This indicates the winding temperature.

[0128] The efficiency sub-model is used to calculate the motor operating efficiency based on output power and total losses. Specifically, the output power of the reverse osmosis high-pressure pump motor is obtained by multiplying the electromagnetic torque by the mechanical angular velocity:

[0129] ;

[0130] The operating efficiency of the reverse osmosis high-pressure pump motor is obtained by dividing the output power by the sum of the output power and the total losses:

[0131] .

[0132] In practice, to facilitate calculations in each sampling period, the digital twin model can be converted into a discrete state update mapping form:

[0133] ;

[0134] This represents the state vector predicted by the digital twin model at the (k+1)th sampling time. This represents the current state vector after estimation or correction at the k-th sampling time. This represents the control vector for the k-th control cycle; This represents the external disturbance vector at the k-th sampling time. This represents the model parameter vector at the k-th sampling time. This represents the discrete state update mapping.

[0135] The digital twin model outputs predicted states for several future cycles in each control cycle, including predicted current, predicted speed, predicted torque, predicted temperature, predicted losses, and predicted efficiency.

[0136] S4. Based on historical load data, current load torque, and load change rate, predict the load trend within future control cycles; whereby the current load torque is estimated in real time by a digital twin model.

[0137] In practice, the load on the reverse osmosis high-pressure pump motor, when operating in the fan, water pump, compressor, and servo system, typically varies over time. Adjusting parameters solely based on the current load can easily lead to response lag. (See also...) Figure 5 Load trend prediction can predict future load based on historical load, current load, load change rate, and operating scenario characteristics.

[0138] In a specific implementation, a sliding window weighted method can be used to predict load trends:

[0139] ;

[0140] in, This represents the load torque predicted at the (k+1)th sampling time. This represents the actual load torque at the kr-th sampling time. This represents the corresponding weighting coefficient; n represents the prediction window size, which can be 10 control periods.

[0141] Furthermore, in application scenarios where the load change rate exceeds the set conditions, the prediction window size is shortened by a set ratio, and the weight coefficients are allocated according to the principle that the weight coefficients are higher the closer to the k+1 sampling time; in application scenarios where the load change rate is lower than the set conditions, the prediction window size is extended by a set ratio.

[0142] In addition to the sliding window weighted method, recursive least squares, time series models, neural networks, or other prediction methods can also be used.

[0143] S5. Based on the d-axis current, q-axis current, rotational speed, and winding temperature under the current state vector, calculate the total loss value under the current state using the digital twin model, and predict the winding temperature rise in future control cycles based on the total loss value.

[0144] S6. Based on the current state digital twin model, the predicted load trend, and the predicted winding temperature rise, generate multiple candidate control parameter combinations, and use the digital twin model to calculate the predicted efficiency, predicted temperature rise, predicted torque pulsation, and predicted vibration noise corresponding to each candidate parameter combination.

[0145] In implementation, the candidate control parameters, i.e., the control vector, must include at least the d-axis current setpoint, the q-axis current setpoint, the flux linkage setpoint, and the switching frequency of the power device. The search range of the candidate control parameters is not a fixed constant, but a dynamic boundary determined by the current operating state, the predicted load trend, and the predicted temperature rise. The range determination process is explained below for each of the four parameters:

[0146] I. d-axis current setpoint

[0147] Step 1: Determine hard constraint boundaries from the current state

[0148] The range of the d-axis current is primarily limited by the inverter's maximum allowable current. The sum of the squares of the d-axis and q-axis currents cannot exceed the square of the inverter's maximum current; therefore, as the q-axis current increases, the usable range of the d-axis current is compressed.

[0149] Meanwhile, motor speed also directly affects the range of d-axis current. When the motor operates in the low-speed range, the back electromotive force (EMF) is low, the inverter has sufficient voltage margin, and the d-axis current can be freely selected within a wide range, typically extending in both positive and negative directions from zero. However, when the motor enters the high-speed range, the back EMF increases, and the inverter output voltage approaches its limit. At this time, in order to maintain high-speed operation, the system must apply a negative d-axis current (i.e., a magnetic weakening current) to weaken the permanent magnet's magnetic field and reduce the back EMF. Therefore, the higher the speed, the lower the lower limit (negative limit) of the d-axis current, and the overall range of the d-axis current will shift in the negative direction.

[0150] Furthermore, the d-axis current cannot increase indefinitely in the negative direction, because an excessively large negative d-axis current will produce a demagnetizing effect, which may cause irreversible damage to the permanent magnet. This demagnetizing current limit constitutes the absolute lower limit of the d-axis current.

[0151] Step 2: Adjust the center of the range based on the predicted load trend.

[0152] Load trends affect the offset of the d-axis current range in both positive and negative directions. When an increase in load is predicted, a larger q-axis current is needed to output torque. To allow sufficient headroom for the q-axis current within the current limit circle, the range of the d-axis current needs to be adjusted appropriately. Conversely, when a decrease in load is predicted, the d-axis current has more room to move towards zero or positive values ​​because there is no need to reserve excessive capacity for the q-axis current.

[0153] Step 3: Correcting the boundary based on predicted temperature rise

[0154] The winding temperature directly affects the tolerance of copper losses. When the predicted temperature is high, it is necessary to prioritize reducing copper losses, and the d-axis current has a direct impact on copper losses. A larger d-axis current (whether positive or negative) will increase copper losses. Therefore, when the temperature is high, the search range of the d-axis current needs to be appropriately narrowed to avoid generating excessively large d-axis current components. If the temperature is low, the range of the d-axis current can be appropriately widened, allowing for the search for the optimal efficiency point within a wider range.

[0155] II. q-axis current setpoint

[0156] Step 1: Determine the center value and upper limit based on the current status and load demand.

[0157] The range of the q-axis current is primarily determined by the inverter's maximum allowable current. Given the current d-axis current, the physical upper limit of the q-axis current is determined by the current limit circle boundary. Simultaneously, the q-axis current is also constrained by the voltage limit ellipse; in the high-speed region, even if the current does not reach its limit, the voltage limit may further shrink the maximum usable value of the q-axis current.

[0158] The required output torque serves as the central basis for determining the q-axis current range. Since the electromagnetic torque is approximately proportional to the q-axis current, a necessary q-axis current value can be calculated from the current torque command; this value becomes the center point of the search range. The q-axis current search range fluctuates around this center point.

[0159] Step 2: Adjust the range and position based on the predicted load trend.

[0160] Load trends are the most important predictive basis for determining the q-axis current range. When the load is predicted to increase, it indicates a greater torque output is expected in the future. Therefore, the center point of the q-axis current search range is moved upwards (in the direction of increase), while ensuring that the upper limit of the search range can cover the maximum possible future load demand. When the load is predicted to decrease, the center point of the search range is moved downwards to avoid wasting computational resources in the high q-axis current region and to better reduce copper losses.

[0161] Step 3: Adjust the upper limit based on the predicted temperature rise

[0162] Winding temperature has a significant limiting effect on the q-axis current range. The q-axis current is the primary current component generating torque and also the main source of copper losses. When the predicted winding temperature approaches the maximum allowable temperature, the upper limit of the q-axis current must be limited, even if the physical current limit circle allows for a larger current; temperature constraints will compress this. This limitation is essentially a trade-off between temperature safety and torque output: the higher the temperature, the lower the upper limit of the q-axis current, tending to sacrifice some torque capability for temperature safety. Conversely, when the predicted temperature is lower, the upper limit of the q-axis current is determined by physical limits, resulting in a wider range.

[0163] III. Magnet flux setpoint

[0164] Step 1: Determine hard constraint boundaries from the current state

[0165] The upper limit of the given flux linkage value is determined by the magnetic saturation characteristics of the core material. When the flux linkage exceeds this value, the magnetic flux density of the core enters the saturation region. Continuing to increase the flux linkage will not only fail to effectively increase the torque, but will also lead to a sharp increase in iron losses, thus constituting a hard upper limit.

[0166] At high speeds, due to the contraction of the voltage limit ellipse, the flux linkage setpoint must be reduced accordingly; otherwise, the required terminal voltage will exceed the DC bus voltage that the inverter can provide. Therefore, the higher the speed, the lower the available upper limit of the flux linkage setpoint.

[0167] Step 2: Adjusting the range based on predicted load trends

[0168] The flux linkage setpoint directly affects the motor's output torque capability. When the predicted load increases, a larger torque output is required, necessitating a higher flux linkage setpoint to fully utilize the reluctance torque. When the predicted load decreases, the flux linkage can be appropriately reduced to decrease iron losses and improve light-load efficiency.

[0169] Step 3: Correcting the boundary based on predicted temperature rise

[0170] The flux linkage setpoint is directly related to iron losses. A higher flux linkage setpoint means a higher magnetic flux density, which leads to higher iron losses (especially hysteresis losses), and consequently, a greater temperature rise. When the predicted winding temperature is higher than expected, the upper limit of the flux linkage setpoint needs to be reduced to avoid excessive iron losses from excessive flux linkage, which would further increase the temperature rise. When the temperature is within a safe range, the range of the flux linkage setpoint is determined by the magnetic saturation characteristics and the rotational speed, and is not constrained by temperature limits.

[0171] IV. Switching Frequency of Power Devices

[0172] Step 1: Determine hard constraint boundaries from the current state

[0173] The lower limit of the switching frequency is determined by both current harmonic distortion requirements and audible noise requirements. Too low a frequency will lead to increased current ripple, increased harmonic losses, and may generate audible noise. The upper limit of the switching frequency is determined by the switching losses and heat dissipation capacity of the power device. The higher the frequency, the more switching operations per unit time, the greater the switching losses, and the higher the temperature rise of the power device.

[0174] Step 2: Adjusting the range based on predicted load trends

[0175] The load size directly affects the magnitude of switching losses. When the load current is predicted to increase, the switching losses themselves will increase (switching losses are proportional to current), and in this case, a lower switching frequency is preferred to control the total losses. When the load is predicted to decrease, the switching losses are relatively small, and the switching frequency can be appropriately increased to obtain better current waveform quality and lower harmonic losses.

[0176] Step 3: Adjusting the range based on the predicted temperature rise

[0177] Switching losses in power devices are a significant source of temperature rise. When the predicted motor winding temperature or power device temperature is higher than expected, it is necessary to lower the upper limit of the switching frequency, prioritizing temperature rise suppression by reducing the switching frequency. The higher the temperature, the lower the maximum permissible switching frequency. When the temperature is within a safe range, the search range for the switching frequency is determined by the physical capabilities of the power device and the load requirements.

[0178] The process of determining the search range for the above four parameters follows a unified logical framework: the hard constraint boundary is determined by the current physical state (speed, bus voltage, temperature), the range center is determined by the predicted load trend, and the range contraction is determined by the predicted temperature rise trend. All three are indispensable, jointly ensuring that the candidate parameter combination is both within the physical safety range and covers future operating conditions, while avoiding failure due to excessive temperature. Furthermore, technicians can also directly limit the fixed parameter range corresponding to each parameter based on their experience with actual operating conditions.

[0179] Each candidate control parameter combination contains control parameters for multiple subsequent control cycles. During generation, multiple candidate control parameter combinations can be randomly generated within a defined parameter range; alternatively, they can be discretized within their respective parameter ranges using a set step size. The discrete values ​​of the four parameters are then combined in all possible permutations to generate the complete set of candidate parameters. For example, the sampling step size can be set as follows: d-axis current step size 0.5~2A, q-axis current step size 0.5~2A, flux linkage step size 0.01~0.02Wb, and switching frequency discretely selected from set levels (e.g., 2kHz, 4kHz, 6kHz, 8kHz, 10kHz). If the number of candidate control parameter combinations is too large, causing the computational load to exceed the controller's real-time requirements, a two-stage search strategy (coarse-to-fine) can be adopted: the first stage uses a large step size (e.g., d / q-axis current step size of 2A) to quickly locate the parameter region with lower comprehensive objective function; the second stage uses a small step size (e.g., d / q-axis current step size of 0.5A) within a local small range to finely search for the optimal value.

[0180] Finally, the digital twin model was used to calculate the predicted efficiency, predicted temperature rise, predicted torque pulsation, predicted vibration noise, and control parameter variation for each candidate combination.

[0181] S7. Using the d-axis current setpoint, q-axis current setpoint, flux linkage setpoint, and power device switching frequency in the candidate control parameter combination as independent variables and the predicted efficiency corresponding to the digital twin model as the dependent variable, construct a real-time efficiency map, and classify the candidate control parameter combination with a predicted efficiency not lower than a set threshold into the high-efficiency region under the premise of meeting safety constraints.

[0182] During implementation, see Figure 6Under the premise of meeting the safety constraints of stator current, terminal voltage, winding temperature, electromagnetic torque, and mechanical speed, candidate control parameter combinations with predicted efficiencies not lower than a set threshold are classified into the high-efficiency region; otherwise, they are discarded. The high-efficiency region is not fixed; see [link to relevant documentation]. Figure 7 When the reverse osmosis high-pressure pump motor temperature rises, the load changes, or the bus voltage fluctuates, the location and range of the high-efficiency region will change accordingly. The set threshold corresponding to the predicted efficiency can be set according to the motor type and application scenario, with a typical value of 85% to 95%. Candidate control parameter combinations below this value are not included in the high-efficiency region.

[0183] S8. Construct a comprehensive objective function, which includes efficiency optimization index, temperature rise constraint index, torque pulsation index, vibration and noise index, and control smoothness index. At the same time, dynamically adjust the weight of each index according to the current operating conditions. Solve for the optimal control parameter sequence that minimizes the comprehensive objective function among the candidate control parameter combinations in the high-efficiency region, and generate control commands based on the optimal control parameter sequence.

[0184] Multi-objective predictive optimization is used to evaluate candidate control parameter combinations over several future control cycles and output the optimal control parameter sequence for those cycles. This invention does not only pursue maximum efficiency but also considers temperature rise, torque ripple, vibration noise, and control parameter smoothness. This avoids excessively high temperatures, large torque fluctuations, or frequent control command jumps caused by prioritizing short-term efficiency improvements. Specifically, the comprehensive objective function is expressed as:

[0185] ;

[0186] in, Indicates the length of the prediction time domain; These are the weights of each indicator; Indicates efficiency optimization metrics; Indicates the temperature rise constraint index; This indicates the torque ripple index; Indicates vibration and noise levels; This indicates the smoothness control index.

[0187] The efficiency optimization index is used to evaluate the prediction efficiency corresponding to the candidate control parameter combination. The calculation principle is as follows:

[0188] ;

[0189] in, This represents the efficiency optimization index for the k-th control cycle. Indicates the predictive efficiency of the digital twin model;

[0190] The temperature rise constraint index is used to evaluate the extent to which the maximum allowable temperature of the winding is predicted. The calculation principle is as follows:

[0191] ;

[0192] in, This indicates the temperature rise constraint index. This represents the winding temperature predicted by the digital twin model for the k-th control cycle. The safe winding temperature threshold is generally taken as 80% to 85% of the maximum allowable temperature corresponding to the motor insulation class. Indicates the maximum permissible temperature;

[0193] The torque ripple index is used to evaluate the variation in predicted electromagnetic torque between adjacent control cycles. The calculation principle is as follows:

[0194] ;

[0195] in, Indicates the torque ripple value. This represents the electromagnetic torque predicted in the k-th control cycle. This represents the electromagnetic torque predicted over k-1 control cycles. Indicates the torque reference value;

[0196] Vibration and noise levels are used to evaluate the vibration and noise levels of reverse osmosis high-pressure pump motors. The calculation principle is as follows:

[0197] ;

[0198] in, Indicates vibration and noise index, This represents the normalized vibration acceleration. This represents the normalized noise equivalent value. The noise equivalent value can be based on the historical measured values ​​of vibration and noise of the motor under different operating conditions (speed, load, etc.). When evaluating a candidate combination of control parameters, historical data with similar operating conditions can be found as the equivalent value. Alternatively, an empirical mapping model (such as lookup table, regression model, etc.) between vibration and noise and operating parameters (speed, torque, current, etc.) can be established. The operating parameters of the candidate combination are input, and the predicted vibration and noise is output.

[0199] The control smoothness index is used to evaluate the magnitude of change in the control parameter vector between two adjacent control cycles in the current candidate control parameter combination. The calculation principle is as follows:

[0200] ;

[0201] in, This indicates the smoothness control index. This represents the control parameter vector for the k-th control cycle. This represents the control parameter vector for k-1 control cycles. This represents the weight matrix.

[0202] The importance of each evaluation indicator changes dynamically under different working conditions. When the weight of a certain optimization objective is increased, the weights of all other indicators are reduced synchronously according to their original proportions, achieving fully automatic linkage updates and ensuring that the total weight remains unchanged. The process of dynamically adjusting the weights of each indicator includes:

[0203] All weights satisfy the normalization constraint: ;

[0204] Based on the current operating conditions, determine the target set S that needs to be prioritized for enhancement: when the predicted winding temperature is below the safety threshold, include efficiency optimization indicators in the target set; when the predicted winding temperature is close to the maximum allowable temperature, include temperature rise constraint indicators in the target set; when the load change rate exceeds the set threshold, include torque ripple indicators and control smoothness indicators in the target set; when vibration or noise exceeds the set threshold, include vibration and noise indicators in the target set.

[0205] Calculate the total weight of enhancements in the target set. and the total weight of non-enhancing terms outside the target set. ;

[0206] Determine the amplification factor based on the degree of deviation from the current operating conditions. The weights within the target set are then amplified and updated. ;

[0207] Calculate the overall reduction factor for the non-enhancing terms: The weights of non-enhancing terms are reduced using the same reduction factor: ;

[0208] Perform boundary constraint checks and set a minimum value for each weight. If the adjusted weight of a certain item is lower than the set minimum value, then the weight will be locked. . The value can be in the range of 0.02 to 0.05 to prevent a single optimization objective from being completely discarded during dynamic weight adjustment.

[0209] S9. Send the control command to the power drive module of the reverse osmosis high-pressure pump motor, thereby driving the reverse osmosis high-pressure pump motor to run according to the control command.

[0210] In practice, to provide real-time control optimization, only the first set of control parameters in the optimal control parameter sequence can be sent to the power drive module as the control quality. After entering the next control cycle, the operating data is re-acquired, the digital twin model is re-updated, the load and temperature rise are re-predicted, the real-time efficiency map is recalculated, and the optimal control parameter sequence is re-solved.

[0211] S10. Collect the current running actual state vector and calculate the actual running efficiency;

[0212] S11. Based on the difference between the actual operating efficiency and the predicted efficiency corresponding to the control command, the corresponding efficiency value in the efficiency map is corrected. At the same time, based on the error between the actual state vector and the corresponding predicted state vector, the parameters of the digital twin model are corrected.

[0213] During implementation, the parameters of the reverse osmosis high-pressure pump motor will change with temperature, operating time, mechanical wear, and component aging. If the initial model is always used, the prediction results of the digital twin model will gradually deviate from the actual situation; therefore, it is necessary to continuously adjust the model based on the error between the actual operating data and the predicted data.

[0214] The difference between the actual operating efficiency and the predicted efficiency corresponding to the control commands is expressed as:

[0215] ;

[0216] Indicates actual operating efficiency; This indicates the prediction efficiency.

[0217] The updated efficiency map value is determined by the original efficiency map value and the efficiency prediction error:

[0218] ;

[0219] This represents the updated efficiency map value; This represents the efficiency map value before the update. This represents the learning coefficient, which ranges from 0.01 to 0.1. A larger value results in faster correction but may cause oscillations, while a smaller value results in more stable convergence but a slower response. This represents the efficiency prediction error for the current sampling period.

[0220] The parameters of the digital twin model are adjusted based on the state prediction error, as shown below:

[0221] ;

[0222] This represents the updated parameters of the digital twin model; This represents the parameters of the digital twin model before the update; The learning gain matrix can be a diagonal matrix, diag(0.01, 0.01, ..., 0.01), and can be adaptively adjusted according to the convergence speed during actual operation. This represents the state prediction error.

[0223] The present invention also includes an abnormal operating condition protection process: when overcurrent, overvoltage, undervoltage, overtemperature, stalled rotor, sudden load change, communication abnormality or sensor failure is detected, the execution of energy-saving optimization results is suspended and the system switches to safety control mode.

[0224] Safety control modes may include current limiting, torque limiting, speed reduction, reducing flux setpoint, reducing switching frequency, stopping online learning, outputting alarm signals, or stopping the reverse osmosis high-pressure pump motor.

[0225] Once the abnormal operating conditions are eliminated and no new abnormal operating conditions occur within the set time, the recovery conditions are met. The digital twin model is then reinitialized, and multi-objective predictive optimization control is gradually restored.

[0226] This invention also discloses a system for implementing the above-mentioned reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization, comprising:

[0227] The data acquisition and processing module is used to acquire real-time status parameters and control parameters from the previous control cycle during the operation of the reverse osmosis high-pressure pump motor; it performs time synchronization, outlier removal, filtering, and normalization on the acquired parameters to obtain the status vector and control vector.

[0228] The digital twin module is used to input the obtained state vector and control vector into the digital twin model of the reverse osmosis high-pressure pump motor for state synchronization, so that the current state of the digital twin model is synchronized with the actual reverse osmosis high-pressure pump motor, and the state vector of the next moment is predicted based on the discrete state update mapping of the digital twin model.

[0229] The load trend prediction module is used to predict the load trend in future control cycles based on historical load data, current load torque, and load change rate.

[0230] The winding temperature rise prediction module is used to calculate the total loss value under the current state based on the d-axis current, q-axis current, rotational speed and winding temperature under the current state vector using a digital twin model, and to predict the winding temperature rise in future control cycles based on the total loss value.

[0231] The parameter generation module is used to generate multiple candidate control parameter combinations based on the current state digital twin model, the predicted load trend, and the predicted winding temperature rise. The digital twin model is used to calculate the predicted efficiency, predicted temperature rise, predicted torque pulsation, and predicted vibration noise corresponding to each candidate parameter combination.

[0232] The efficiency map module is used to construct a real-time efficiency map with the d-axis current setpoint, q-axis current setpoint, flux linkage setpoint and power device switching frequency in the candidate control parameter combination as independent variables and the predicted efficiency corresponding to the digital twin model as the dependent variable. Under the premise of meeting safety constraints, candidate control parameter combinations with predicted efficiency not lower than a set threshold are classified into the high-efficiency region.

[0233] The multi-objective optimization module is used to construct a comprehensive objective function, which includes efficiency optimization index, temperature rise constraint index, torque ripple index, vibration and noise index, and control smoothness index. At the same time, the weights of each index are dynamically adjusted according to the current operating conditions. The module finds the optimal control parameter sequence that minimizes the comprehensive objective function from the candidate control parameter combinations in the high-efficiency region, and generates control commands based on the optimal control parameter sequence.

[0234] The control correction module is used to send control commands to the power drive module of the reverse osmosis high-pressure pump motor, thereby driving the reverse osmosis high-pressure pump motor to operate according to the control commands; it collects the actual operating state vector and calculates the actual operating efficiency; based on the difference between the actual operating efficiency and the predicted efficiency corresponding to the control command, it corrects the corresponding efficiency value in the efficiency map, and at the same time, based on the error between the actual state vector and the corresponding predicted state vector, it corrects the parameters of the digital twin model.

[0235] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of the invention. All equivalent changes made based on the description and drawings of the present invention are included within the scope of the present invention.

Claims

1. A reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization, characterized in that, Includes the following steps: S1. Collect real-time status parameters of the reverse osmosis high-pressure pump motor during operation and control parameters of the previous control cycle; S2. Perform time synchronization, outlier removal, filtering and normalization on the collected parameters to obtain the state vector and control vector; S3. Input the obtained state vector and control vector into the digital twin model of the reverse osmosis high-pressure pump motor for state synchronization, so that the current state of the digital twin model is synchronized with the actual reverse osmosis high-pressure pump motor, and predict the state vector at the next moment based on the discrete state update mapping of the digital twin model. S4. Based on historical load data, current load torque, and load change rate, predict the load trend within future control cycles; S5. Based on the d-axis current, q-axis current, rotational speed, and winding temperature under the current state vector, calculate the total loss value under the current state using the digital twin model, and predict the winding temperature rise in future control cycles based on the total loss value. S6. Based on the digital twin model of the current state, the predicted load trend and the predicted winding temperature rise, generate multiple candidate control parameter combinations, and use the digital twin model to calculate the predicted efficiency, predicted temperature rise, predicted torque pulsation and predicted vibration noise corresponding to each candidate parameter combination. S7. Using the d-axis current setpoint, q-axis current setpoint, flux linkage setpoint, and power device switching frequency in the candidate control parameter combination as independent variables and the predicted efficiency corresponding to the digital twin model as the dependent variable, construct a real-time efficiency map, and classify the candidate control parameter combination with a predicted efficiency not lower than a set threshold into the high efficiency region under the premise of meeting safety constraints. S8. Construct a comprehensive objective function, which includes efficiency optimization index, temperature rise constraint index, torque ripple index, vibration and noise index, and control smoothness index. At the same time, dynamically adjust the weight of each index according to the current operating conditions. Solve for the optimal control parameter sequence that minimizes the comprehensive objective function from the candidate control parameter combinations in the high-efficiency region, and generate control commands based on the optimal control parameter sequence. S9. Send the control command to the power drive module of the reverse osmosis high-pressure pump motor, thereby driving the reverse osmosis high-pressure pump motor to run according to the control command; S10. Collect the current running actual state vector and calculate the actual running efficiency; S11. Based on the difference between the actual operating efficiency and the predicted efficiency corresponding to the control command, the corresponding efficiency value in the efficiency map is corrected. At the same time, based on the error between the actual state vector and the corresponding predicted state vector, the parameters of the digital twin model are corrected.

2. The reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization according to claim 1, characterized in that, Digital twin models include electromagnetic sub-models, mechanical sub-models, loss sub-models, thermal sub-models, and efficiency sub-models; among them, The electromagnet model is used to describe the relationship between d-axis current, q-axis current, d-axis voltage, q-axis voltage, rotational speed, magnetic flux, and inductance. The d-axis and q-axis voltage equations are used to predict current changes, respectively. The mechanical sub-model is used to describe the relationship between electromagnetic torque, load torque, frictional resistance torque, and changes in mechanical speed; The loss sub-model is used to calculate copper loss, iron loss, mechanical loss, and additional loss, and then to obtain the total loss; The thermal model is used to describe the winding temperature change, which depends on the total losses of the reverse osmosis high-pressure pump motor, ambient temperature, equivalent thermal resistance, and equivalent heat capacity. The efficiency sub-model is used to calculate the motor operating efficiency based on the output power and total losses.

3. The reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization according to claim 1, characterized in that, S4 uses a sliding window weighted method to predict load trends: ; in, This represents the load torque predicted at the (k+1)th sampling time. This represents the actual load torque at the kr-th sampling time. Indicates the corresponding weight coefficient; n represents the prediction window size; In application scenarios where the load change rate exceeds the set conditions, the prediction window size is shortened by a set ratio, and the weight coefficients are allocated according to the principle that the weight coefficients are higher the closer to the k+1 sampling time. In application scenarios where the load change rate is lower than the set conditions, the prediction window size is extended by a set ratio.

4. The reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization according to claim 1, characterized in that, The comprehensive objective function in S8 is expressed as follows: ; in, Indicates the length of the prediction time domain; These are the weights of each indicator; Indicates efficiency optimization metrics; Indicates the temperature rise constraint index; This indicates the torque ripple index; Indicates vibration and noise levels; This indicates the smoothness control index.

5. The reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization according to claim 4, characterized in that, The efficiency optimization index is used to evaluate the prediction efficiency corresponding to the candidate control parameter combination. The calculation principle is as follows: ; in, This represents the efficiency optimization index for the k-th control cycle. Indicates the predictive efficiency of the digital twin model; The temperature rise constraint index is used to evaluate the extent to which the maximum allowable temperature of the winding is predicted. The calculation principle is as follows: ; in, This indicates the temperature rise constraint index. This represents the winding temperature predicted by the digital twin model for the k-th control cycle. Indicates the safety winding temperature threshold. Indicates the maximum permissible temperature; The torque ripple index is used to evaluate the variation in predicted electromagnetic torque between adjacent control cycles. The calculation principle is as follows: ; in, Indicates the torque ripple value. This represents the electromagnetic torque predicted in the k-th control cycle. This represents the electromagnetic torque predicted over k-1 control cycles. Indicates the torque reference value; Vibration and noise levels are used to evaluate the vibration and noise levels of reverse osmosis high-pressure pump motors. The calculation principle is as follows: ; in, Indicates vibration and noise index, This represents the normalized vibration acceleration. This represents the normalized noise equivalent value; The control smoothness index is used to evaluate the magnitude of change in the control parameter vector between two adjacent control cycles in the current candidate control parameter combination. The calculation principle is as follows: ; in, This indicates the control smoothness index. This represents the control parameter vector for the k-th control cycle. This represents the control parameter vector for k-1 control cycles. This represents the weight matrix.

6. The reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization according to claim 1, characterized in that, The process of dynamically adjusting the weights of each indicator based on the current operating conditions in S8 includes: All weights satisfy the normalization constraint: ; Based on the current operating conditions, determine the target set S that needs to be prioritized for enhancement: when the predicted winding temperature is below the safety threshold, include efficiency optimization indicators in the target set; when the predicted winding temperature is close to the maximum allowable temperature, include temperature rise constraint indicators in the target set; when the load change rate exceeds the set threshold, include torque ripple indicators and control smoothness indicators in the target set; when vibration or noise exceeds the set threshold, include vibration and noise indicators in the target set. Calculate the total weight of enhancements in the target set. and the total weight of non-enhancing terms outside the target set. ; Determine the amplification factor based on the degree of deviation from the current operating conditions. The weights within the target set are then amplified and updated. ; Calculate the overall reduction factor for the non-enhancing terms: The weights of non-enhancing terms are reduced using the same reduction factor: ; Perform boundary constraint checks and set a minimum value for each weight. If the adjusted weight of a certain item is lower than the set minimum value, then the weight will be locked. .

7. The reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization according to claim 1, characterized in that, Safety constraints in S7 include stator current constraints, terminal voltage constraints, winding temperature constraints, electromagnetic torque constraints, and mechanical speed constraints.

8. A system for implementing the reverse osmosis high-pressure pump motor control method based on digital twin and multi-objective optimization as described in any one of claims 1-7, characterized in that, include: The data acquisition and processing module is used to acquire real-time status parameters of the reverse osmosis high-pressure pump motor during operation and control parameters of the previous control cycle. The collected parameters are time-synchronized, outlier-removed, filtered, and normalized to obtain the state vector and control vector. The digital twin module is used to input the obtained state vector and control vector into the digital twin model of the reverse osmosis high-pressure pump motor for state synchronization, so that the current state of the digital twin model is synchronized with the actual reverse osmosis high-pressure pump motor, and the state vector of the next moment is predicted based on the discrete state update mapping of the digital twin model. The load trend prediction module is used to predict the load trend in future control cycles based on historical load data, current load torque, and load change rate. The winding temperature rise prediction module is used to calculate the total loss value under the current state based on the d-axis current, q-axis current, rotational speed and winding temperature under the current state vector, and predict the winding temperature rise in future control cycles based on the total loss value. The parameter generation module is used to generate multiple candidate control parameter combinations based on the digital twin model of the current state, the predicted load trend, and the predicted winding temperature rise. The digital twin model is used to calculate the predicted efficiency, predicted temperature rise, predicted torque pulsation, and predicted vibration noise corresponding to each candidate parameter combination. The efficiency map module is used to construct a real-time efficiency map with the d-axis current setpoint, q-axis current setpoint, flux linkage setpoint and power device switching frequency in the candidate control parameter combination as independent variables and the predicted efficiency corresponding to the digital twin model as the dependent variable. Under the premise of meeting safety constraints, candidate control parameter combinations with predicted efficiency not lower than a set threshold are classified into the high-efficiency region. The multi-objective optimization module is used to construct a comprehensive objective function, which includes efficiency optimization index, temperature rise constraint index, torque ripple index, vibration and noise index, and control smoothness index. At the same time, the weights of each index are dynamically adjusted according to the current operating conditions. The module finds the optimal control parameter sequence that minimizes the comprehensive objective function from the candidate control parameter combinations in the high-efficiency region, and generates control commands based on the optimal control parameter sequence. The control correction module is used to send control commands to the power drive module of the reverse osmosis high-pressure pump motor, thereby driving the reverse osmosis high-pressure pump motor to operate according to the control commands; it also collects the actual operating state vector and calculates the actual operating efficiency. Based on the difference between the actual operating efficiency and the predicted efficiency corresponding to the control commands, the corresponding efficiency values ​​in the efficiency map are corrected. At the same time, based on the error between the actual state vector and the corresponding predicted state vector, the parameters of the digital twin model are corrected.