High efficiency propeller power management optimization method and system

Abstract

The application relates to the technical field of data processing, and discloses a high-efficiency propeller power management optimization method and system. The method comprises the following steps: judging load mutation in advance and distributing power in advance through magnetic coupling slip characteristics extraction; realizing power mutual feeding among propellers through an energy transfer channel established by a multi-propeller power state matrix; inversely calculating a bearing friction coefficient through a power balance method and calculating compensation power by using a nonlinear mapping self-learning algorithm; superimposing pre-distributed power, transferred power and compensation power to generate an output instruction, realizing predictive power distribution, energy recycling and adaptive optimization management. The application realizes predictive power distribution of underwater propellers, bidirectional recycling of energy and adaptive optimization management in multiple working conditions.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a high-efficiency thruster power management optimization method and system. Background Technology

[0002] As the core power unit of underwater vehicles, the power management technology of underwater thrusters directly affects the system's energy efficiency and operational reliability. Existing underwater thruster power management methods mainly adopt a fixed threshold protection strategy. By monitoring parameters such as motor temperature and current, a protection mechanism is triggered to reduce output power when these parameters exceed preset thresholds. At the same time, external capacitor banks are used to passively absorb the back electromotive force energy generated during motor operation. In multi-thrust systems, each thruster is independently powered and unrelated to the others. Performance degradation is addressed through regular inspection and replacement maintenance after bearing wear.

[0003] However, existing technologies have the following shortcomings: First, the overheat protection strategy uses a fixed power reduction ratio, uniformly reducing the output power by 50% when the motor temperature exceeds the threshold. This fixed ratio cannot be finely adjusted according to the actual degree of temperature exceeding the threshold, load status, and operating conditions. This results in excessive power reduction when there is slight overheating, affecting propulsion efficiency, and insufficient power reduction when there is severe overheating, failing to effectively protect the equipment. Second, the back electromotive force energy is passively absorbed by a large-capacity capacitor bank configured at both ends of the power supply. The capacitor capacity needs to be manually selected according to different power supply types, and the absorbed energy is ultimately dissipated as heat and cannot be recovered, resulting in energy waste. Third, bearing wear in muddy and watery environments leads to increased frictional resistance. Existing solutions only passively increase power or replace the bearing after the thrust decreases significantly due to severe bearing wear. They cannot actively compensate for power in the early stages of wear to maintain constant thrust output.

[0004] Further analysis revealed that the fundamental flaw of the fixed threshold protection strategy lies in its lack of ability to anticipate load surges, allowing only a passive response after a fault has occurred. In contrast, the phase difference of the magnet array, speed slip, and load torque in the magnetically coupled transmission mechanism exhibit characteristic patterns 0.1-0.3 seconds before load surges. If these characteristics can be extracted and a predictive model established, some power can be transferred to the buffer module before a fault occurs, rather than passively absorbing it after back EMF generation. Furthermore, the power state, buffer energy, thermal margin, and load level of each thruster in a multi-thrust system differ in real time. If a distributed power state matrix can be constructed and the mutual feedback priority between thrusters calculated, a bidirectional energy transfer channel can be established to redistribute buffer energy, transferring the stored energy of high-buffer-energy thrusters to high-load thrusters, instead of allowing energy to dissipate independently within each thruster's buffer module. Furthermore, the frictional power consumption of the bearing can be separated from the input power and the friction coefficient can be calculated by power balance method. If a nonlinear mapping relationship between the friction coefficient and power compensation can be established, the power can be fed forward to compensate when the wear degree is low. With the help of self-learning algorithm to optimize the compensation parameters, the thrust can be kept stable throughout the wear process, rather than passively adjusting when the wear is severe. Summary of the Invention

[0005] This application provides a high-efficiency thruster power management optimization method and system. By constructing a magnetic coupling transmission slip feature extraction and power pre-allocation mechanism, a multi-thruster distributed power state matrix and energy mutual feedback channel, and a bearing wear degree back calculation and feedforward compensation self-learning algorithm, it solves the technical problems in the prior art, such as fixed and single power management strategy that cannot be predicted in advance, passive dissipation of back electromotive force energy that cannot be recovered and utilized, isolated operation of multi-thruster power that cannot be coordinated and optimized, and passive adjustment after bearing wear that leads to thrust fluctuations. It realizes predictive allocation of underwater thruster power, bidirectional recycling of energy, and adaptive optimization management under multiple operating conditions.

[0006] In a first aspect, this application provides a high-efficiency thruster power management optimization method, the high-efficiency thruster power management optimization method comprising:

[0007] Step S1: Collect the phase difference, rotational speed difference, and load torque of the magnet array, and calculate the slip characteristic value of the magnetic coupling drive using the asymmetric slip response kernel function;

[0008] Step S2: Perform differential operation on the slip characteristic value of the magnetic coupling drive to determine the load change trend, calculate the pre-allocated power value according to the judgment result, and divert the power corresponding to the pre-allocated power value from the main circuit to the back electromotive force energy buffer module.

[0009] Step S3: Construct a multi-thruster power state matrix, calculate the mutual feedback weight based on the buffer energy and load difference in the back EMF energy buffer module, establish a bidirectional energy transfer channel between thrusters, and determine the energy transfer power through the bidirectional energy transfer channel;

[0010] Step S4: Calculate the bearing friction coefficient using the power balance method, calculate the feedforward compensation power based on the nonlinear mapping of wear degree, and optimize the compensation parameters using a self-learning algorithm;

[0011] Step S5: The pre-allocated power value, the energy transfer power, and the feedforward compensation power are superimposed to generate the output power command for each thruster.

[0012] Secondly, this application provides a high-efficiency thruster power management optimization system, the high-efficiency thruster power management optimization system comprising:

[0013] The calculation module is used to collect the phase difference, speed difference and load torque of the magnet array, and calculate the slip characteristic value of the magnetic coupling drive through the asymmetric slip response kernel function;

[0014] The shunt module is used to perform differential calculation on the slip characteristic value of the magnetic coupling drive to determine the load change trend, calculate the pre-allocated power value based on the judgment result, and shunt the power corresponding to the pre-allocated power value from the main circuit to the back EMF energy buffer module.

[0015] The transfer module is used to construct the power state matrix of multiple thrusters, calculate the mutual feedback weight based on the buffer energy and load difference in the back EMF energy buffer module, establish a bidirectional energy transfer channel between thrusters, and determine the energy transfer power through the bidirectional energy transfer channel.

[0016] The optimization module is used to back-calculate the bearing friction coefficient using the power balance method, calculate the feedforward compensation power based on the nonlinear mapping of wear degree, and optimize the compensation parameters using a self-learning algorithm.

[0017] The superposition module is used to superimpose the pre-allocated power value, energy transfer power, and feedforward compensation power to generate the output power command for each thruster.

[0018] Thirdly, a high-efficiency thruster power management optimization device is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the high-efficiency thruster power management optimization device to execute the above-described high-efficiency thruster power management optimization method.

[0019] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the above-described efficient thruster power management optimization method.

[0020] In the technical solution provided in this application, by using magnetic coupling transmission slip characteristic value extraction and asymmetric slip response kernel function calculation in the power management of underwater thrusters, the characteristic change law of magnet array phase difference, speed slip rate and load torque can be identified 0.1-0.3 seconds before the load change occurs. Compared with the passive response mode of the prior art that waits for overcurrent or overheating to occur before starting protection, the prediction mechanism of this application changes power management from post-event protection to pre-event prevention. By performing first-order difference operation and moving average filtering on the slip characteristic value time series, the load change trend is accurately judged and the pre-allocated power value is calculated. Before the main circuit current reaches the overload state, part of the power is diverted to the back EMF energy buffer module in advance, avoiding the power device from being subjected to instantaneous overload impact. At the same time, the three-level cascaded supercapacitor group, electrolytic capacitor group and film capacitor group structure realizes hierarchical energy management of rapid absorption, medium-term storage and stable output, overcoming the contradictory requirements of the single capacitor group in the prior art that both rapid response and long-term storage are required. The multi-thruster distributed power state matrix constructed in this application centrally manages the previously isolated thruster power information by recording the current output power, reserved power margin, back EMF buffer energy, motor thermal state, and load level of each thruster. The mutual feedback priority weight algorithm comprehensively considers four dimensions: reserved margin, buffer energy, thermal margin, and load difference. The sigmoid function is used to fuse the multi-dimensional information into weight values ​​in the range of 0 to 1. Compared with the existing technology where each thruster operates independently and cannot support each other, the bidirectional energy transfer channel established in this application can transfer the stored energy of the high-buffered-energy thruster to the signal power bus of the high-load thruster through an isolated DC-DC converter. This realizes the transformation of back EMF energy from passive dissipation to active reuse, allowing the energy that was originally lost as heat in a single thruster buffer module to be recycled within the thruster cluster.

[0021] The power balance method employed in this application can accurately separate bearing friction power consumption from motor input power. By back-calculating the bearing friction coefficient and defining wear degree as the ratio of the current friction coefficient to the initial friction coefficient, a quantitative evaluation index for wear degree is established. In the feedforward compensation power calculation, the wear degree is raised to the power of 1.6 to reflect the bearing wear acceleration characteristics, i.e., the nonlinear relationship where the impact is small in the early stage of wear but rapidly increases in the later stage. Compared with the lag response of passively increasing power only after the thrust decreases due to severe bearing wear in the prior art, this application starts to apply compensation power when the wear degree just exceeds 1.0. Combined with the product of water density and propeller velocity squared as the second compensation term, the influence of both mechanical wear and hydrodynamics on power demand is comprehensively considered. The gradient descent self-learning algorithm constructs an optimization objective function of the sum of squares of the difference between the target thrust and the actual thrust, and iteratively updates the first and second compensation coefficients, so that the compensation parameters can adapt to the individual differences of different propellers and the environmental characteristics of different waters, overcoming the limitation of fixed compensation parameters in the prior art that cannot adapt to diverse working conditions. The dual-core processor architecture separates the fast-response magnetic coupling slip feature extraction from the computationally intensive multi-thruster mutual feed scheduling. The first core's 200-microsecond execution cycle matches the 5000Hz sampling frequency of the Hall sensor, ensuring timely processing of each sampled data. The second core's 1-millisecond execution cycle provides ample time for solving the power balance equation and calculating the mutual feed weights. The high-speed serial peripheral interface transmits 16-bit power shunting instructions in just 1 microsecond at a 40MHz clock frequency. The controller LAN bus with a 1Mbps baud rate, combined with the token ring protocol, enables orderly communication between multiple thrusters. The 20kHz frequency and 12-bit resolution of the pulse width modulation signal provide approximately 0.024% power regulation accuracy. The differential signal line transmission has strong anti-common-mode interference capability. The coordinated operation of these hardware architectures and communication mechanisms ensures that the pre-allocated power value, energy transfer power, and feedforward compensation power can be superimposed in real time to generate the final output power command. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of one embodiment of the high-efficiency thruster power management optimization method in this application.

[0024] Figure 2 This is a schematic diagram of one embodiment of the high-efficiency thruster power management optimization system in this application.

[0025] Figure 3 This is a schematic block diagram of the structure of the high-efficiency thruster power management optimization device in an embodiment of the present invention. Detailed Implementation

[0026] This application provides an efficient thruster power management optimization method and system. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0027] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the high-efficiency thruster power management optimization method in this application includes:

[0028] Step S1: Collect the phase difference, rotational speed difference, and load torque of the magnet array, and calculate the slip characteristic value of the magnetic coupling drive using the asymmetric slip response kernel function;

[0029] Step S2: Perform differential calculation on the slip characteristic value of the magnetic coupling drive to determine the load change trend, calculate the pre-allocated power value based on the judgment result, and divert the power corresponding to the pre-allocated power value from the main circuit to the back EMF energy buffer module.

[0030] Step S3: Construct a multi-thruster power state matrix, calculate the mutual feedback weight based on the buffer energy and load difference in the back EMF energy buffer module, establish a bidirectional energy transfer channel between thrusters, and determine the energy transfer power through the bidirectional energy transfer channel;

[0031] Step S4: Calculate the bearing friction coefficient using the power balance method, calculate the feedforward compensation power based on the nonlinear mapping of wear degree, and optimize the compensation parameters using a self-learning algorithm;

[0032] Step S5: Superimpose the pre-allocated power value, energy transfer power, and feedforward compensation power to generate the output power command for each thruster.

[0033] It is understood that the executing entity of this application can be a high-efficiency thruster power management optimization system, or it can be a terminal or a server; no specific limitation is made here. This application's embodiments use a server as an example for illustration.

[0034] Specifically, addressing the technical problems in existing underwater thruster power management, such as overheat protection relying solely on a fixed power reduction strategy, passive dissipation of back EMF, isolated power operation across multiple thrusters, and unpredictable compensation for bearing wear, this paper proposes several solutions. These solutions include: pre-predicting load surges through magnetic coupling drive slip characteristic extraction; transferring impending back EMF energy to a buffer module via a power pre-allocation mechanism instead of passively absorbing it after generation; establishing an energy feedback channel through a multi-thruster power state matrix to achieve dynamic power distribution among thrusters; and proactively increasing power in worn thrusters through bearing friction coefficient back-calculation and self-learning compensation instead of passively adjusting after thrust decreases. A Hall sensor array acquires the phase difference between the magnet inside the thruster housing and the magnet inside the propeller hub at a frequency of 5000 times per second. Photoelectric encoders and magnetic encoders synchronously measure the motor speed and propeller speed. The difference between the two speeds is divided by the motor speed to calculate the slip ratio. A torque sensor measures the load torque borne by the magnetic coupling drive. The asymmetric slip response kernel function performs a 1.4-fold power operation on the phase difference to reflect the nonlinear attenuation of magnetic field coupling, and a 2.1-fold power operation on the slip rate to reflect the superlinear growth of eddy current losses. The load torque is used as a negative exponential function variable to calculate the attenuation weight, reflecting the decrease in the overload protection threshold. These three terms are multiplied by their respective weighting coefficients and summed to obtain the slip characteristic value. A 30ms sliding window is constructed to sample this characteristic value to obtain a time series. The difference between adjacent sampling points in the series is divided by the sampling interval to obtain the instantaneous rate of change. When the rate of change continuously exceeds 15 / s and lasts for more than 10ms, it is determined that the load is about to change abruptly. At this time, the current characteristic value is subtracted from the baseline value of 3, multiplied by 0.02, and then added to obtain the pre-allocation ratio coefficient. This coefficient multiplied by the rated power of the thruster gives the pre-allocation power value. The silicon carbide field-effect transistor switching circuit shunts the power corresponding to this value from the main circuit at a frequency of 100kHz and injects it into the back EMF energy buffer module, which consists of a supercapacitor bank, an electrolytic capacitor bank, and a film capacitor bank cascaded in three stages. The supercapacitor has a capacity of 2700F to quickly absorb the peak, the electrolytic capacitor has a capacity of 18000μF for medium-term storage, and the film capacitor has a capacity of 4700μF for stable output. The energy flow is achieved by cascading the three stages through a bidirectional converter.

[0035] A power state matrix with the number of rows equal to the number of thrusters is constructed for the multi-thruster system. Each row contains 5 columns recording the current output power, reserved power margin, back EMF buffer energy, motor thermal state, and load level. The reserved power margin is calculated by subtracting the current output power from the rated power, and the load level is the ratio of the current torque to the rated torque. When calculating the priority weight of mutual feedback between any two thrusters, the first weight term is obtained by multiplying the ratio of the receiver's reserved power margin to its rated power by a first weighting coefficient of 0.4; the second weight term is obtained by multiplying the ratio of the transmitter's buffer energy to its maximum buffer energy by a second weighting coefficient of 0.3; the third weight term is obtained by normalizing and inverting the ratio of the receiver's temperature to its limit by a third weighting coefficient of 0.2; and the fourth weight term is obtained by multiplying the square of the difference between the two load levels by a fourth weighting coefficient of 0.1 and taking the negative value. The sum of these four terms is then input into a sigmoid function and mapped to the interval between 0 and 1. All thruster pairs are iterated through to calculate their respective weights. Thruster pairs with weights greater than 0.5 are selected and sorted in descending order of weight. The first few pairs are selected to establish a bidirectional energy transfer channel. The isolated DC-DC converter uses a full-bridge resonant topology to extract DC voltage from the transmitting thin-film capacitor bank. After the transformer turns ratio is changed, it is injected into the receiving signal power bus. The transfer power is determined by the minimum of the transmitting power gap, the receiving available power, and the converter rated power. The transfer time is calculated by dividing the transmitting buffer energy by the transfer power. The converter efficiency is 94%.

[0036] When calculating the bearing friction coefficient using the power balance method, the propeller maintains a constant speed. The motor input power is measured by a wattmeter, and the thrust is measured by a triaxial force sensor. Based on the thrust value, the induced flow velocity is calculated using propeller theory. The effective propulsion power is obtained by multiplying the thrust by the flow velocity. Fluid resistance power consumption is calculated using a lookup table based on water density, rotational speed, and propeller diameter. Motor copper loss is calculated as the square of the current multiplied by the phase resistance, and iron loss is calculated as the product of the frequency to the power of 1.3 and the square of the magnetic flux density, multiplied by the iron loss coefficient. The bearing friction power consumption is obtained by subtracting the propulsion power, fluid power consumption, copper loss, and iron loss from the input power. This power consumption is then divided by the product of twice pi, rotational speed, bearing radius, and radial load to calculate the friction coefficient. The ratio of the current friction coefficient to the initial friction coefficient in a brand-new state is defined as the bearing wear degree. The wear degree is multiplied by the first compensation coefficient to the power of 1.6 to obtain the first compensation term. The product of water density and the square of the propeller flow velocity is multiplied by the second compensation coefficient to obtain the second compensation term. The sum of these two terms is the feedforward compensation power. When optimizing the compensation coefficients using the gradient descent method, the objective function is constructed as the sum of squares of the difference between the target thrust and the actual thrust. Partial derivatives are calculated for the first and second compensation coefficients, and the coefficient values ​​are iteratively updated with a learning rate of 0.001. The initial first coefficient of 120 and the second coefficient of 0.015 converge to 147 and 0.019 after iteration, so that the thrust after compensation is kept near the rated value instead of increasing the power and causing overheating when wear becomes severe.

[0037] The dual-core processor's first core performs slip feature extraction and pre-allocation decision-making at a 200μs cycle, while the second core performs mutual feedback scheduling and bearing compensation at a 1ms cycle. Based on the pre-allocated power value, a 16-bit power shunting command is generated, containing a header, shunting ratio, and checksum, and sent to the switching circuit via a 40MHz serial peripheral interface. The controller area network bus uses a token ring protocol at a 1Mbps baud rate to collect the status of other thrusters, with each thruster occupying a 2ms communication slot. After calculating the compensation power based on wear and operating parameters, a 20kHz frequency, 12-bit resolution pulse-width modulation signal is generated to adjust the driver's duty cycle. The pre-allocated power value, energy transfer power, and feedforward compensation power are added to obtain the final output power command value, which is transmitted to the drive control unit via a differential signal line for power allocation. This solves the technical problems of existing technologies, such as fixed and singular power management strategies that cannot adapt to operating conditions, passive dissipation of back EMF energy that cannot be recovered, independent power supply for multiple thrusters that cannot support each other, and thrust fluctuations caused by passive adjustment after bearing wear.

[0038] In one specific embodiment, step S1 includes:

[0039] The phase difference between the magnet array inside the thruster housing and the magnet array inside the propeller hub is collected in real time by a Hall sensor array. The sampling frequency is set to 5000Hz to obtain phase difference time series data.

[0040] The output shaft speed of the motor is measured by an optical encoder, and the actual speed of the propeller is measured by a magnetic encoder. The difference between the motor speed and the propeller speed is divided by the motor speed to obtain the slip ratio of the magnetic coupling transmission.

[0041] The load torque is monitored in real time by a torque sensor installed on the shaft connecting the magnetic coupling transmission mechanism and the propeller. The measurement accuracy is set to 0.5% of the full scale to obtain the load torque data.

[0042] The first characteristic component is obtained by raising the phase difference to the sine function value and then raising it to the 1.4th power. The second characteristic component is obtained by raising the slip ratio to the 2.1st power. The attenuation weight is calculated by using the load torque as the independent variable of the negative exponential function. The first and second characteristic components are multiplied by preset weight coefficients and then multiplied by the attenuation weight. The three terms are weighted and summed to obtain the slip characteristic value of the magnetic coupling drive.

[0043] Specifically, the Hall sensor array detects changes in magnetic field strength to capture the relative positional difference between the magnet array inside the propeller housing and the magnet array inside the propeller hub in real time. The sampling frequency is set to 5000Hz to ensure that 5000 phase difference data points are acquired per second to form a continuous time series. This frequency is chosen based on the consideration that the fastest response time for phase changes during magnetic coupling transmission is approximately 0.2ms. The 0.2ms sampling interval can capture the complete process of phase abrupt changes without signal distortion. The photoelectric encoder is installed at the end of the motor output shaft. It generates pulse signals by blocking the phototube when the grating disk rotates. The pulse frequency directly corresponds to the motor speed. The magnetic encoder is installed on the propeller hub shaft system. It uses the magnetic field change signal generated by the magnetic encoder disk to measure the actual propeller speed. The two encoders sample synchronously to ensure the consistency of the speed data time. The slip ratio calculation process for magnetic coupling transmission involves subtracting the motor speed measured by the photoelectric encoder from the propeller speed measured by the magnetic encoder to obtain the speed difference. This difference, divided by the motor speed, is the slip ratio. The slip ratio reflects the relative slippage between the inner and outer magnet arrays in the magnetic coupling transmission. When the propeller load increases, the magnets in the hub lag behind, leading to an increase in the slip ratio. The torque sensor uses a strain gauge structure and is installed on the shaft connecting the magnetic coupling transmission mechanism and the propeller. When the shaft is subjected to torsion, shear strain is generated, and the resistance value of the strain gauge changes accordingly. The resistance change is converted into a voltage signal through a Wheatstone bridge circuit, amplified by an amplifier, and output as a voltage value proportional to the torque. The measurement accuracy is set to 0.5% of the full scale, meaning that for a sensor with a range of 40 Nm, the measurement error does not exceed 0.2 Nm. This accuracy meets the requirements for real-time monitoring of load torque.

[0044] The first step in processing the asymmetric slip response kernel function data is to substitute the phase difference data into a sine function. The phase difference is expressed in radians, and the sine function converts the angular quantity into a dimensionless value between -1 and 1. The sine function value reflects the projected component of the magnetic field vector of the magnet array in the direction of torque transmission. When the phase difference is 0 degrees, the sine value is 0, indicating that the magnets are perfectly aligned and the torque transmission efficiency is the highest. When the phase difference is 90 degrees, the sine value is 1, indicating that the torque transmission begins to suffer from magnet misalignment. The purpose of raising the sine value to the 1.4th power is to amplify the influence of the phase difference on the magnetic field coupling efficiency. The exponent 1.4 comes from the fitting results of the experimental data of the magnetization curve of neodymium iron boron permanent magnets. This nonlinear relationship reflects the accelerated decay characteristic of the magnetic field strength as the phase difference increases. The value obtained after raising the 1.4th power is used as the first characteristic component. The second step processes the slip ratio data by directly exponentiating the slip ratio value to the power of 2.1. The exponent of 2.1 is based on the physical principle that eddy current loss is proportional to the square of the speed difference in the law of electromagnetic induction. The additional increment of 0.1 considers the enhancing effect of the conductive environment of seawater on the eddy current effect. The power of 2.1 results in a much larger increase in the characteristic component corresponding to the slip ratio increasing from 0.01 to 0.05 than linear growth. This superlinear relationship accurately describes the physical process of eddy current loss accelerating with slip. The calculation result is used as the second characteristic component. The third step processes the load torque data by multiplying the load torque value by a preset attenuation coefficient and taking the negative sign as the exponent of the natural exponential function. The attenuation coefficient is determined according to the magnet strength and air gap size of the magnetic coupling transmission mechanism. The calculation result of the negative exponential function is the attenuation weight between 0 and 1. This weight reflects the suppressive effect of the load torque on the slip tolerance. When the load torque is small, the attenuation weight is close to 1, indicating that the magnetic coupling transmission has a large margin to withstand slip. When the load torque is close to the rated value, the attenuation weight drops rapidly to below 0.5, indicating that the slip tolerance has significantly decreased and is approaching the critical slip state.

[0045] The three-term weighted summation process begins by multiplying the first feature component by a preset weighting coefficient to obtain the first weighted term. This weighting coefficient reflects the contribution of the phase difference to the overall slip characteristics. The weighting coefficient value is determined by collecting data on the relationship between phase difference and power loss under different load conditions during offline training. The second feature component is then multiplied by another preset weighting coefficient to obtain the second weighted term. This weighting coefficient reflects the contribution of eddy current loss to the slip characteristics and is also determined through fitting experimental data. Finally, the first and second weighted terms are multiplied by an attenuation weight, respectively. This multiplication operation achieves the coupling modulation of the two feature components by the load torque. When the load is small, the attenuation weight is close to 1, having little impact on the feature components. When the load increases, the attenuation weight decreases, causing the feature component value to be reduced proportionally, reflecting the actual situation where slip tolerance decreases under high loads. Finally, the two multiplication results are added together to obtain the slip characteristic value of the magnetic coupling drive. This characteristic value comprehensively reflects the combined influence of three physical quantities—phase difference, slip rate, and load torque—on the state of the magnetic coupling drive. The characteristic value range is set from 0 to 10. The larger the value, the closer the magnetic coupling drive is to the overload slip state. A value less than 3 indicates the normal working range, a value between 3 and 6 indicates the warning range, and a value greater than 6 indicates that the slip state has been entered and the load needs to be reduced or the power input cut off immediately.

[0046] In one specific embodiment, step S2 includes:

[0047] The slip characteristic values ​​of the magnetic coupling drive are sampled in a sliding window with a time window length of 30ms. Each time window contains 150 sampling points to obtain the time series of slip characteristic values.

[0048] Perform a first-order difference operation on the slip eigenvalue time series, calculate the change in slip eigenvalue between adjacent sampling points and divide it by the sampling time interval to obtain the slip eigenvalue change rate.

[0049] Determine whether the rate of change of the slip characteristic value continuously exceeds 15 per second and lasts for more than 10 ms. When the condition is met, trigger the load change warning signal. Subtract the reference value 3 from the current slip characteristic value, multiply by 0.02 and add 0.1 to obtain the pre-allocation ratio coefficient. Multiply the pre-allocation ratio coefficient by the rated power of the thruster to obtain the pre-allocation power value.

[0050] The power corresponding to the pre-allocated power value is diverted from the main circuit of the power supply through a fast power switching circuit. The switching circuit uses silicon carbide field-effect transistors and the switching frequency is set to 100kHz. The diverted power is injected into the back EMF energy buffer module, which consists of a supercapacitor group, an electrolytic capacitor group, and a film capacitor group in a three-stage cascade.

[0051] Specifically, the sliding window sampling extracts a fixed-duration data segment from the continuous slip characteristic value data stream by setting a time window length of 30ms. The sampling frequency of 5000Hz means that 5000 slip characteristic value data points are generated per second. The number of sampling points contained in the 30ms time window is 5000 multiplied by 0.03, which equals 150 data points. The sliding window moves forward by one sampling point each time, which is 0.2ms. The new window contains 149 points after the previous window plus the newly acquired point. By continuously extracting data through the sliding window, multiple slip characteristic value time series of 150 points are formed. The time series retains the complete information of the evolution of slip characteristic values ​​over time. The selection of the window length of 30ms is based on the typical time scale of the transition of magnetic coupling drive from normal state to overload slip state. This duration can capture the complete process of load change without causing early warning lag due to excessively long windows. The first-order difference operation performs a subtraction operation on adjacent sampling points in the time series. The slip feature value of the i-th sampling point is extracted from the time series and recorded as the current value. The slip feature value of the i-th minus 1 sampling point is extracted and recorded as the previous value. The change in slip feature value is obtained by subtracting the previous value from the current value. This change reflects the increase or decrease of slip feature value between adjacent sampling points. The change is divided by the sampling time interval of 0.2ms to obtain the instantaneous rate of change at that sampling point. The instantaneous rate of change is measured in units of seconds and its magnitude represents the speed at which the slip feature value increases or decreases. The above difference operation is performed sequentially on 149 sampling points from the 2nd to the 150th in the time series to generate a new sequence containing 149 instantaneous rate of change values.

[0052] Moving average filtering smooths the instantaneous rate of change sequence to eliminate high-frequency noise. The filtering window contains several adjacent instantaneous rate of change data points. The arithmetic mean of the instantaneous rate of change of all data points in the window is used as the smoothed rate of change of the center point of the window. The filtering window slides along the instantaneous rate of change sequence to cover all data points. Each data point is filtered to obtain the corresponding smoothed rate of change. The filtered rate of change sequence eliminates sensor measurement noise and random fluctuations in the magnetic coupling transmission process, while retaining the main trend information of load change. The condition check for determining load mutation includes two dimensions. The first dimension checks whether the value of the smoothed rate of change continuously exceeds the threshold of 15 seconds. Continuous exceedance means that there are several adjacent data points in the filtered rate of change sequence whose rate of change values ​​are all greater than 15 seconds. This threshold is set based on the statistical characteristic that the slip characteristic value change rate is usually less than 10 seconds under normal operating conditions. A change rate exceeding 15 seconds indicates that the slip characteristic value is rising rapidly. The second dimension checks whether the duration of exceeding the threshold exceeds 10 ms. The duration is calculated by counting the number of data points that continuously exceed the threshold and multiplying it by the sampling interval of 0.2 ms. The requirement that the duration exceeds 10 ms means that at least 50 consecutive sampling points exceed the threshold. This time condition excludes brief instantaneous disturbances to avoid false triggering. Only when both conditions are met simultaneously—a rate of change exceeding 15 seconds and a duration exceeding 10 ms—is the load mutation warning signal triggered.

[0053] The pre-allocation ratio coefficient is calculated by subtracting a reference value of 3 from the current slip characteristic value. The reference value of 3 corresponds to the boundary point where the magnetic coupling drive enters the warning zone from the normal operating zone. The difference reflects the depth of the slip characteristic value within the warning zone. The difference is multiplied by a scaling factor of 0.02, which establishes a linear relationship between the depth of the slip characteristic value and the pre-allocation ratio. The multiplication result is then added to a base of 0.1. The base of 0.1 ensures that at least 10% of the power is pre-allocated even when the difference is close to zero when the slip characteristic value just enters the warning zone. The calculated pre-allocation ratio coefficient ranges from 0.1 to a larger value. The ratio coefficient is multiplied by the rated power of the thruster to obtain the pre-allocated power value. The rated power is the maximum continuous output power specified on the thruster nameplate. The multiplication operation converts the dimensionless ratio coefficient into a dimensional power value in watts. After receiving the pre-allocated power value command, the fast power switching circuit performs a power shunt operation. The core device of the circuit is a silicon carbide field-effect transistor (MOSFET). Compared with traditional silicon-based MOSFETs, this device has lower on-resistance and faster switching speed. The switching frequency is set to 100kHz, which means that the MOSFET performs 100,000 switching actions per second, with each switching cycle being 10 microseconds. The on-time ratio of the MOSFET in each cycle is adjusted by pulse width modulation technology. The on-time ratio is proportional to the pre-allocated power value. When the MOSFET is on, the current in the main circuit of the power supply is partially guided to the shunt branch, which is connected to the back EMF energy buffer module. When the MOSFET is off, the current in the main circuit returns to the normal path. The high-frequency switching action makes the shunt current appear as a continuous and controllable power transfer on a macroscopic scale.

[0054] The back EMF energy buffer module adopts a three-stage cascaded capacitor array structure. The first stage is a supercapacitor bank, with each individual capacitor having a capacitance of 2700 farads and a withstand voltage of 48 volts, and an equivalent series resistance of less than 0.3 milliohms. Supercapacitors store charge using the electrochemical double-layer principle. Their extremely high capacitance and extremely low internal resistance enable them to absorb large current pulses within milliseconds. When the power switching circuit shunts the main circuit power to the buffer module, the supercapacitor bank receives the shunt current first. Its low internal resistance characteristic prevents voltage spikes caused by sudden current changes from damaging subsequent circuits. The second stage is an electrolytic capacitor bank, with each individual capacitor having a capacitance of 18000 microfarads and a withstand voltage of 100 volts, and an equivalent series resistance of approximately 5 milliohms. Electrolytic capacitors store charge through a chemical reaction between the electrolyte and the metal foil. Their capacitance is larger than that of film capacitors but smaller than that of supercapacitors, and their internal resistance is higher than that of supercapacitors but lower than that of film capacitors. The electrolytic capacitor bank undertakes the medium-term energy storage task, gradually transferring the energy rapidly absorbed by the supercapacitors to the electrolytic capacitors to achieve energy buffering over time. The third stage is a film capacitor bank, with each individual capacitor having a capacitance of 4700 microfarads and a withstand voltage of 63 volts. The equivalent series resistance is about 2 milliohms. The film capacitors use metallized polypropylene film as the dielectric, which is characterized by good voltage stability and excellent frequency characteristics. The film capacitor bank is responsible for stabilizing the output voltage. When buffered energy is needed later, a stable DC power supply with low ripple can be obtained from the output of the film capacitor bank. The three capacitors are cascaded together via a bidirectional DC-DC converter. The converter topology is a bidirectional CLLC resonant structure. The converter is positioned between the supercapacitor and the electrolytic capacitor, and between the electrolytic capacitor and the film capacitor. The bidirectional characteristic allows energy to flow bidirectionally between adjacent stages. The resonant structure achieves soft switching and reduces switching losses through inductor-capacitor resonance. The converter automatically adjusts the energy flow direction and power based on the voltage state of each capacitor. When the supercapacitor voltage is higher than a set threshold, the converter transfers energy to the electrolytic capacitor. When the electrolytic capacitor voltage is higher than the threshold, it continues to transfer energy to the film capacitor. When buffered energy needs to be released, the energy flows in the reverse direction from the film capacitor through the electrolytic capacitor back to the supercapacitor and then outputs to the load. The three-stage cascaded structure realizes hierarchical power management with rapid absorption, medium-term storage, and stable output. It solves the technical problem of passive dissipation of back EMF energy in existing technologies, which cannot be recovered and utilized. The power pre-allocation mechanism avoids the problem of power devices being subjected to instantaneous overload impacts by predicting the sudden change trend of the load in advance and transferring power in advance, which is the case in existing technologies.

[0055] In one specific embodiment, a first-order difference operation is performed on the slip eigenvalue time series to calculate the change in slip eigenvalue between adjacent sampling points, divided by the sampling time interval, to obtain the slip eigenvalue change rate, including:

[0056] Extract the slip feature value of the current sampling point and the slip feature value of the previous sampling point from the slip feature value time series, calculate the difference between the two, and obtain the change in slip feature value;

[0057] The sampling time interval between adjacent sampling points is calculated based on the sampling frequency. The change in the slip characteristic value is divided by the sampling time interval to obtain the instantaneous rate of change of a single sampling point.

[0058] The above difference operation is performed sequentially on all sampling points in the slip eigenvalue time series to generate an instantaneous rate of change sequence with the same length as the original time series;

[0059] The instantaneous rate of change sequence is processed by moving average filtering. The sliding window contains several sampling points. The arithmetic mean of the instantaneous rate of change within the window is calculated to obtain the smoothed slip characteristic rate of change.

[0060] Specifically, the calculation of slip characteristic value change involves locating the current sampling point position from the slip characteristic value time series, reading the slip characteristic value data stored at that position as the current value, then locating the previous sampling point position (i.e., decrementing the current position index by 1), reading the slip characteristic value data stored at the previous position as the previous value, and performing a subtraction operation between the current value and the previous value. The subtraction result is the change in slip characteristic value between two adjacent sampling points. A positive change indicates that the slip characteristic value is increasing, and a negative change indicates that the slip characteristic value is decreasing. The absolute value of the change reflects the degree of change. This differential operation eliminates the DC component of the slip characteristic value while retaining dynamic change information. The sampling time interval is calculated from the reciprocal of the sampling frequency. Taking the reciprocal of the sampling frequency of 5000Hz, the sampling time interval is 1 divided by 5000, which equals 0.0002 seconds or 0.2 milliseconds. This time interval is a fixed value that remains unchanged throughout the sampling process. The change in slip characteristic value is divided by the sampling time interval of 0.2 milliseconds. The division operation converts the change into a rate of change. The dimension of the rate of change is converted from a dimensionless change to a unit per second. The result of the division is called the instantaneous rate of change of a single sampling point. The instantaneous rate of change value represents the amount by which the slip characteristic value will increase or decrease per second if it continues to change at the current speed.

[0061] When performing differencing operations sequentially on all sampling points in a slip eigenvalue time series, assuming the time series contains 150 sampling points, starting from the second sampling point, since the first sampling point has no previous value, the instantaneous rate of change at the second sampling point is equal to the slip eigenvalue at the second sampling point minus the slip eigenvalue at the first sampling point, divided by 0.2 milliseconds. The instantaneous rate of change at the third sampling point is equal to the slip eigenvalue at the third sampling point minus the slip eigenvalue at the second sampling point, divided by 0.2 milliseconds, and so on. Extending to the 150th sampling point, the instantaneous rate of change at the 150th sampling point is equal to the slip characteristic value at the 150th sampling point minus the slip characteristic value at the 149th sampling point, divided by 0.2 milliseconds. The difference operation produces 149 instantaneous rate of change values. These 149 values ​​are arranged in chronological order to form an instantaneous rate of change sequence. The length of the instantaneous rate of change sequence is one data point shorter than the original slip characteristic value time series. Each element in the sequence corresponds to the slip characteristic value change rate at the corresponding moment in the original time series.

[0062] Moving average filtering smooths the instantaneous rate of change sequence, eliminating measurement noise and short-term fluctuations. The filtering window contains several adjacent instantaneous rate of change data points. The window size determines the smoothing effect; more data points in the window result in a stronger smoothing effect but a slower response time, while fewer data points result in a faster response time but weaker noise suppression. The calculation process of moving average filtering is as follows: First, the filtering window is placed at the beginning of the instantaneous rate of change sequence. All instantaneous rate of change data points covered by the window are read, and the values ​​of these data points are summed. The sum is divided by the number of data points in the window to obtain the arithmetic mean, which is used as the smoothed rate of change at the sampling point at the center of the window. Then, the filtering window is moved one sampling point to the right. After the window moves, it includes the latter part of the data points of the previous window plus a new data point. The above summation and averaging calculation is repeated at the new window position to obtain the smoothed rate of change at the center point of the new position. The window continues to slide and the calculation is repeated until the window moves to the end of the sequence. The length of the smoothed rate of change sequence generated by moving average filtering is the same as or slightly shorter than the original instantaneous rate of change sequence, depending on the boundary processing method.

[0063] In one specific embodiment, step S3 includes:

[0064] Construct a distributed power state matrix containing n thrusters. Each row of the matrix corresponds to one thruster. The five columns are the current output power, reserved power margin, back EMF buffer energy, motor thermal state, and load level, respectively. The reserved power margin is calculated by subtracting the current output power from the rated power of the thruster, and the load level is calculated by dividing the current load torque by the rated torque.

[0065] For any two thrusters, the mutual feed-in priority weight is calculated by multiplying the ratio of the reserved power margin of the first thruster to its rated power by a first weighting coefficient to obtain the first weighting term; multiplying the ratio of the back EMF buffer energy of the second thruster to its maximum buffer energy by a second weighting coefficient to obtain the second weighting term; normalizing the ratio of the motor temperature of the first thruster to its temperature limit and multiplying it by a third weighting coefficient to obtain the third weighting term; and multiplying the square of the difference in load levels between the two thrusters by a fourth weighting coefficient to obtain the fourth weighting term. The four terms are then weighted and summed and mapped to the 0 to 1 interval using a sigmoid function to obtain the mutual feed-in priority weight.

[0066] Iterate through all thruster pairs and calculate their respective mutual feedback priority weights. Filter thruster pairs with weight values ​​greater than the threshold of 0.5 as energy transfer candidates. Sort them by weight from largest to smallest and select the top few thruster pairs.

[0067] A bidirectional energy transfer channel is established for the selected thruster pair. The energy of the high back EMF buffered energy thruster is converted into voltage and injected into the signal power bus of the low energy thruster through an isolated DC-DC converter. The transfer power is determined according to the minimum value among the power gap of the transmitter, the available power of the receiver, and the rated power of the converter.

[0068] Specifically, when constructing the distributed power state matrix, the matrix dimensions are first determined. The number of rows, n, equals the number of thrusters carried by the underwater vehicle, and the number of columns is fixed at 5. The matrix is ​​stored using a two-dimensional array data structure, with each row index corresponding to a unique identifier for a thruster. The first column stores the current output power, which is collected in real time by the power measurement module of the thruster driver. The measurement module synchronously samples the motor input voltage and current, multiplies the voltage by the current to obtain the instantaneous power, and then applies a moving average filter to the instantaneous power to obtain the current output power. The second column stores the reserved power margin, calculated by subtracting the current output power from the rated power indicated on the thruster's nameplate. The rated power is the maximum power that the thruster can continuously output under rated operating conditions. The reserved power margin reflects how much more output power the thruster can increase without exceeding the rated limit. When the thruster is running at full load, the reserved margin is close to zero; when the thruster is running at light load, the reserved margin is close to the rated power. The third column stores the back EMF buffer energy, which is read from the energy management unit of the back EMF energy buffer module. The energy management unit monitors the voltage of the three-stage capacitor array in real time, calculates the energy stored in each stage of the capacitor based on the relationship between capacitance and the square of voltage, and sums the three stages of energy to obtain the total buffer energy. The fourth column stores the motor thermal status. The winding temperature is measured by a PT1000 platinum resistance temperature sensor embedded in the motor winding. The resistance value of the PT1000 changes linearly with temperature. The measuring circuit converts the resistance value into a temperature value, and the motor thermal status is directly recorded as the Celsius value measured by the temperature sensor. The fifth column stores the load level. The calculation process involves reading the current load torque measured by the torque sensor, dividing this torque by the rated torque of the thruster. The rated torque is the torque output by the thruster at rated power and rated speed. The load level is a dimensionless ratio; a value less than 1 indicates that the load is below the rated value, a value equal to 1 indicates that the load has reached the rated value, and a value greater than 1 indicates that the load exceeds the rated value and enters an overload state.

[0069] The mutual feedback priority weight calculation is performed on any two thrusters. One thruster is designated as the energy receiver, denoted as thruster i, and the other as the energy transmitter, denoted as thruster j. The first weight term is calculated by reading the reserved power margin and rated power of thruster i from the power state matrix. The reserved power margin is divided by the rated power to obtain the margin ratio. This ratio reflects the ability of thruster i to receive additional energy. A margin ratio close to 1 indicates that thruster i is in a light-load state with sufficient margin to receive energy, while a margin ratio close to 0 indicates that thruster i is close to full load and cannot receive more energy. The margin ratio is multiplied by the first weight coefficient of 0.4 to obtain the first weight term. The weight coefficient of 0.4 indicates that the reserved margin accounts for 40% of the weight in the mutual feedback decision. The second weighting term is calculated by reading the back EMF buffer energy of thruster j from the power state matrix and dividing this energy by the maximum buffer energy of the buffer module to obtain the energy ratio. The maximum buffer energy is the total energy that the three-stage capacitor array can store under rated voltage. The energy ratio reflects the energy reserve that thruster j can output. An energy ratio close to 1 indicates that the buffer module of thruster j is fully charged and has sufficient energy to transfer. An energy ratio close to 0 indicates that the buffer module of thruster j is almost empty and has no energy to transfer. The energy ratio is multiplied by the second weighting coefficient of 0.3 to obtain the second weighting term. The weighting coefficient of 0.3 indicates that the buffer energy accounts for 30% of the weight in the mutual feedback decision. The third weighting term is calculated by reading the motor temperature of thruster i from the power state matrix and dividing this temperature by the motor temperature limit to obtain the temperature ratio. The temperature limit is the highest temperature that the motor insulation material can withstand for a long time, usually 90 degrees Celsius. The temperature ratio reflects the thermal margin of thruster i. A temperature ratio close to 1 indicates that the motor temperature is close to the limit and the thermal margin is small. A temperature ratio much less than 1 indicates that the motor temperature is low and the thermal margin is large. The normalization process is to subtract the temperature ratio from 1 to convert the temperature ratio into a thermal margin ratio. The temperature ratio of 0.8 is normalized to 0.2, which means that there is still a 20% thermal margin. Multiplying the normalized thermal margin ratio by the third weighting coefficient of 0.2 gives the third weighting term. The weighting coefficient of 0.2 indicates that the thermal margin accounts for 20% of the weight in the mutual feedback decision. The fourth weighting term is calculated by reading the load levels of thruster i and thruster j from the power state matrix, calculating the difference between them, and then squaring the difference. The square of the difference reflects the degree of load imbalance between the two thrusters. The larger the square of the difference, the greater the load difference, and the more necessary it is to achieve load balancing through energy feedback. The fourth weighting term is obtained by multiplying the square of the difference by a negative fourth weighting coefficient of -0.1. The negative sign indicates that the larger the load difference, the smaller the weight, suppressing energy transfer under excessive load difference and avoiding exacerbating the imbalance. The weighting coefficient of 0.1 indicates that the load difference accounts for 10% of the weight in the feedback decision.The four weights are added together to obtain a weighted sum. This weighted sum is then substituted into the sigmoid function. The sigmoid function is formed by dividing 1 by 1 and raising the weighted sum to the power of the negative weighted sum of the natural exponential function. The sigmoid function maps any real number to the interval between 0 and 1. When the weighted sum is positive, the sigmoid output is greater than 0.5. The larger the weighted sum, the closer the output is to 1. When the weighted sum is negative, the sigmoid output is less than 0.5. The smaller the weighted sum, the closer the output is to 0. The output of the sigmoid function is the mutual feed-in priority weight between thruster i and thruster j.

[0070] A double-loop structure is used to iterate through all thruster pairs. The outer loop iterates through each thruster as an energy receiver, while the inner loop iterates through each thruster other than the receiver as an energy sender. For each receiver-sender combination, the aforementioned mutual feedback priority weight calculation is performed. The calculated weight values, along with the thruster pair identifiers, are stored in a weight list. The length of the weight list is n multiplied by n minus 1 because thrusters do not form mutual feedback pairs with themselves. The filtering process iterates through the weight list, checking if each weight value is greater than a threshold of 0.5. Thruster pairs with weights greater than 0.5 are added to the candidate list. The threshold of 0.5 is used as the filtering criterion based on the characteristics of the sigmoid function; a weight greater than 0.5 indicates a positive weighted sum, meaning the overall mutual feedback conditions are favorable. The sorting process arranges the candidate list in descending order of weight values ​​using a quicksort algorithm to ensure sorting efficiency. After sorting, the thruster pair with the highest weight is placed at the top of the list. The selection process starts from the first candidate in the sorted list and selects thruster pairs in sequence. The number of selections is limited by the power system capacity and the number of converters. Usually, the first 3 to 5 pairs of thrusters are selected to establish an energy transfer channel. During the selection process, it is checked whether the same thruster is simultaneously the sender or receiver in multiple pairs. If the same thruster has been selected, subsequent thruster pairs containing that thruster are skipped to avoid energy flow conflicts.

[0071] When establishing a bidirectional energy transfer channel, an isolated DC-DC converter is configured based on the selected thruster. The converter input is connected to the output of the thin-film capacitor bank of the back EMF energy buffer module of the energy transmitter thruster, and the converter output is connected to the signal power bus of the energy receiver thruster. The converter adopts a full-bridge LLC resonant topology. The four switches on the input side form a full-bridge inverter to convert DC voltage into high-frequency AC voltage. The resonant inductor and resonant capacitor, along with the transformer, form an LLC resonant cavity. The resonant cavity operates near the resonant frequency to achieve soft switching and reduce switching losses. The transformer provides electrical isolation and voltage conversion. The rectifier bridge and filter capacitor on the output side rectify and filter the high-frequency AC voltage into DC voltage. When determining the transferred power, the power gap of the transmitter is first calculated. The current output power and rated power are read from the power state matrix of the energy receiver. If the receiver is in an overload state and the current output power exceeds the rated power, the power gap is equal to the current output power minus the rated power. If the receiver is not overloaded, the power gap is zero. Then, the available power of the receiver is calculated. The reserved power margin is read from the power state matrix of the energy sender. The reserved margin, or available power, indicates how much the sender can reduce its output without affecting its mission. The rated power specified on the converter's nameplate is read. The rated power is the maximum power that the converter can continuously transmit. By comparing the power gap, available power, and converter rated power, the minimum value is selected as the actual transferred power. This minimum value ensures that it does not exceed the receiver's demand, the transmitter's capacity, or the converter's capacity. The transferred power value is sent to the converter control unit via the controller LAN bus. The control unit adjusts the duty cycle and operating frequency of the converter's switching transistors based on the transferred power. The duty cycle and frequency together determine the amount of power transmitted by the converter. The transferred power, or energy transfer power, is recorded in the energy management log. This solves the technical problem of isolated power island operation for multiple thrusters in existing technologies, where they cannot support each other. The distributed power state matrix enables centralized management of power information for multiple thrusters. The mutual feedback priority weight algorithm comprehensively considers four dimensions—reservation margin, buffer energy, thermal margin, and load balance—to achieve intelligent mutual feedback decision-making. The bidirectional energy transfer channel achieves energy redistribution between thrusters through the DC-DC converter, allowing high-load thrusters to receive energy support from low-load thrusters.

[0072] In one specific embodiment, step S4 includes:

[0073] The motor input power and thrust output are measured when the thruster is running at a constant speed. The effective propulsion power, fluid resistance power consumption, motor copper loss and iron loss are calculated based on the power balance relationship. The bearing friction power consumption is obtained by subtracting the power consumption from the input power.

[0074] The bearing friction coefficient is calculated by back-calculating the bearing friction power consumption, propeller speed, bearing radius and radial load, and the ratio of the current friction coefficient to the initial friction coefficient is taken as the bearing wear degree.

[0075] The first compensation term is obtained by multiplying the wear rate by the first compensation coefficient to the power of 1.6. The second compensation term is obtained by multiplying the product of the water density and the square of the propeller velocity by the second compensation coefficient. The two terms are summed to obtain the feedforward compensation power.

[0076] The objective function is constructed as the sum of squares of the difference between the target thrust and the actual thrust. The gradient descent method is used to iteratively update the first and second compensation coefficients to obtain the optimized compensation parameters.

[0077] Specifically, the constant-speed operation of the thruster is achieved through closed-loop speed control. The controller continuously adjusts the output power of the driver to maintain the motor speed at the set value. A constant-speed state is considered achieved when the speed fluctuation is less than 1% of the rated speed. The motor input power is measured by a power analyzer. The analyzer simultaneously collects the voltage and current waveforms at the motor input, multiplies the voltage and current waveforms point by point, and takes the average value to obtain the average power. The sampling period covers at least 10 motor electrical cycles to ensure measurement accuracy. Thrust output is measured by a triaxial force sensor installed between the thruster's fixed bracket and the hull of the vessel. The strain gauge inside the sensor converts the thrust into an electrical signal, and the signal conditioning circuit amplifies the signal and digitizes the thrust value. Effective propulsion power is calculated by multiplying the thrust value by the propeller velocity. The propeller velocity is calculated using propeller momentum theory, which states that the thrust of the propeller accelerating the water flow is equal to the rate of change of water momentum. The velocity is equal to the thrust divided by the square root of the product of the water density and the propeller disk area. The thrust multiplied by the velocity is the power of the thruster doing work on the water flow. Fluid resistance power consumption includes the power consumed by the resistance encountered by the propeller casing, fairing, and propeller when moving in water. Fluid resistance power consumption is calculated using a lookup table created through CFD simulation. The input to the lookup table is water density, propeller speed, and propeller diameter, and the output is the fluid resistance power consumption under the corresponding operating condition. The lookup table data comes from a large number of simulation conditions covering the actual operating range of the propeller. Motor copper loss is calculated by measuring the motor phase current and phase resistance. The single-phase copper loss is obtained by multiplying the square of the phase current by the phase resistance. The three-phase copper loss is calculated by multiplying the single-phase copper loss by 3. The phase resistance is measured as a reference value at room temperature using a bridge voltammeter. During operation, the phase resistance is corrected according to the winding temperature. The correction formula is: operating resistance equals reference resistance multiplied by 1 plus the temperature coefficient multiplied by the temperature difference. The temperature coefficient of copper is 0.00393 per degree Celsius. Motor iron loss calculation is based on the hysteresis loop and eddy current loss characteristics of the motor core material. Iron loss is related to the motor's operating frequency and magnetic flux density. The frequency is calculated by dividing the motor speed by 60 and then multiplying by the number of pole pairs. The magnetic flux density is calculated using motor design parameters and input voltage. Iron loss equals the iron loss coefficient multiplied by the frequency to the power of 1.3 and then multiplied by the square of the magnetic flux density. The iron loss coefficient is determined through motor bench testing. Subtracting the effective propulsion power, fluid resistance power consumption, motor copper loss, and motor iron loss from the motor input power leaves the bearing friction power consumption, which is generated by propeller bearing friction.

[0078] The bearing friction coefficient is calculated by inversely based on the relationship between frictional power consumption and frictional force. Bearing frictional force equals the friction coefficient multiplied by the radial load, while frictional power consumption equals the frictional force multiplied by the frictional velocity. The frictional velocity is the linear velocity of the relative motion between the inner and outer rings of the bearing, which is equal to the propeller angular velocity multiplied by the bearing radius. The propeller angular velocity is calculated by multiplying the propeller speed by twice pi and then dividing by 60 to convert to radians per second. The bearing friction coefficient is obtained by dividing the frictional power consumption by twice pi, the propeller speed, the bearing radius, and the radial load. The radial load includes the radial component of the propeller's gravity and the radial force generated by hydrodynamics. The propeller gravity is calculated by multiplying the propeller mass by gravitational acceleration. The radial component of the hydrodynamic force is obtained through propeller hydrodynamic analysis and is proportional to the thrust; the proportionality coefficient is determined based on the propeller's geometric parameters. The friction coefficient measured under standard operating conditions in a brand-new state of the propeller is stored as the initial friction coefficient in the controller's non-volatile memory. The initial friction coefficient of water-lubricated ceramic bearings is typically between 0.06 and 0.10. Bearing wear is defined as the current coefficient of friction divided by the initial coefficient of friction. A wear value of 1 indicates that the bearing is in brand new condition. A wear value greater than 1 indicates that the bearing has worn down and the coefficient of friction has increased. The higher the wear value, the more severe the wear. A wear value exceeding 1.5 usually indicates that the bearing needs to be replaced.

[0079] In one specific embodiment, step S5 includes:

[0080] A dual-core processor is used to perform magnetic coupling slip feature extraction and power pre-allocation decision, multi-thruster power mutual feedback scheduling and bearing compensation calculation respectively. The execution cycle of the first core is set to 200 microseconds and the execution cycle of the second core is set to 1 millisecond.

[0081] Based on the pre-allocated power value, a power shunting command is generated and sent to the power pre-allocation circuit through a high-speed serial peripheral interface. The interface clock frequency is set to 40MHz. Power status data of other thrusters are collected through the controller LAN bus and the energy transfer power is calculated. The bus baud rate is set to 1Mbps.

[0082] The feedforward compensation power is calculated based on the bearing wear and current operating parameters. The motor driver output is adjusted by the pulse width modulation signal. The pulse width modulation frequency is set to 20kHz and the resolution is 12 bits.

[0083] The pre-allocated power value, energy transfer power, and feedforward compensation power are numerically superimposed to generate the final output power command value for each thruster, which is then transmitted to the corresponding thruster's drive control unit via differential signal lines to execute power allocation control.

[0084] Specifically, the dual-core processor architecture uses an ARM Cortex-M7 processor, integrating two independent processing cores. Each core has its own instruction cache, data cache, and floating-point unit. The two cores share peripheral resources and memory space through an on-chip high-speed bus. The first core is dedicated to magnetic coupling slip feature extraction and power pre-allocation decision-making. The execution cycle is set to 200 microseconds, meaning that the first core completes a full round of data acquisition, feature calculation, and pre-allocation decision-making every 200 microseconds. The choice of execution cycle is based on the Hall sensor sampling frequency of 5000Hz, corresponding to a sampling interval of 0.2 milliseconds. The 200-microsecond execution cycle matches the sampling interval to ensure that each sampled data is processed in a timely manner. The second core is responsible for multi-thruster power mutual feedback scheduling and bearing compensation calculation. The execution cycle is set to 1 millisecond. Mutual feedback scheduling requires communication with other thrusters through the controller area network bus. Communication latency and data processing time require the execution cycle to not be too short. Bearing compensation calculation involves solving complex power balance equations, and the calculation time also requires a sufficient cycle. The 1-millisecond execution cycle strikes a balance between response speed and computational sufficiency. The two cores exchange data through a shared memory region. The first core writes the calculated pre-allocated power value into the shared memory, and the second core reads the value from the shared memory. The shared memory access uses a hardware semaphore mechanism to avoid data conflicts caused by the two cores accessing it at the same time.

[0085] The power shunting command generation calculates the shunting ratio based on the pre-allocated power value. The shunting ratio equals the pre-allocated power value divided by the current output power of the power supply. The shunting ratio is converted into a 10-bit binary number representing a ratio between 0 and 1. This 10-bit binary number provides 1024 resolution levels, each corresponding to approximately 0.1% of the shunting ratio. The command data packet uses a 16-bit format. The first two bits are the command header, identifying this as a power shunting command and not another command. The middle 10 bits contain the shunting ratio data, and the last 4 bits are a cyclic redundancy check (CRC) code. The CRC code is calculated by performing a CRC algorithm on the first 12 bits of data. The receiving end uses the CRC code to verify whether data transmission has errors. The high-speed serial peripheral interface SPI adopts a master-slave mode, with the processor as the master device and the power pre-allocation circuit as the slave device. The interface includes four signal lines: clock line, master-output-slave-in data line, master-in-slave-output data line, and chip select line. The clock frequency is set to 40MHz, which means that each clock cycle is 25 nanoseconds. It takes 16 clock cycles, or 400 nanoseconds, to transmit a 16-bit instruction data packet. Including the chip select setup and hold time, the entire transmission process takes about 1 microsecond. The transmission speed is much faster than the 200-microsecond execution cycle, ensuring that the instruction is delivered in time. The Controller Area Network (CAN) bus adopts a multi-master mode, with each thruster's controller acting as a bus node. The bus baud rate is set to 1 Mbps, meaning 1 megabit of data is transmitted per second. The CAN bus uses differential signaling, providing strong anti-interference capabilities suitable for underwater electromagnetic environments. The bus protocol employs a token ring mechanism, where virtual tokens are passed sequentially according to node addresses. The node holding the token has sending privileges. Each node is allocated a 2-millisecond time window to send its thruster's power status data. The data frame contains five data fields: current output power, reserved power margin, buffer energy, temperature, and load level. Each field is 32 bits, totaling 160 bits. Including the frame header, arbitration segment, control segment, and CRC checksum, the entire data frame is approximately 200 bits. Transmission at 1 Mbps takes 0.2 milliseconds, leaving 1.8 milliseconds to wait for the token to be passed to the next node. The second core collects data from all thrusters and updates the distributed power status matrix, calculating the mutual feedback weights of each thruster pair. Based on these weights, it filters and sorts the data to determine the energy transfer power, which is then written to shared memory for subsequent power aggregation.

[0086] The feedforward compensation power calculation reads the current bearing wear degree from shared memory. The wear degree is calculated and updated by the second core in the previous execution cycle using the power balance method. The second core also reads the current water density and propeller flow rate from the sensor interface, and performs a 1.6 power operation on the wear degree. The power operation is implemented using a lookup table to avoid the time consumption of floating-point operations. The lookup table pre-stores the 1.6 power values ​​corresponding to the wear degree in the range of 0.8 to 2.0 with a step size of 0.01. The actual wear degree is obtained by linear interpolation to obtain the corresponding power value. The power value is multiplied by the first compensation coefficient to obtain the first compensation term. The water density is multiplied by the square of the flow rate, and the square operation is performed directly. The product is then multiplied by the second compensation coefficient to obtain the second compensation term. The two compensation terms are added together to obtain the feedforward compensation power. The pulse width modulation (PWM) signal is generated by adjusting the duty cycle based on the feedforward compensation power. The duty cycle equals the total power after compensation divided by the maximum power of the motor. The duty cycle is converted into a 12-bit binary number, providing 4096 resolution levels, each corresponding to approximately 0.024% duty cycle adjustment accuracy. The PWM frequency is set to 20kHz, meaning each PWM cycle is 50 microseconds. The 12-bit duty cycle controls the duration of the high level of the PWM signal within the 50-microsecond cycle. A duty cycle of 0.5 corresponds to a high level of 25 microseconds and a low level of 25 microseconds. Increasing the duty cycle prolongs the high level time and shortens the low level time. The PWM signal drives the IGBT power transistor of the motor driver. When the level is high, the IGBT conducts current and flows to the motor. When the level is low, the IGBT is turned off, and the current is maintained through the freewheeling diode. The larger the duty cycle, the higher the average voltage of the motor and the greater the output power.

[0087] The power superposition calculation reads three values ​​from shared memory: pre-allocated power, energy transfer power, and feedforward compensation power. All three are expressed in watts. The three values ​​are added directly to obtain the power increment. The current reference output power of the thruster is read from the power state matrix. The reference output power is added to the power increment to obtain the final output power command value. The command value is compared with the rated power of the thruster. If the command value exceeds the rated power, it is limited to the rated power to prevent the thruster from overloading. If the command value is negative, it is limited to zero to prevent the thruster from reversing. The final output power command value is converted into a voltage command and output through a digital-to-analog converter (DAC). The DAC has a 16-bit resolution and provides 65,536 voltage levels. The voltage command range is 0 to 5 volts, corresponding to a power command from 0 to the rated power. The voltage command is transmitted to the drive control unit via differential signal lines. The differential signal lines adopt a twisted-pair structure containing positive and negative signal lines. Differential transmission has strong anti-common-mode interference capability. The differential receiver at the drive control unit restores the differential signal to a single-ended voltage signal. The drive control unit adjusts the power output according to the voltage command. This solves the technical problem of fixed and singular power management strategies in existing technologies that cannot be dynamically optimized. The dual-core parallel processing architecture enables fast-response slip feature extraction and computationally intensive mutual feedback scheduling to be performed simultaneously without interference. The high-speed serial interface and differential signal transmission ensure the real-time performance and reliability of command transmission. The power superposition mechanism comprehensively considers three dimensions—pre-allocation, mutual feedback, and compensation—to achieve multi-objective collaborative optimization.

[0088] The above describes the high-efficiency thruster power management optimization method in the embodiments of this application. The following describes the high-efficiency thruster power management optimization system in the embodiments of this application. Please refer to [link / reference]. Figure 2 One embodiment of the high-efficiency thruster power management optimization system in this application includes:

[0089] The calculation module is used to collect the phase difference, speed difference and load torque of the magnet array, and calculate the slip characteristic value of the magnetic coupling drive through the asymmetric slip response kernel function;

[0090] The shunt module is used to perform differential calculation on the slip characteristic value of the magnetic coupling drive to determine the load change trend, calculate the pre-allocated power value based on the judgment result, and shunt the power corresponding to the pre-allocated power value from the main circuit to the back EMF energy buffer module.

[0091] The transfer module is used to construct the power state matrix of multiple thrusters, calculate the mutual feedback weight based on the buffer energy and load difference in the back EMF energy buffer module, establish a bidirectional energy transfer channel between thrusters, and determine the energy transfer power through the bidirectional energy transfer channel.

[0092] The optimization module is used to back-calculate the bearing friction coefficient using the power balance method, calculate the feedforward compensation power based on the nonlinear mapping of wear degree, and optimize the compensation parameters using a self-learning algorithm.

[0093] The superposition module is used to superimpose the pre-allocated power value, energy transfer power, and feedforward compensation power to generate the output power command for each thruster.

[0094] above Figure 2 The medium- and high-efficiency thruster power management optimization system in this embodiment of the invention is described in detail from the perspective of modular functional entities. The high-efficiency thruster power management optimization device in this embodiment of the invention is described in detail from the perspective of hardware processing.

[0095] Reference Figure 3 This invention also provides a high-efficiency thruster power management optimization device, which can be a server, and its internal structure can be as follows: Figure 3 As shown, the high-efficiency thruster power management optimization device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor, designed as a computer, provides computational and control capabilities. The memory of the high-efficiency thruster power management optimization device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the high-efficiency thruster power management optimization device stores the data corresponding to this embodiment. The network interface of the high-efficiency thruster power management optimization device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0096] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the high-efficiency thruster power management optimization device to which the present invention is applied.

[0097] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the high-efficiency thruster power management optimization method.

[0098] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0099] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a high-efficiency thruster power management optimization device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0100] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high-efficiency thruster power management optimization method, characterized in that, The method includes: Step S1: Collect the phase difference, speed difference, and load torque of the magnet array, and calculate the slip characteristic value of the magnetic coupling transmission using the asymmetric slip response kernel function. This includes: real-time acquisition of the phase difference between the magnet array inside the propeller housing and the magnet array inside the propeller hub using a Hall sensor array, with the sampling frequency set to 5000Hz, to obtain phase difference time series data; measuring the motor output shaft speed using an optical encoder and the actual propeller speed using a magnetic encoder, dividing the difference between the motor speed and the propeller speed by the motor speed to obtain the magnetic coupling transmission slip rate; and using the magnetic coupling... A torque sensor on the shaft connecting the transmission mechanism and the propeller monitors the load torque in real time, with the measurement accuracy set to 0.5% of the full scale, to obtain the load torque data. The phase difference is taken as a sine function value and then raised to the 1.4th power to obtain the first characteristic component. The slip ratio is raised to the 2.1st power to obtain the second characteristic component. The load torque is used as the independent variable of a negative exponential function to calculate the attenuation weight. The first and second characteristic components are multiplied by a preset weight coefficient and then multiplied by the attenuation weight. The three terms are weighted and summed to obtain the magnetic coupling transmission slip characteristic value. Step S2: Perform differential operation on the slip characteristic value of the magnetic coupling drive to determine the load change trend, calculate the pre-allocated power value according to the judgment result, and divert the power corresponding to the pre-allocated power value from the main circuit to the back electromotive force energy buffer module. Step S3: Construct a multi-thruster power state matrix, calculate the mutual feedback weight based on the buffer energy and load difference in the back EMF energy buffer module, establish a bidirectional energy transfer channel between thrusters, and determine the energy transfer power through the bidirectional energy transfer channel; Step S4: Calculate the bearing friction coefficient using the power balance method, calculate the feedforward compensation power based on the nonlinear mapping of wear degree, and optimize the compensation parameters using a self-learning algorithm; Step S5: The pre-allocated power value, the energy transfer power, and the feedforward compensation power are superimposed to generate the output power command for each thruster.

2. The high-efficiency thruster power management optimization method according to claim 1, characterized in that, Step S2 includes: The slip characteristic value of the magnetic coupling drive is sampled in a sliding window with a time window length of 30ms, and each time window contains 150 sampling points to obtain the slip characteristic value time series. Perform a first-order difference operation on the slip eigenvalue time series, calculate the change in slip eigenvalue between adjacent sampling points and divide it by the sampling time interval to obtain the slip eigenvalue change rate; Determine whether the rate of change of the slip characteristic value continuously exceeds 15 seconds and lasts for more than 10ms. When the condition is met, trigger a load change warning signal. Subtract the reference value 3 from the current slip characteristic value, multiply by 0.02 and add 0.1 to obtain the pre-allocation ratio coefficient. Multiply the pre-allocation ratio coefficient by the rated power of the thruster to obtain the pre-allocation power value. The power corresponding to the pre-allocated power value is diverted from the main circuit of the power supply through a fast power switching circuit. The switching circuit uses silicon carbide field-effect transistors and the switching frequency is set to 100kHz. The diverted power is injected into the back EMF energy buffer module, which consists of a supercapacitor group, an electrolytic capacitor group, and a film capacitor group cascaded in three stages.

3. The high-efficiency thruster power management optimization method according to claim 2, characterized in that, The step of performing a first-order difference operation on the slip eigenvalue time series, calculating the change in slip eigenvalue between adjacent sampling points divided by the sampling time interval to obtain the slip eigenvalue change rate, includes: Extract the slip feature value of the current sampling point and the slip feature value of the previous sampling point from the slip feature value time series, calculate the difference between the two, and obtain the change in slip feature value; The sampling time interval between adjacent sampling points is calculated based on the sampling frequency, and the change in the slip characteristic value is divided by the sampling time interval to obtain the instantaneous change rate of a single sampling point. The above difference operation is performed sequentially on all sampling points in the slip characteristic time series to generate an instantaneous rate of change sequence with the same length as the original time series; The instantaneous rate of change sequence is subjected to moving average filtering. A sliding window is set to contain several sampling points. The arithmetic mean of the instantaneous rate of change within the window is calculated to obtain the smoothed slip characteristic value rate of change.

4. The high-efficiency thruster power management optimization method according to claim 1, characterized in that, Step S3 includes: Construct a distributed power state matrix containing n thrusters. Each row of the matrix corresponds to one thruster. The five columns are the current output power, reserved power margin, back EMF buffer energy, motor thermal state, and load level, respectively. The reserved power margin is calculated by subtracting the current output power from the rated power of the thruster, and the load level is calculated by dividing the current load torque by the rated torque. For any two thrusters, the mutual feed-in priority weight is calculated by multiplying the ratio of the reserved power margin of the first thruster to its rated power by a first weighting coefficient to obtain the first weighting term; multiplying the ratio of the back EMF buffer energy of the second thruster to its maximum buffer energy by a second weighting coefficient to obtain the second weighting term; normalizing the ratio of the motor temperature of the first thruster to its temperature limit and multiplying it by a third weighting coefficient to obtain the third weighting term; and multiplying the square of the difference in load levels between the two thrusters by a fourth weighting coefficient to obtain the fourth weighting term. The four terms are then weighted and summed and mapped to the 0 to 1 interval using a sigmoid function to obtain the mutual feed-in priority weight. Iterate through all thruster pairs and calculate their respective mutual feedback priority weights. Filter thruster pairs with weight values ​​greater than the threshold of 0.5 as energy transfer candidates. Sort them by weight from largest to smallest and select the top few thruster pairs. A bidirectional energy transfer channel is established for the selected thruster pair. The energy of the high back EMF buffered energy thruster is converted into voltage and injected into the signal power bus of the low energy thruster through an isolated DC-DC converter. The transfer power is determined according to the minimum value among the power gap of the transmitter, the available power of the receiver, and the rated power of the converter.

5. The high-efficiency thruster power management optimization method according to claim 1, characterized in that, Step S4 includes: The motor input power and thrust output are measured when the thruster is running at a constant speed. The effective propulsion power, fluid resistance power consumption, motor copper loss and iron loss are calculated based on the power balance relationship. The bearing friction power consumption is obtained by subtracting the power consumption from the input power. The bearing friction coefficient is calculated based on the bearing friction power consumption, propeller speed, bearing radius and radial load, and the ratio of the current friction coefficient to the initial friction coefficient is taken as the bearing wear degree. The wear degree is raised to the power of 1.6 and multiplied by the first compensation coefficient to obtain the first compensation term. The product of water density and the square of propeller velocity is multiplied by the second compensation coefficient to obtain the second compensation term. The sum of the two terms is the feedforward compensation power. The objective function is constructed as the sum of squares of the difference between the target thrust and the actual thrust. The first and second compensation coefficients are iteratively updated using the gradient descent method to obtain the optimized compensation parameters.

6. The high-efficiency thruster power management optimization method according to claim 5, characterized in that, Step S5 includes: A dual-core processor is used to perform magnetic coupling slip feature extraction and power pre-allocation decision, multi-thruster power mutual feedback scheduling and bearing compensation calculation respectively. The execution cycle of the first core is set to 200 microseconds and the execution cycle of the second core is set to 1 millisecond. A power shunting command is generated based on the pre-allocated power value. The power shunting command is sent to the power pre-allocation circuit through a high-speed serial peripheral interface. The interface clock frequency is set to 40MHz. Power status data of other thrusters are collected through the controller local area network bus and the energy transfer power is calculated. The bus baud rate is set to 1Mbps. The feedforward compensation power is calculated based on the bearing wear and current operating parameters. The motor driver output is adjusted by a pulse width modulation signal. The pulse width modulation frequency is set to 20kHz and the resolution is 12 bits. The pre-allocated power value, the energy transfer power, and the feedforward compensation power are numerically superimposed to generate the final output power command value for each thruster, which is then transmitted to the corresponding thruster's drive control unit via a differential signal line to execute power allocation control.

7. A high-efficiency thruster power management optimization system, characterized in that, For implementing the high-efficiency thruster power management optimization method as described in any one of claims 1-6, the high-efficiency thruster power management optimization system comprises: The calculation module is used to collect the phase difference, speed difference and load torque of the magnet array, and calculate the slip characteristic value of the magnetic coupling drive through the asymmetric slip response kernel function; The shunt module is used to perform differential calculation on the slip characteristic value of the magnetic coupling drive to determine the load change trend, calculate the pre-allocated power value based on the judgment result, and shunt the power corresponding to the pre-allocated power value from the main circuit to the back EMF energy buffer module. The transfer module is used to construct the power state matrix of multiple thrusters, calculate the mutual feedback weight based on the buffer energy and load difference in the back EMF energy buffer module, establish a bidirectional energy transfer channel between thrusters, and determine the energy transfer power through the bidirectional energy transfer channel. The optimization module is used to back-calculate the bearing friction coefficient using the power balance method, calculate the feedforward compensation power based on the nonlinear mapping of wear degree, and optimize the compensation parameters using a self-learning algorithm. The superposition module is used to superimpose the pre-allocated power value, energy transfer power, and feedforward compensation power to generate the output power command for each thruster.

8. A high-efficiency thruster power management optimization device, characterized in that, It includes a memory and a processor, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement the high-efficiency thruster power management optimization method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute the high-efficiency thruster power management optimization method as described in any one of claims 1 to 6.

Images