Oil-cooled motor oil passage flow prediction regulation method and system

Abstract

The present application relates to the oil-cooled motor thermal management technical field, specifically relates to a kind of oil-cooled motor oil passage flow prediction regulation method and system.The method includes: obtaining the real-time current of oil-cooled motor, each branch flow, oil inlet and oil return temperature;According to the current change in current window, determine the overall relative change trend of motor temperature and short-time relative current, calculate the expected motor temperature index;Calculate the heat state of each branch;The expected motor temperature index and heat state are input into the first prediction model, and the total oil passage flow and each branch flow distribution ratio are output;According to the heat state and short-time relative current, the local heat instantaneous trend is calculated, which is input into the second prediction model together with the output of the first prediction model, and the deviation of the total oil passage flow and the deviation of each branch flow distribution ratio are output, and the flow control instruction is generated and output according to the correction.The present application improves the flow control response speed of oil-cooled motor under transient operating conditions and the distribution accuracy of each branch flow.

Description

Technical Field

[0001] This invention relates to the field of thermal management technology for oil-cooled motors, specifically to a method and system for predicting and controlling the oil flow in an oil-cooled motor. Background Technology

[0002] Oil-cooled motors use cooling oil as the cooling medium. Compared to traditional water-cooling or air-cooling methods, they have higher cooling efficiency and power density, and superior insulation performance, effectively reducing the risk of permanent magnet demagnetization. The cooling effect of an oil-cooled motor directly depends on whether the cooling oil can be delivered to each heat-generating part as needed and evenly. Therefore, it is necessary to adjust the flow rate of each oil channel branch in real time to adapt to complex and changing operating conditions.

[0003] Currently, traditional solutions for oil flow control in oil-cooled motors mainly include rule-based control strategies and model-predictive feedforward control methods. However, rule-based control strategies suffer from insufficient control accuracy and poor flow distribution uniformity. Taking a heavy-duty truck oil-cooled flat wire motor (model 270-580-02, rated / peak power 205 / 330kW, rated / peak torque 327 / 580N·m, peak speed 12000rpm) as an example, in actual bench testing, its stator-rotor loss ratio is approximately 10:1, while the flow distribution design target is 4:1. Test results show that when the oil pump speed decreases, the total oil supply decreases, and the rotor oil circuit cannot meet the minimum heat dissipation flow requirement of 3L / min, leading to a significant increase in the risk of local overheating. Traditional rule-based control strategies are insufficient to address the flow distribution imbalance problem under such transient operating conditions.

[0004] To address the aforementioned issues, current mainstream technologies primarily employ PID closed-loop control or model predictive control for feedforward optimization. However, PID control is prone to response lag under transient conditions such as rapid acceleration, leading to instantaneous overheating of the windings. In the bench test of the aforementioned 270-580-02 motor, when the oil level was below the air gap height, the internal temperature distribution of the motor was significantly uneven, with a large temperature difference between local high and low points. Increasing the oil level reduced the temperature difference somewhat, but the overall temperature remained high, and the increased oil churning loss led to a decrease in motor efficiency, indicating that PID control relying solely on lag feedback is insufficient for precise control under uneven temperature conditions. While feedforward optimization methods based on model predictive control can perform forward-looking adjustments, their control effectiveness is highly dependent on the accuracy of the thermal model. The model establishment and parameter calibration process is complex, making engineering implementation difficult. Summary of the Invention

[0005] This invention provides a method and system for predicting and controlling the oil flow in an oil-cooled motor to solve existing problems.

[0006] The present invention provides a method for predicting and controlling the oil flow rate of an oil-cooled motor, which adopts the following technical solution:

[0007] One embodiment of the present invention provides a method for predicting and controlling the oil flow rate in an oil-cooled motor, the method comprising the following steps:

[0008] The real-time current of the oil-cooled motor, the real-time flow rate of each oil passage branch, the inlet temperature of the cooling oil, and the return temperature of each oil passage branch are obtained.

[0009] For the current window at the current moment, the overall relative change trend of motor temperature is determined based on the real-time current in the current window and the real-time current in the previous adjacent current window. The short-time relative current is calculated based on the real-time current in the current window, and the expected motor temperature index is calculated based on the overall relative change trend of motor temperature and the short-time relative current.

[0010] The heating state of each oil passage branch is calculated based on the product of the real-time flow rate and the return oil temperature rise. The return oil temperature rise of each oil passage branch is determined based on the difference between the cooling oil inlet temperature and the return oil temperature of the oil passage branch.

[0011] The expected motor temperature index and the heating status of each oil passage branch are input into the first prediction model, and the total oil passage flow and the flow distribution ratio of each oil passage branch are output through the first prediction model.

[0012] The instantaneous trend of local heating in each oil passage branch is calculated based on the heating state of each oil passage branch and the short-time relative current.

[0013] The output of the first prediction model and the instantaneous trend of local heating in each oil passage branch are input into the second prediction model. The deviation of the total oil passage flow and the deviation of the flow distribution ratio of each oil passage branch are output by the second prediction model.

[0014] Based on the total oil passage flow rate, flow distribution ratio, deviation of total oil passage flow rate, and deviation of flow distribution ratio, the corrected total oil passage flow rate and the corrected flow distribution ratio of each oil passage branch are calculated, and flow control commands are generated and output.

[0015] Furthermore, for the current window at the current moment, the overall relative trend of motor temperature change is determined based on the real-time current within this current window and the real-time current within the previous adjacent current window, specifically including:

[0016] For the current window at the current moment, calculate the first integral value of the real-time current within the current window, and calculate the second integral value of the real-time current within the previous adjacent current window.

[0017] Calculate the ratio of the first integral value to the second integral value, and take the larger value of the ratio and 1 as the overall relative trend of motor temperature change.

[0018] The length of the current window at the current moment is determined based on the time delay from the start of current change to the temperature change in the oil-cooled motor. The previous adjacent current window is the current window that is adjacent to this current window in time and has no intersection with it. The length of the previous adjacent current window is the same as the length of the current window at the current moment.

[0019] Furthermore, the short-time relative current is calculated based on the real-time current within the current window, and the expected motor temperature index is calculated based on the overall relative change trend of the motor temperature and the short-time relative current, specifically including:

[0020] The short-time relative current is obtained by normalizing the real-time current within the current window and taking the average value.

[0021] The expected motor temperature index is determined by multiplying the overall relative trend of motor temperature change with the short-term relative current.

[0022] Furthermore, the heating status of each oil passage branch is calculated based on the product of the real-time flow rate and the return oil temperature rise, specifically including:

[0023] The product of the real-time flow rate of each oil passage branch and the return oil temperature rise is normalized, and the normalized value is determined as the heating state of that oil passage branch.

[0024] Furthermore, the first and second prediction models are LSTM models.

[0025] Furthermore, the training process of the first prediction model is as follows:

[0026] Historical data where the real-time current change within the current window is less than a preset first threshold is obtained as training samples. The heating state of each oil passage branch and the expected motor temperature index within the current window are used as training inputs. The historical total oil passage flow and the historical flow distribution ratio of each oil passage branch at the time after the current window are used as training labels to train the LSTM model and obtain the first prediction model.

[0027] Furthermore, the training process of the second prediction model is as follows:

[0028] Historical data showing that the real-time current change within a current window exceeds a preset second threshold is used as training samples. The instantaneous trend of local heating in each oil passage branch within the current window, as well as the total oil passage flow and the flow distribution ratio of each oil passage branch output by the first prediction model after inputting the historical data within the current window, are used as training inputs. The difference between the historical total oil passage flow and the total oil passage flow output by the first prediction model at a time after the current window, and the difference between the historical flow distribution ratio of each oil passage branch at a time after the current window and the flow distribution ratio of each oil passage branch output by the first prediction model, are used as training labels to train the LSTM model and obtain the second prediction model.

[0029] Furthermore, based on the heating state of each oil passage branch and the short-time relative current, the instantaneous trend of local heating in each oil passage branch is calculated, specifically including:

[0030] Multiplying the heating state of each oil passage branch by the short-term relative current yields the instantaneous local heating trend of each oil passage branch.

[0031] Furthermore, based on the total oil passage flow rate, flow distribution ratio, deviation of the total oil passage flow rate, and deviation of the flow distribution ratio, the corrected total oil passage flow rate and the corrected flow distribution ratio of each oil passage branch are calculated, specifically including:

[0032] Add the total oil passage flow rate to the deviation of the total oil passage flow rate to obtain the corrected total oil passage flow rate;

[0033] The flow distribution ratio of each oil duct branch is added to the corresponding flow distribution ratio deviation, and the result is normalized to obtain the corrected flow distribution ratio of each oil duct branch.

[0034] This invention proposes an oil-cooled motor oil passage flow prediction and control system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the oil-cooled motor oil passage flow prediction and control method described above.

[0035] The beneficial effects of the technical solution of the present invention are:

[0036] This invention proactively predicts the overall heating trend of the motor by observing the window changes of real-time current. Combined with the heating state represented by the product of the return oil temperature rise and flow rate in each branch, it constructs a dual LSTM collaborative prediction architecture of "long-term trend prediction + short-term fluctuation correction." This allows the system to plan the oil flow rate in advance before the actual temperature rise, fundamentally overcoming the response delay and local overheating problems caused by lag feedback in traditional PID control under transient conditions. Simultaneously, this invention adopts a completely data-driven approach, using easily measurable current, flow rate, and temperature signals to indirectly represent the thermal state of each branch. It eliminates the need for complex motor thermophysical models, avoiding the engineering implementation bottlenecks such as calibration difficulties and model mismatch in model predictive control methods, significantly reducing system deployment difficulty and computational complexity. Furthermore, the architecture of independent training and online collaborative inference of the two models balances the control accuracy under steady-state conditions and the response speed under transient conditions, effectively ensuring the balanced flow distribution of each oil branch under various operating conditions. This reduces the risk of permanent magnet demagnetization and winding insulation damage caused by local overheating, improving motor operational reliability and overall energy efficiency. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 A flowchart illustrating a method for predicting and controlling oil flow in an oil-cooled motor, as provided in an embodiment of the present invention.

[0039] Figure 2 This is a structural diagram of an oil-cooled motor oil passage flow prediction and control system provided in one embodiment of the present invention. Detailed Implementation

[0040] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a method for predicting and controlling oil flow in an oil-cooled motor according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0041] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0042] The following description, in conjunction with the accompanying drawings, details a specific scheme for the oil passage flow prediction and control method for an oil-cooled motor provided by the present invention.

[0043] This invention provides a method and system for predicting and controlling the oil flow rate in an oil-cooled motor. Please refer to [link / reference]. Figure 1 The diagram illustrates a flowchart of a method for predicting and controlling the oil flow rate of an oil-cooled motor according to an embodiment of the present invention. The method includes the following steps:

[0044] S101. Obtain the real-time current of the oil-cooled motor, the real-time flow rate of each oil passage branch, the inlet temperature of the cooling oil, and the return temperature of each oil passage branch.

[0045] For example, this step mainly involves acquiring data related to the heat dissipation status of the oil-cooled motor during operation, specifically including the acquisition and processing of the following parameters:

[0046] Real-time current acquisition: A Hall current sensor is installed on the busbar or cable connecting the inverter output to the three-phase leads of the motor. Using Kirchhoff's current law, the third-phase current is calculated by measuring any two of the three-phase currents, thus obtaining the real-time equivalent total current of the oil-cooled motor. The unit of current is ampere. Hall current sensors have advantages such as fast response, non-contact measurement, and good isolation, making them suitable for use in high-voltage environments in motor drive systems.

[0047] Real-time flow rate acquisition for each oil passage branch: The real-time flow rate of each oil passage branch is directly read from the flow control module, in liters per minute. This flow control module includes an electronic oil pump and / or a flow control valve. There is a known correspondence between its control signal and the current output flow rate (such as the speed and flow characteristic curve of the electronic oil pump, and the opening degree and flow characteristic curve of the flow control valve). Therefore, the real-time flow rate value of each branch can be obtained without the need for an additional flow meter.

[0048] The cooling oil temperature is obtained by measuring two parts: a single-point inlet temperature and a multi-point return temperature. The inlet temperature (single point) is measured by installing a sheathed platinum resistance temperature sensor on the pipeline between the oil cooler outlet and the branch manifold. The sensor probe penetrates the pipe wall and is inserted into the center of the oil flow cross-section to measure the temperature of the cooling oil after cooling by the oil cooler and before entering each branch oil channel. This temperature is recorded as the inlet temperature of the oil-cooled motor. Sheathed platinum resistance temperature sensors offer advantages such as high accuracy, good stability, and oil and temperature resistance, making them suitable for long-term use in lubricating oil or cooling oil environments. The return oil temperatures (multiple points) are measured by installing sheathed platinum resistance temperature sensors at each return port before each branch oil channel merges into the main return oil pipe. The sensor probe is inserted into the return oil flow to measure the return oil temperature after heat absorption in each branch. For the first... The return oil temperature of the branch oil channel is denoted as... The unit is degrees Celsius. The return oil temperature of each branch reflects the temperature rise of the cooling oil after it flows through the corresponding heat-generating components (such as windings, iron cores, bearings, etc.) and absorbs heat.

[0049] After obtaining the above data, since the sampling frequencies of the current sensor and the temperature sensor are different, a timestamp alignment operation is first performed on all the collected data. Specifically, the data of each channel is interpolated and resampled using a unified time base to ensure that at the same point in time, the current data, the flow data of each branch, the inlet oil temperature, and the corresponding return oil temperature data can strictly correspond, providing a synchronized data foundation for subsequent time series analysis.

[0050] After time alignment, the data undergoes denoising. The current signal contains high-frequency noise generated by inverter switching operations, and the temperature signal may contain quantization noise and random measurement fluctuations. These noises are filtered out using low-pass filtering or moving average filtering to preserve the true trend of the signal. Furthermore, depending on the quality of the actual acquired data, conventional data augmentation processing can be performed, such as fine-tuning the time axis or adding small amounts of noise, to enrich the diversity of the training samples. However, this augmentation is not a necessary step in this method and can be selectively used based on the amount of data in actual engineering deployments.

[0051] S102. For the current window at the current moment, determine the overall relative change trend of the motor temperature based on the real-time current in the current window and the real-time current in the previous adjacent current window, calculate the short-time relative current based on the real-time current in the current window, and calculate the expected motor temperature index based on the overall relative change trend of the motor temperature and the short-time relative current.

[0052] In this embodiment, for the current window at the current moment, the overall relative change trend of the motor temperature is determined based on the real-time current within the current window and the real-time current within the previous adjacent current window, specifically including:

[0053] For the current window at the current moment, calculate the first integral value of the real-time current within the current window, and calculate the second integral value of the real-time current within the previous adjacent current window.

[0054] Calculate the ratio of the first integral value to the second integral value, and take the larger value of the ratio and 1 as the overall relative trend of motor temperature change.

[0055] The length of the current window at the current moment is determined based on the time delay from the start of current change to the temperature change in the oil-cooled motor. The previous adjacent current window is the current window that is adjacent to this current window in time and has no intersection with it. The length of the previous adjacent current window is the same as the length of the current window at the current moment.

[0056] The short-time relative current is calculated based on the real-time current within the current window, and the expected motor temperature index is calculated based on the overall relative change trend of motor temperature and the short-time relative current, specifically including:

[0057] The short-time relative current is obtained by normalizing the real-time current within the current window and taking the average value.

[0058] The expected motor temperature index is determined by multiplying the overall relative trend of motor temperature change with the short-term relative current.

[0059] For example, in an oil-cooled motor, the cooling oil is typically stored in a cooling oil tank when not in use. When the motor starts running and requires cooling, the cooling oil is drawn from the tank by an oil pump and first sent to an oil cooler for cooling, becoming low-temperature oil, before being sent into the motor's internal oil passages. The low-temperature oil is guided through the internal oil passages to the heat-generating parts, absorbing heat at the high-temperature components inside the motor. Finally, the cooling oil carrying the heat flows back to the cooling oil tank, forming a complete cooling cycle.

[0060] The heating of an electric motor is strongly correlated with the current state: the main heat in an electric motor comes from Joule heating during operation, and the higher the current level over a period of time, the more heat is generated. In the test of the 270-580-02 motor, the ratio of stator winding loss (copper loss) to rotor loss (core loss, magnet loss, and wind wear loss) was approximately 10:1, indicating that the impact of current changes on stator temperature rise is much greater than that on rotor temperature rise.

[0061] The instantaneous current of a motor fluctuates dramatically depending on operating conditions: during rapid acceleration, overtaking, and other transient conditions, the current increases significantly in a short period, while the current fluctuation is relatively small during constant speed movement or coasting. During this process, the heating components in the motor generate heat synchronously with the current due to the Joule effect. This heat accumulates over time, eventually causing a change in the temperature of the heating components. As the current increases, the motor speed often also increases, leading to a corresponding increase in eddy current losses and frictional losses, which are indirectly related to the current magnitude.

[0062] After the motor current responds to transient operating conditions, the temperature of the heating components does not change immediately. Instead, there is a delayed temperature rise due to the heat capacity effect. This delayed temperature rise allows for the planning of oil flow rate in advance before the actual temperature increase. It should also be noted that when the motor current decreases, only the generated heat decreases relatively; this does not directly lead to a decrease in the temperature of the heating components. The heating components actually maintain their original temperature and only decrease slowly with heat dissipation. Therefore, the oil flow rate cannot be reduced immediately.

[0063] Based on the above analysis, this method uses real-time motor current to predict the subsequent temperature change trend of the heating component. The specific calculation method is as follows:

[0064] A local current time window is established on the sequence of current changes over time: starting from the current moment, the window is divided in reverse chronological order. The length of the current window at the current moment is set to half the time delay from when the motor current begins to change to when the temperature changes. This time delay is obtained by averaging multiple measurements. The previous adjacent current window is the current window that is temporally adjacent to this current window and has no overlap with it. The length of the previous adjacent current window is the same as the length of the current window at the current moment.

[0065] After dividing the window, calculate the integral value of the real-time current in the current window (denoted as the first integral value), and then calculate the integral value of the real-time current in the previous adjacent window (denoted as the second integral value). Calculate the ratio of the first integral value to the second integral value, and record this ratio as the instantaneous current change trend. .

[0066] Obtain the instantaneous current change trend After that, take The larger value in 1 is denoted as the overall relative trend of motor temperature change. . That is, take .

[0067] The physical significance of this step is that when the current decreases (the instantaneous current change trend is less than 1), although the heat generation is relatively reduced, the heating components still maintain a relatively high actual temperature and will not cool down immediately. Therefore, the oil flow supply should not be reduced immediately. By taking a larger value, the overall relative change trend of the motor temperature is made equal to 1, that is, the current flow level is maintained without decreasing. This design ensures that the system will not erroneously reduce the oil quantity prediction at the moment the current drops, avoiding the risk of residual heat not being able to dissipate in time due to premature reduction in flow.

[0068] Real-time current at each moment , change the current value from interval compression to The interval, in which This represents the maximum possible current of the motor, a value calibrated and fixed in the electronic control unit at the factory and not updated during operation. The compressed current is averaged within the current current window and recorded as the short-time relative current. This value is dimensionless.

[0069] Motor heating primarily originates from Joule heating of the windings, and the heat generation is directly proportional to the square of the current. To more accurately reflect the nonlinear effect of current changes on heat generation in the input characteristics, this embodiment employs a square compression method: reducing the square value of the real-time current from... Interval linear compression to The mean value is taken after the interval. The squared compression method enables the input features to directly correspond to the physical laws of Joule heating, and has higher prediction accuracy under transient conditions with drastic current fluctuations.

[0070] In other embodiments, linear compression can also be used, that is, the real-time current is compressed from... interval compression to The mean value is taken after the interval. The difference between the two compression methods lies in the different mapping functions. Linear compression reflects the changing trend of the current itself, while square compression reflects the changing trend of the heat generation. The purpose of both is to make the current value dimensionless, so that it can be integrated with the overall relative change trend of the motor temperature on a unified scale. In practical applications, an appropriate compression method can be selected according to the motor's heating characteristics and operating conditions, and both can achieve the predictive control purpose of this invention.

[0071] In this embodiment of the invention, "normalization" refers to scaling the data proportionally to a certain scale. The interval is designed to eliminate the influence of dimensions and allow different physical quantities to be calculated on the same scale. In this embodiment, the calculation of short-time relative current adopts square compression as the preferred method, that is, the square value of the current is linearly compressed to a value based on Joule's thermophysical laws. The interval. In other embodiments, other normalization methods such as linear compression, maximum-minimum normalization, or Z-score normalization can also be used, as long as the current value can be mapped to... An interval is sufficient. Regardless of the normalization method used, the goal is to make the current value dimensionless so that it can be integrated with the overall relative change trend of the motor temperature on a unified scale.

[0072] short-time relative current relative trend of motor temperature The product of these is used as the expected motor temperature index. Expected motor temperature index It integrates two dimensions: the current current level (short-term relative current) and the current change trend (overall relative change trend of motor temperature): even if the current current level is not high ( Smaller, but if the current is rising sharply ( (The temperature is relatively high), and the expected motor temperature index will also increase accordingly, thus indicating the upcoming temperature rise in advance and providing a forward-looking basis for subsequent flow control.

[0073] S103. Calculate the heating state of each oil passage branch based on the product of the real-time flow rate and the return oil temperature rise of each oil passage branch. The return oil temperature rise of each oil passage branch is determined based on the difference between the inlet temperature of the cooling oil and the return oil temperature of the oil passage branch.

[0074] In this embodiment, the heating state of each oil passage branch is calculated based on the product of the real-time flow rate and the return oil temperature rise, specifically including:

[0075] The product of the real-time flow rate of each oil passage branch and the return oil temperature rise is normalized, and the normalized value is determined as the heating state of that oil passage branch.

[0076] For example, in the preceding steps, the overall temperature change of the heat-generating components was estimated using real-time motor current. In existing oil-cooled motors, a multi-loop design is commonly used to divide the cooling oil into multiple independent branches, each precisely cooling different structures. The expected motor temperature index as a whole is insufficient to support differentiated flow control for each branch.

[0077] In the actual operation of oil-cooled motors, when the overall motor temperature is expected to rise based on real-time current prediction, the heat-generating components corresponding to each oil passage branch will not generate equal amounts of heat synchronously and uniformly. Specifically, the structural thermal characteristics of the heat-generating components corresponding to each oil passage branch are different: for example, the windings are mainly characterized by copper losses and have relatively small heat capacity, resulting in a rapid temperature rise when the current increases; while the losses of the iron core and bearings have different sources and heat capacity, leading to a relatively gradual temperature rise. Therefore, the same expected motor temperature index translates into actual heat dissipation requirements at different oil passage branches at varying rates, and the expected motor temperature index cannot be simply used as a uniform basis for flow control in each branch of the oil passage.

[0078] The actual thermal state of the heat-generating structures corresponding to each oil channel branch differs. The return oil temperature rise, which is the difference between the return oil temperature when it leaves the branch and the inlet oil temperature when it enters, reflects the degree of matching between the oil flow rate and the actual heat dissipation demand in that branch. It represents the heat dissipation already completed by the cooling oil and lags further in time compared to the lag in component temperature rise. Specifically, in immersion oil cooling, for a single branch, as the oil flow rate increases, the total heat exchange increases before the marginal effect. However, with the increase in flow rate, the cooling oil velocity also increases, shortening the residence time of the oil in the oil channel. This results in less heat absorption per unit volume of cooling oil, leading to less sufficient heat exchange and a smaller temperature difference before and after the cooling oil. Therefore, the thermal state of the heat-generating structure corresponding to each branch can be determined by the oil flow rate and return oil temperature rise.

[0079] Based on the above analysis, this method establishes a long-term window, building upon the expected motor temperature index estimated in the previous steps, to analyze the specific demand for flow regulation in each oil passage branch. The specific calculation method is as follows:

[0080] Calculation of return oil temperature rise: First, obtain the cooling oil temperature downstream of the oil cooler as the inlet oil temperature. For each oil passage branch, obtain the cooling oil temperature at the return oil inlet of that branch as the return oil temperature. Subtract the inlet oil temperature from the return oil temperature of that branch to obtain the return oil temperature rise of that branch. The return oil temperature rise reflects the temperature increase of the cooling oil after it flows through the corresponding heat-generating component of that branch and absorbs heat; it is a direct representation of the actual heat dissipation of that branch.

[0081] Calculation of heating state: Multiply the return oil temperature rise of each oil channel branch by the real-time flow rate of that branch to obtain the product. From a heat transfer perspective, the heat absorbed by the cooling oil in the branch is equal to the product of the cooling oil's mass flow rate, specific heat capacity, and return oil temperature rise. Under the condition that the cooling oil density and specific heat capacity are basically stable, the heat carried away by the cooling oil is proportional to the product of the flow rate and the return oil temperature rise. Therefore, the product of the flow rate and the return oil temperature rise is proportional to the actual heat carried away by the cooling oil in that branch, that is, the actual heat dissipated by the corresponding heat-generating component in that branch.

[0082] Furthermore, under the same heat generation conditions, if the product of the flow rate and the return oil temperature rise in a certain branch is small, it indicates that the cooling oil in that branch carries away less heat, meaning that the corresponding heat-generating component in that branch has a higher temperature and insufficient heat dissipation; conversely, if the product is large, it indicates sufficient heat dissipation and a lower component temperature. Therefore, the product of the flow rate and the return oil temperature rise can be used as an indirect measure of the relative temperature state of the corresponding heat-generating component in that branch.

[0083] The product result is normalized, and the normalized value is determined as the heating state of the oil passage branch. .

[0084] The upper and lower limits of normalization are determined statistically by analyzing the maximum and minimum values ​​in historical operating data (e.g., using the 99th percentile as the upper limit and the 1st percentile as the lower limit). Data exceeding the range is truncated to the boundary values. A higher heat state value indicates a higher temperature of the corresponding heat-generating component in that branch, and a more urgent need for heat dissipation. In other embodiments, other forms of normalization methods (such as maximum / minimum value normalization, Z-score normalization, etc.) can also be used, as long as the product result can be mapped to... Any range is acceptable, and all of them can achieve the purpose of this invention.

[0085] The above-mentioned method for calculating the heating state does not require the placement of temperature sensors on rotating components such as rotors. The thermal state of the corresponding heating components in each branch can be indirectly inferred simply by measuring the temperature and flow rate at the oil return port, thus overcoming the engineering challenge of difficult rotor temperature measurement in traditional solutions.

[0086] S104. Input the expected motor temperature index and the heating status of each oil passage branch into the first prediction model, and output the total oil passage flow and the flow distribution ratio of each oil passage branch through the first prediction model.

[0087] In this embodiment, the first prediction model is an LSTM model.

[0088] The training process of the first prediction model is as follows:

[0089] Historical data where the real-time current change within the current window is less than a preset first threshold is obtained as training samples. The heating state of each oil passage branch and the expected motor temperature index within the current window are used as training inputs. The historical total oil passage flow and the historical flow distribution ratio of each oil passage branch at the time after the current window are used as training labels to train the LSTM model and obtain the first prediction model.

[0090] For example, at each time point, the heating state of each oil channel branch constitutes a temperature state set, and the number of heating states in the set is the same as the total number of oil channel branches. A moving average is applied to the heating state sequence of each branch to eliminate instantaneous trends. The heating state itself is calculated based on the return oil temperature rise, which is a lagging signal, and its instantaneous changes are not suitable for direct trend judgment. The smoothing window length is designed to be 0.1 times the length of subsequent long-term windows. For example, if the smoothing window length is 3 sampling points, for the heating state value of a branch at a certain time, the average of the heating state values ​​at the previous time, the current time itself, and the next time is taken as the smoothed heating state value of that branch at that time. The set of smoothed heating states is denoted as the set of smoothed heating states of each branch.

[0091] In this invention, in practical implementation, the heating state of each oil passage branch can refer to the heating state after moving average processing. This moving average processing is a preprocessing of the original heating state data, the purpose of which is to eliminate instantaneous measurement noise in the hysteresis signal of return oil temperature rise. The value obtained after smoothing still falls within the scope of the heating state of each oil passage branch.

[0092] Then, starting from the current moment, long time windows are divided in reverse chronological order. The window length is designed to be the time from when the cooling oil enters the oil cooler to when it flows back to the cooling oil tank. This time is obtained by averaging multiple measurements. This forms a sequence of smoothed heating states at various moments within a long time window.

[0093] The first prediction model is constructed based on this, and it is a Long Short-Term Memory (LSTM) network model. The input to this model is a smoothed sequence of heating states at each time point within a long time window, and the expected motor temperature index at each time point obtained in the preceding steps. That is, each time point in the input data contains the smoothed heating state of each branch and the expected motor temperature index at that time, with the time step corresponding to the number of time points included in the long time window. The output of this model is the predicted total oil flow rate and the flow distribution ratio of each branch, with the sum of the flow distribution ratios of each branch being 1.

[0094] The first prediction model is trained using historical data. The selection of training samples follows these principles:

[0095] For each current window in the historical data, the magnitude of the real-time current change within that window is calculated. Specifically, the mean of the absolute values ​​of the current change rate within that window is calculated as the evaluation index. The smaller the index value, the smoother the current change within that window, and the more stable the operating condition; the larger the index value, the more drastic the current fluctuation within that window, indicating a transient operating condition.

[0096] A preset first threshold is used to distinguish between steady-state and transient operating conditions. In this embodiment, the current change amplitudes of all current windows in the historical data are sorted by magnitude, and the median is used as the preset first threshold. That is, half of the historical data with current change amplitudes less than this threshold are selected as steady-state operating condition data. In other embodiments, this threshold can also be directly set to a fixed value based on engineering experience.

[0097] Historical data with current variation amplitudes less than a preset first threshold are selected as training samples for the first prediction model. The heating state of each oil passage branch and the expected motor temperature index within the current window are used as training inputs, and the historical total oil passage flow rate and the historical flow distribution ratio of each oil passage branch at times after the current window are used as training labels to train the LSTM model and obtain the first prediction model.

[0098] Using stable historical data with small current fluctuations for training is to allow the first prediction model to focus on learning the steady-state mapping between the thermal state and the required flow rate under stable operating conditions, eliminating the interference of transient fluctuations on model learning, and ensuring the prediction accuracy of the model under steady-state operating conditions.

[0099] After training is completed, during online operation, a long time window is extracted with the current time as the endpoint. The smoothed heating state set sequence of each time within the window and the expected motor temperature index are input into the first prediction model. The model then outputs the predicted total oil flow rate and the flow distribution ratio of each branch.

[0100] S105. Calculate the instantaneous local heating trend of each oil passage branch based on the heating state of each oil passage branch and the short-time relative current.

[0101] In this embodiment, the instantaneous trend of local heating in each oil passage branch is calculated based on the heating state of each oil passage branch and the short-time relative current, specifically including:

[0102] Multiplying the heating state of each oil passage branch by the short-term relative current yields the instantaneous local heating trend of each oil passage branch.

[0103] For example, in the aforementioned steps, the first prediction model has established a long-term trend analysis and completed preliminary flow allocation based on the thermal state represented by the return oil temperature rise of each oil channel branch. However, the return oil temperature rise itself has a strong hysteresis effect and is only suitable for capturing slowly changing long-term trends. When the oil-cooled motor operates stably, this allocation strategy can well meet the heat dissipation needs of the heat-generating components in each branch; but once the operating state fluctuates rapidly, relying solely on the control of the return oil temperature rise will produce a significant hysteresis and will not be able to respond in time to changes in heat dissipation needs under transient operating conditions. Therefore, it is necessary to further capture the changing trend of heat dissipation needs of each branch in a short period of time to correct the preliminary allocation results.

[0104] The preceding steps have already estimated the expected motor temperature index based on the instantaneous current change trend, using it as a measure of the overall motor heating level. However, it should be noted that the newly generated temperature is not only directly related to the current intensity but also to the actual temperature of the corresponding heat-generating component. For branches with higher heating values, it means that the corresponding heat-generating component is already at a high temperature. As the temperature rises, the conductor resistance increases synchronously, and the same amount of current will generate more Joule heat. Therefore, it is necessary to further increase the oil flow rate in this branch to ensure that the cooling oil can provide sufficient heat dissipation capacity to meet the heat dissipation requirements of this component after the temperature rises.

[0105] This physical law was confirmed in the actual verification of the 270-580-02 motor. In the bench test of this motor, the stator-rotor oil flow distribution was approximately 3.7:1 when the oil pump speed was 3500 rpm, with a rotor oil flow rate of approximately 3 liters per minute, which met the minimum heat dissipation flow requirement. However, when the oil pump speed dropped to 3000 rpm or below, the total oil supply decreased, and the rotor oil flow rate could not meet the minimum heat dissipation requirement of 3 liters per minute, resulting in insufficient rotor heat dissipation. This phenomenon indicates that, based on long-term window prediction, it is necessary to capture the impact of transient factors such as oil pump speed fluctuations on flow distribution through short-term trends and adjust the distribution ratio in a timely manner to ensure that the heat dissipation requirements of each branch are met.

[0106] Based on the above analysis, this method constructs short-term trend characteristics based on the heating state and short-term relative current. The specific calculation method is as follows:

[0107] The heating state value represents the current temperature state of the corresponding heating component in that branch. A higher heating state value indicates a higher temperature for the corresponding heating component. With a constant current level, higher resistance results in more intense heat generation and a more pronounced temperature rise trend. Multiplying the heating state of each oil passage branch by the short-time relative current yields the instantaneous local heating trend of each oil passage branch. .

[0108] The instantaneous trend of localized heating takes into account both the current temperature state and current level of the component: the higher the temperature and the greater the current, the stronger the instantaneous trend of localized heating in that branch, indicating a more urgent pressure for temperature rise in that branch and a need to prioritize obtaining more cooling oil flow. Through this feature, the system can detect in advance the differentiated heat dissipation requirements of each branch under transient operating conditions, providing a basis for decision-making for subsequent short-term adjustments.

[0109] S106. Input the output of the first prediction model and the instantaneous trend of local heating of each oil channel branch into the second prediction model, and output the deviation of the total oil channel flow and the deviation of the flow distribution ratio of each oil channel branch through the second prediction model.

[0110] In this embodiment, the second prediction model is an LSTM model.

[0111] The training process for the second prediction model is as follows:

[0112] Historical data showing that the real-time current change within a current window exceeds a preset second threshold is used as training samples. The instantaneous trend of local heating in each oil passage branch within the current window, as well as the total oil passage flow and the flow distribution ratio of each oil passage branch output by the first prediction model after inputting the historical data within the current window, are used as training inputs. The difference between the historical total oil passage flow and the total oil passage flow output by the first prediction model at a time after the current window, and the difference between the historical flow distribution ratio of each oil passage branch at a time after the current window and the flow distribution ratio of each oil passage branch output by the first prediction model, are used as training labels to train the LSTM model and obtain the second prediction model.

[0113] For example, a short time series is constructed based on the instantaneous local heating trends of each oil channel branch calculated in the aforementioned steps. At each time point, the instantaneous local heating trends of all branches constitute a set of instantaneous local heating trends, and the number of elements in the set is the same as the total number of oil channel branches.

[0114] Starting from the current moment, short time windows are divided in reverse chronological order. The length of a short time window is designed to be half the length of a long time window in the preceding steps, thus forming a sequence of local heating instantaneous trends at each moment within a short time window.

[0115] The training labels are constructed as follows: for each training sample, its corresponding short-time window covers time from time... The transient data segment is input into the already trained first prediction model to obtain the result of the first prediction model at the end of the window. The output shows the predicted total oil flow rate and the predicted flow distribution ratio of each branch. This data is obtained from historical records at the same time point. The actual total oil flow rate and actual flow distribution ratio are calculated. The difference between the actual total oil flow rate and the predicted total oil flow rate output by the first prediction model is used as the total oil flow rate deviation label; the difference between the actual flow distribution ratio and the predicted flow distribution ratio of each branch output by the first prediction model is used as the flow distribution ratio deviation label of each branch. These differences are used as the training labels for the second prediction model. Through training with a large number of transient historical samples, the second prediction model learns the mapping relationship between the instantaneous trend of local heating and the deviation of the first prediction model.

[0116] A second prediction model is constructed based on this. The input to this model is a sequence of localized heating instantaneous trends at each moment within a short time window, and the total oil flow rate output by the first prediction model. That is, each moment in the input data includes the localized heating instantaneous trends of each branch and the total oil flow rate at that moment, with the time step corresponding to the number of moments included in the short time window. The output of this model is the deviation value relative to the prediction result of the first prediction model, specifically including the deviation in total oil flow rate and the deviation in the flow distribution ratio of each branch.

[0117] The second prediction model is trained using historical operating data, and the selection principle for training samples is complementary to that of the first prediction model. For each current window in the historical data, the same evaluation metric as the first prediction model is used, namely, calculating the mean of the absolute values ​​of the current change rate within that window. A preset second threshold is used to distinguish between transient and steady-state operating conditions. In this embodiment, the current change amplitudes of all current windows in the historical data are sorted by magnitude, and the median is used as the preset second threshold. That is, half of the historical data with current change amplitudes greater than this threshold are selected as transient operating condition data. This median is the same as the preset first threshold used in the first prediction model; both are the same value. In other embodiments, the first and second thresholds can also be set independently to different values ​​based on engineering experience, as long as the first threshold is lower than the second threshold. The specific values ​​can be determined based on the actual operating condition distribution characteristics of the motor.

[0118] Historical data with current fluctuations exceeding a preset second threshold are selected as training samples for the second prediction model. It should be noted that this transient data was not used to train the first prediction model, which had not encountered such drastic fluctuations during training; therefore, it inevitably exhibits prediction bias under transient conditions. The second prediction model corrects the first prediction model by learning the patterns of this bias.

[0119] The training input is constructed as follows: historical data within the current window is input into the already trained first prediction model. The total oil flow rate and the flow distribution ratio of each oil branch output by the first prediction model are extracted and used together with the instantaneous local heating trend of each oil branch as training input. The training labels are constructed as follows: the historical true total oil flow rate and the historical true flow distribution ratio of each branch at times after the current window are obtained. The difference between the historical true values ​​and the output value of the first prediction model is calculated, and this difference is used as the training label. Through training with a large number of transient historical samples, the second prediction model learns the mapping relationship between the instantaneous local heating trend and the deviation of the first prediction model.

[0120] S107. Based on the total oil passage flow rate, flow distribution ratio, deviation of total oil passage flow rate, and deviation of flow distribution ratio, calculate the corrected total oil passage flow rate and the corrected flow distribution ratio of each oil passage branch, thereby generating and outputting flow control commands.

[0121] In this embodiment, based on the total oil passage flow rate, flow distribution ratio, deviation of the total oil passage flow rate, and deviation of the flow distribution ratio, the corrected total oil passage flow rate and the corrected flow distribution ratio of each oil passage branch are calculated, specifically including:

[0122] Add the total oil passage flow rate to the deviation of the total oil passage flow rate to obtain the corrected total oil passage flow rate;

[0123] The flow distribution ratio of each oil duct branch is added to the corresponding flow distribution ratio deviation, and the result is normalized to obtain the corrected flow distribution ratio of each oil duct branch.

[0124] For example, after obtaining the prediction result of the second prediction model, the prediction result of the first prediction model is corrected using this result. Specifically, the deviation between the total oil passage flow rate output by the first prediction model and the total oil passage flow rate output by the second prediction model is added to obtain the corrected total oil passage flow rate; the deviations between the flow distribution ratios of each branch output by the first prediction model and the flow distribution ratios of each branch output by the second prediction model are added to obtain the corrected flow distribution ratios of each branch. Simultaneously, the corrected flow distribution ratios of each branch are normalized so that the sum of the flow distribution ratios of all branches is 1. This yields the final total oil passage flow rate and the flow distribution ratios of each branch in the oil-cooled motor.

[0125] Based on the corrected total oil passage flow rate and the flow distribution ratio of each branch, the target flow rate of each oil passage branch can be calculated: the target flow rate of each branch is equal to the corrected total oil passage flow rate multiplied by the corrected flow distribution ratio of that branch.

[0126] Then the oil flow control execution phase begins:

[0127] The target flow rate values ​​of each oil passage branch are converted into specific control commands, which are then transmitted to the corresponding flow regulating valves or electronic oil pumps through the signal output interface of the electronic control unit. The control commands can be in the form of PWM signals, CAN bus signals, or analog signals (such as 4-20mA or 0-10V), depending on the interface type of the actuator.

[0128] Upon receiving a control command, the flow control valve or electronic oil pump executes the corresponding action. For the flow control valve, it drives the valve core to move, changing the opening degree of each oil outlet on the valve body, thereby regulating the oil flow in each branch. For the electronic oil pump, it adjusts the motor speed to change the total oil supply and the flow distribution in each branch. The response time of the actuator should be less than one-tenth of the thermal time constant of the oil-cooled motor to ensure real-time control.

[0129] Various sensors installed in the system (current sensors, temperature sensors, etc.) continuously monitor the system and transmit the monitoring data back to the electronic control unit in real time. The electronic control unit repeats all the above steps, from data acquisition, trend prediction, deviation correction to command output, forming a continuous closed-loop predictive control to ensure that the flow distribution of each oil passage branch of the oil-cooled motor always meets the heat dissipation requirements under various operating conditions.

[0130] In a specific embodiment of the 270-580-02 motor, the actuator for oil flow regulation includes an oil pump, a flow regulating valve, and an oil-water heat exchanger. The oil circuit structure includes a three-section stator oil circuit and a rotor cooling scheme, with the rotor oil circuit nozzle having a diameter of 4 mm. The oil sump position has been laterally shifted from below the motor side to directly below the motor to optimize oil return smoothness. These hardware configurations provide the physical basis for the execution of flow prediction and regulation commands. It should be noted that the specific model and hardware parameters of the above-mentioned actuator are only a specific application example of the present invention on the 270-580-02 motor. In other embodiments, actuators of different brands and parameters may be used according to the actual motor structure and selection, without affecting the implementation of the method of the present invention.

[0131] In summary, in this embodiment of the invention, by real-time acquisition of motor current, flow rate of each branch, and temperature data, and by utilizing the current window integral ratio and taking the larger value operation, an overall relative trend of motor temperature change with unidirectional safety characteristics is constructed. The expected motor temperature index is obtained based on the product of the normalized mean current and the temperature change trend, achieving a forward-looking prediction of the overall heating trend. On this basis, the heating state of each branch is characterized by the product of the return oil temperature rise and the branch flow rate. A first prediction model is constructed in combination with the expected motor temperature index, realizing preliminary oil passage flow allocation based on long-term trends. Then, the instantaneous trend of local heating is constructed by the product of the heating state and the short-term relative current. A second prediction model is constructed in combination with the output of the first prediction model, realizing the transient deviation correction of the preliminary allocation results. Finally, the corrected total oil passage flow rate and the flow allocation ratio of each branch are obtained by deviation superposition and normalization, and a control command output is generated. The entire solution utilizes a dual-model collaborative architecture of "long-term trend prediction + short-term fluctuation correction" to balance steady-state control accuracy and transient response speed without relying on a precise thermophysical model. This effectively overcomes the problems of lag in traditional PID control and difficulty in model predictive control calibration, and achieves accurate prediction and control of the flow rate of each oil passage branch of the oil-cooled motor.

[0132] This invention also proposes an oil passage flow prediction and control system for oil-cooled motors; please refer to [link / reference]. Figure 2 The diagram shows a structural diagram of an oil-cooled motor oil passage flow prediction and control system provided in an embodiment of the present invention. The system includes: a data acquisition module 101, a data processing module 102, and a flow control module 103.

[0133] The data acquisition module 101 is used to acquire the real-time current of the oil-cooled motor, the real-time flow rate of each oil passage branch, the inlet temperature of the cooling oil, and the return temperature of each oil passage branch.

[0134] The data processing module 102 is used to determine the overall relative change trend of motor temperature based on the real-time current in the current window at the current moment and the real-time current in the previous adjacent current window, calculate the short-time relative current based on the real-time current in the current window, and calculate the expected motor temperature index based on the overall relative change trend of motor temperature and the short-time relative current.

[0135] The heating state of each oil passage branch is calculated based on the product of the real-time flow rate and the return oil temperature rise. The return oil temperature rise of each oil passage branch is determined based on the difference between the cooling oil inlet temperature and the return oil temperature of the oil passage branch.

[0136] The expected motor temperature index and the heating status of each oil passage branch are input into the first prediction model, and the total oil passage flow and the flow distribution ratio of each oil passage branch are output through the first prediction model.

[0137] The instantaneous trend of local heating in each oil passage branch is calculated based on the heating state of each oil passage branch and the short-time relative current.

[0138] The output of the first prediction model and the instantaneous trend of local heating in each oil passage branch are input into the second prediction model. The deviation of the total oil passage flow and the deviation of the flow distribution ratio of each oil passage branch are output by the second prediction model.

[0139] The flow control module 103 is used to calculate the corrected total oil passage flow and the corrected flow distribution ratio of each oil passage branch based on the total oil passage flow, flow distribution ratio, deviation of total oil passage flow, and deviation of flow distribution ratio, thereby generating and outputting flow control commands.

[0140] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the oil-cooled motor oil passage flow prediction and control system and the oil-cooled motor oil passage flow prediction and control method embodiment provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiment, which will not be repeated here.

[0141] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0142] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0143] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting and controlling the oil flow rate in an oil-cooled motor, characterized in that, include: The real-time current of the oil-cooled motor, the real-time flow rate of each oil passage branch, the inlet temperature of the cooling oil, and the return temperature of each oil passage branch are obtained. For the current window at the current moment, the overall relative change trend of motor temperature is determined based on the real-time current in the current window and the real-time current in the previous adjacent current window. The short-time relative current is calculated based on the real-time current in the current window, and the expected motor temperature index is calculated based on the overall relative change trend of motor temperature and the short-time relative current. The heating state of each oil passage branch is calculated based on the product of the real-time flow rate and the return oil temperature rise. The return oil temperature rise of each oil passage branch is determined based on the difference between the cooling oil inlet temperature and the return oil temperature of the oil passage branch. The expected motor temperature index and the heating status of each oil passage branch are input into the first prediction model, and the total oil passage flow and the flow distribution ratio of each oil passage branch are output through the first prediction model. The instantaneous trend of local heating in each oil passage branch is calculated based on the heating state of each oil passage branch and the short-time relative current. The output of the first prediction model and the instantaneous trend of local heating in each oil passage branch are input into the second prediction model. The deviation of the total oil passage flow and the deviation of the flow distribution ratio of each oil passage branch are output by the second prediction model. Based on the total oil passage flow rate, flow distribution ratio, deviation of total oil passage flow rate, and deviation of flow distribution ratio, the corrected total oil passage flow rate and the corrected flow distribution ratio of each oil passage branch are calculated, and flow control commands are generated and output.

2. The method for predicting and controlling the oil flow rate of an oil-cooled motor according to claim 1, characterized in that, The determination of the overall relative trend of motor temperature change based on the real-time current within the current window and the real-time current within the previous adjacent current window specifically includes: For the current window at the current moment, calculate the first integral value of the real-time current within the current window, and calculate the second integral value of the real-time current within the previous adjacent current window. Calculate the ratio of the first integral value to the second integral value, and take the larger value of the ratio and 1 as the overall relative trend of motor temperature change. The length of the current window at the current moment is determined based on the time delay from the start of current change to the temperature change in the oil-cooled motor. The previous adjacent current window is the current window that is adjacent to this current window in time and has no intersection with it. The length of the previous adjacent current window is the same as the length of the current window at the current moment.

3. The method for predicting and controlling the oil flow rate of an oil-cooled motor according to claim 1, characterized in that, The step of calculating the short-time relative current based on the real-time current within the current window, and calculating the expected motor temperature index based on the overall relative change trend of motor temperature and the short-time relative current, specifically includes: The short-time relative current is obtained by normalizing the real-time current within the current window and taking the average value. The expected motor temperature index is determined by multiplying the overall relative trend of motor temperature change with the short-term relative current.

4. The method for predicting and controlling the oil flow rate of an oil-cooled motor according to claim 1, characterized in that, The calculation of the heating status of each oil passage branch based on the product of the real-time flow rate and the return oil temperature rise specifically includes: The product of the real-time flow rate of each oil passage branch and the return oil temperature rise is normalized, and the normalized value is determined as the heating state of that oil passage branch.

5. The method for predicting and controlling the oil flow rate of an oil-cooled motor according to claim 1, characterized in that, The first and second prediction models are LSTM models.

6. The method for predicting and controlling the oil flow rate of an oil-cooled motor according to claim 5, characterized in that, The training process of the first prediction model is as follows: Historical data where the real-time current change within the current window is less than a preset first threshold is obtained as training samples. The heating state of each oil passage branch and the expected motor temperature index within the current window are used as training inputs. The historical total oil passage flow and the historical flow distribution ratio of each oil passage branch at the time after the current window are used as training labels to train the LSTM model and obtain the first prediction model.

7. The method for predicting and controlling the oil flow rate of an oil-cooled motor according to claim 5, characterized in that, The training process of the second prediction model is as follows: Historical data showing that the real-time current change within a current window exceeds a preset second threshold is used as training samples. The instantaneous trend of local heating in each oil passage branch within the current window, as well as the total oil passage flow and the flow distribution ratio of each oil passage branch output by the first prediction model after inputting the historical data within the current window, are used as training inputs. The difference between the historical total oil passage flow and the total oil passage flow output by the first prediction model at a time after the current window, and the difference between the historical flow distribution ratio of each oil passage branch at a time after the current window and the flow distribution ratio of each oil passage branch output by the first prediction model, are used as training labels to train the LSTM model and obtain the second prediction model.

8. The method for predicting and controlling the oil flow rate of an oil-cooled motor according to claim 1, characterized in that, The calculation of the local heating instantaneous trend of each oil passage branch based on the heating state of each branch and the short-time relative current specifically includes: Multiplying the heating state of each oil passage branch by the short-term relative current yields the instantaneous local heating trend of each oil passage branch.

9. The method for predicting and controlling the oil flow rate of an oil-cooled motor according to claim 1, characterized in that, The process of calculating the corrected total oil passage flow rate and the corrected flow distribution ratio for each oil passage branch based on the total oil passage flow rate, flow distribution ratio, deviation of the total oil passage flow rate, and deviation of the flow distribution ratio specifically includes: Add the total oil passage flow rate to the deviation of the total oil passage flow rate to obtain the corrected total oil passage flow rate; The flow distribution ratio of each oil duct branch is added to the corresponding flow distribution ratio deviation, and the result is normalized to obtain the corrected flow distribution ratio of each oil duct branch.

10. A flow prediction and control system for an oil-cooled motor, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of the oil passage flow prediction and control method for an oil-cooled motor as described in any one of claims 1-9.

Images