Pulse width modulation dynamic optimization method and system combined with artificial intelligence

Abstract

The application discloses a pulse width modulation dynamic optimization method and system combined with artificial intelligence, relates to the technical field of artificial intelligence, and comprises the following steps: obtaining motor operation parameters of a target electric drive system in response to a power transmission system operation signal; performing load prediction based on the motor operation parameters and the power transmission system operation signal to obtain predicted load demand, and generating a plurality of candidate pulse width states according to the predicted load demand; performing operation response analysis and motor influence analysis on the plurality of candidate pulse width states to obtain a plurality of operation response degrees and a plurality of motor influence degrees; obtaining a plurality of pulse width adaptabilities according to the plurality of operation response degrees and the plurality of motor influence degrees, and selecting an optimal pulse width state from the plurality of candidate pulse width states according to the plurality of pulse width adaptabilities; and controlling the target electric drive system based on the optimal pulse width state. The application solves the problem that the conventional motor control method in the prior art is difficult to be adjusted in real time according to the dynamic change of actual operation demand of the motor.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, specifically to a pulse width modulation dynamic optimization method and system that incorporates artificial intelligence. Background Technology

[0002] In electric drive systems, pulse width modulation (PWM) is a core technology for motor drives. Traditional PWM typically employs a fixed PWM strategy, which cannot dynamically adjust based on real-time changes in motor operating parameters and the operating signals of the electric drive system. When the electric drive system receives an operating signal, it can only set the pulse width state based on a fixed PWM strategy, making it difficult to accurately match actual load changes. Specifically, the fixed PWM strategy can lead to inaccurate motor response, failing to achieve the desired operating response effect; furthermore, the fixed PWM strategy only adjusts the pulse width according to preset rules, ignoring the impact of pulse width changes on the motor's own operating state, potentially damaging the motor due to unreasonable pulse width adjustments. Therefore, existing electric drive systems face the technical challenge of failing to optimize and adjust PWM in real-time according to the dynamic changes in the actual operating requirements of the motor. Summary of the Invention

[0003] This application provides a pulse width modulation dynamic optimization method and system that combines artificial intelligence, which solves the technical problem that traditional motor control methods in the prior art are difficult to optimize and adjust in real time according to the dynamic changes in the actual operation requirements of the motor.

[0004] The technical solution to the above-mentioned technical problems in this application is as follows:

[0005] In a first aspect, this application provides a pulse width modulation dynamic optimization method incorporating artificial intelligence, the method comprising:

[0006] Responding to the operation signals of the electric drive system, the operating parameters of the target electric drive system's motor are obtained;

[0007] Load prediction is performed based on the motor operating parameters and the power transmission system operation signals to obtain the predicted load demand, and multiple candidate pulse width states are generated based on the predicted load demand.

[0008] Perform operational response analysis and motor influence analysis on the multiple candidate pulse width states to obtain multiple operational response degrees and multiple motor influence degrees;

[0009] Multiple pulse width fitness values ​​are obtained based on the multiple operational responsiveness values ​​and the multiple motor influence values. The optimal pulse width state is selected from multiple candidate pulse width states based on the multiple pulse width fitness values. The target electric drive system is controlled based on the optimal pulse width state.

[0010] Secondly, this application provides a pulse width modulation dynamic optimization system incorporating artificial intelligence, including:

[0011] The operating parameter acquisition module is used to respond to the operating signals of the electric drive system and acquire the motor operating parameters of the target electric drive system.

[0012] The candidate pulse width state generation module is used to perform load prediction based on the motor operating parameters and the power transmission system operation signal to obtain the predicted load demand, and generate multiple candidate pulse width states according to the predicted load demand.

[0013] The candidate pulse width state analysis module is used to perform operation response analysis and motor influence analysis on the multiple candidate pulse width states, and obtain multiple operation response degrees and multiple motor influence degrees.

[0014] The optimal pulse width state selection module is used to obtain multiple pulse width fitness values ​​based on the multiple operation responsiveness values ​​and the multiple motor influence values, and select the optimal pulse width state from multiple candidate pulse width states based on the multiple pulse width fitness values, and control the target electric drive system based on the optimal pulse width state.

[0015] This application provides one or more technical solutions, which have at least the following technical effects or advantages:

[0016] This application provides a pulse width modulation (PWM) dynamic optimization method and system that incorporates artificial intelligence (AI). By integrating AI technology into the PWM process, it can respond accurately and in real-time to dynamic changes in motor operation requirements. In the load prediction stage, which uses motor operating parameters and power transmission system operating signals, historical data and real-time parameters are used for precise analysis, improving prediction accuracy and avoiding the limitations of fixed parameters in traditional methods. The generation of multiple candidate pulse width states and the selection mechanisms for multiple operation response analyses and motor impact analyses enable the system to achieve a good balance between control response effectiveness and motor protection. Operation response analysis and motor impact analysis evaluate candidate pulse width states from different perspectives, ensuring that the selected optimal pulse width state meets the response requirements of motor operation while minimizing adverse effects on the motor.

[0017] The above technical solutions not only improve the operating efficiency and performance of the motor, but also extend its service life and reduce maintenance costs. Predictive load analysis and multi-candidate solution generation enhance control response speed and accuracy, enabling proactive adaptation to changes in motor operating demands. This ensures effective protection of the motor equipment while pursuing efficient control response, preventing equipment damage caused by aggressive control strategies. It also reduces reliance on manual parameter tuning, improving the intelligence and adaptability of the control system. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the pulse width modulation dynamic optimization method combined with artificial intelligence provided in the embodiments of this application;

[0020] Figure 2 This is a schematic diagram of the structure of the pulse width modulation dynamic optimization system combined with artificial intelligence provided in the embodiments of this application.

[0021] The components represented by each number in the attached diagram are explained below:

[0022] The module includes a parameter acquisition module 11, a candidate pulse width state generation module 12, a candidate pulse width state analysis module 13, and an optimal pulse width state selection module 14. Detailed Implementation

[0023] This application provides a pulse width modulation dynamic optimization method and system that incorporates artificial intelligence, which addresses the technical problem that traditional motor control methods in the prior art are unable to optimize and adjust in real time according to the dynamic changes in the actual operating requirements of the motor.

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0026] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid unnecessarily obscuring the description of this application. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0027] Example 1, as Figure 1 As shown, embodiments of this application provide a pulse width modulation dynamic optimization method incorporating artificial intelligence, including:

[0028] S10: Respond to the operation signal of the electric drive system and obtain the motor operating parameters of the target electric drive system;

[0029] In this embodiment, the system first responds to the operation signal of the electric drive system, which reflects the operator's intention to operate the motor, such as starting or stopping. Then, the operating parameters of the target electric drive system's motor are acquired through sensors and other devices. These parameters include the motor's speed, current, voltage, and temperature.

[0030] Among them, motor operating parameters reflect the current operating status of the motor. For example, motor speed reflects the motor's operating speed, current and voltage reflect the motor's power consumption, and temperature is an important indicator for measuring whether the motor is in normal working condition.

[0031] Then, the process of acquiring the motor operating parameters of the target electric drive system uses high-precision sensors, such as Hall sensors, to accurately measure the motor current and convert the magnetic signal into an electrical signal output, making the measured current value closer to the true value, thereby providing accurate current data for load prediction. The sensors are also regularly calibrated and maintained to ensure the reliability of the acquired data.

[0032] S20: Based on the motor operating parameters and the power transmission system operation signal, perform load prediction to obtain the predicted load demand, and generate multiple candidate pulse width states according to the predicted load demand;

[0033] In this embodiment, firstly, based on the obtained motor operating parameters of the target motor and the operating signals of the electric drive system, a machine learning algorithm is used to perform load prediction, such as a neural network, to train the historical motor operating parameters and the operating signals of the electric drive system and construct a load prediction model.

[0034] Secondly, the acquired motor operating parameters and electric drive system operating signals are input into the trained load prediction model to obtain the predicted load demand. The predicted load demand reflects the load that the motor needs to bear under the current operating signal.

[0035] Finally, multiple candidate pulse width states are generated based on the predicted load demand. Pulse width generation rules are pre-defined, for example, different pulse width generation formulas or value ranges correspond to different load demand ranges. Once the predicted load demand is obtained, multiple potentially applicable candidate pulse width states are generated according to the rules.

[0036] Specifically, step S20 in the method includes:

[0037] Obtain a preset feature template, and extract features from the motor operating parameters based on the preset feature template to obtain the current motor operating features;

[0038] Retrieve the signal encoding table, input the operation signal of the electric drive system into the signal encoding table, and obtain the current operation signal encoding;

[0039] The current motor operating characteristics are used as the first search condition, and the current operation signal encoding is used as the second search condition to jointly retrieve multiple motor load response records;

[0040] Each of the motor load response records includes historical motor operating characteristics, historical operation signal codes, and corresponding historical load requirements. The matching degree between the historical motor operating characteristics and the current motor operating characteristics is greater than a preset feature matching threshold, and the historical operation signal codes are consistent with the current operation signal codes.

[0041] Extract multiple historical load demands from multiple motor load response records, perform aggregated value statistics on the multiple historical load demands, and obtain the predicted load demand.

[0042] In this embodiment, a preset feature template is first obtained, and the motor operating parameters are extracted using the preset feature template. The current motor operating features can reflect the current operating status of the motor. The motor operating parameters include the motor speed, current, voltage, temperature, etc.

[0043] Secondly, the signal encoding table is retrieved, and the operating signals of the electric drive system are converted into the current operating signal code. The electric drive system operating signal code is a digital representation of the operating signals of the electric drive system, which facilitates subsequent retrieval and analysis.

[0044] Next, using the current motor operating characteristics as the first search condition and the current operation signal code as the second search condition, multiple motor load response records are retrieved through a joint search of the database. These records include historical motor operating characteristics, historical operation signal codes, and corresponding historical load demands. By setting two conditions—that the matching degree between historical and current motor operating characteristics is greater than a preset characteristic matching threshold, and that historical and current operation signal codes are identical—historical records with high similarity to the current situation are filtered out.

[0045] For example, with a preset feature matching threshold of 80%, the current motor operating characteristics are: motor speed 1500 rpm, current 20A, voltage 300V, temperature 50℃, and current operation signal code "001". When searching the database, historical record A is found. Its historical motor operating characteristics are: motor speed 1480 rpm, current 21A, voltage 298V, temperature 51℃, and historical operation signal code "001". The calculated matching degree between this historical motor operating characteristic and the current motor operating characteristic is 85%, which is greater than the preset feature matching threshold. Furthermore, the historical operation signal code matches the current operation signal code. Therefore, historical record A will be selected as a motor load response record.

[0046] Furthermore, when the matching degree between the historical motor operating characteristics and the current motor operating characteristics is greater than the preset feature matching threshold, it indicates that the historical motor operating characteristics and the current motor operating characteristics have a high degree of similarity.

[0047] Finally, ensemble value analysis is performed on the extracted historical load demands to find the load demand value that best represents the current situation from a large amount of historical data. Various methods can be used for ensemble value analysis, such as box plot analysis. Box plot analysis can visually display the distribution of data, and by identifying and eliminating outliers, the statistical results can be more accurate. Box plot analysis first sorts the multiple historical load demand data, calculates the median, upper and lower quartiles, and other statistical measures, and then draws a box plot. Based on the box plot, the range of normal data within the retained bins is determined. After removing outliers outside the bins, the data within the bins is then averaged to obtain the predicted load demand.

[0048] For example, suppose we have 8 historical load demand data points: 10, 12, 15, 18, 20, 22, 25, and 100. First, using box plot analysis, we sort the data to obtain: 10, 12, 15, 18, 20, 22, 25, and 100. The median is calculated to be (18+20) / 2 = 19, the lower quartile Q1 is 13.5, the upper quartile Q3 is 23.5, and the interquartile range (IQR) is Q3 - Q1 = 10. According to box plot rules, the lower limit is Q1 - 1.5 × IQR = -1.5, and the upper limit is Q3 + 1.5 × IQR = 38.5. It can be seen that the data point 100 exceeds the upper limit and is an outlier; therefore, data point 100 is removed. Calculate the average of the remaining 7 data points 10, 12, 15, 18, 20, 22, and 25. The result is (10+12+15+18+20+22+25) / 7 = 17. Therefore, the predicted load demand is 17.

[0049] After averaging the data within the enclosure to obtain the predicted load demand, multiple candidate pulse width states are generated according to pre-set pulse width generation rules. These candidate pulse width states are used to select the most suitable pulse width from different possibilities during the subsequent selection process, thereby achieving optimal control of the target electric drive system.

[0050] Furthermore, multiple candidate pulse width states are generated based on the predicted load demand, including:

[0051] Activate the pulse width integration configurator, which includes multiple pulse width configuration units;

[0052] The predicted load demand is input into the pulse width integration configurator, which calls multiple pulse width configuration units to configure the pulse width for the predicted load demand, thereby obtaining the multiple candidate pulse width states.

[0053] In this embodiment, the pulse width integrated configurator is first activated. The multiple pulse width configuration units contained in the activated pulse width integrated configurator are pre-set according to different load conditions and control requirements. The pulse width configuration units can be different machine learning models. Each pulse width configuration unit has unique pulse width generation rules and algorithms, capable of generating appropriate pulse widths for different predicted load requirements.

[0054] Once the predicted load demand is input into the pulse width configuration integrator, multiple pulse width configuration units are automatically invoked to process it. Each pulse width configuration unit generates multiple candidate pulse width states based on its own rules and algorithms, combined with the specific values ​​and characteristics of the predicted load demand. For example, some pulse width configuration units generate pulse widths linearly based on the magnitude of the load demand, while others use a non-linear approach, taking into account the trend of load changes and other relevant factors.

[0055] Different candidate pulse width states represent different control strategies and effects. Some candidate pulse width states may focus more on quickly responding to motor operation needs, enabling the motor to rapidly adjust its output power to meet load changes, but this may cause significant stress on the motor; while other candidate pulse width states may focus more on protecting the motor and reducing damage, but the response speed may be relatively slower. By generating multiple candidate pulse width states, the system can comprehensively consider factors such as operational response and motor protection during subsequent analysis and selection.

[0056] The construction steps of the pulse width integrated configurator include:

[0057] Collect multiple sample load data, construct a sample load dataset, and annotate the pulse width of each sample load data in the sample load dataset to obtain a sample pulse width annotation set;

[0058] Build a multi-configuration unit architecture;

[0059] Using the sample load dataset as input and the sample pulse width annotation set as the supervision target, the multiple configuration unit architectures are trained synchronously to obtain multiple pulse width configuration units.

[0060] The multiple pulse width configuration units are integrated to obtain the pulse width integrated configurator.

[0061] In this embodiment, multiple sample load data are first collected. These sample load data cover the load conditions borne by the motor under different operating scenarios, such as load data under different speeds and different gradients. A sample load dataset is constructed by integrating and classifying the collected sample load data. Pulse width annotation is then performed on each sample load data in the sample load dataset, i.e., the optimal pulse width value corresponding to each sample load data is used to label it, resulting in a sample pulse width annotation set.

[0062] Then, multiple configuration unit architectures are constructed, employing different neural network structures or algorithm models to adapt to varying load conditions and pulse width generation requirements. Using a sample load dataset as input and a sample pulse width annotation set as the supervision target, multiple configuration unit architectures are trained simultaneously. During training, the parameters of the configuration unit architectures are continuously adjusted to ensure the model can accurately generate the corresponding pulse width values ​​based on the input load data. After multiple iterations of training, multiple pulse width configuration units are obtained.

[0063] Finally, multiple pulse width configuration units are integrated, combining the trained pulse width configuration units into a unified pulse width integrated configurator. Based on the input predicted load demand, the pulse width integrated configurator calls different pulse width configuration units to configure the pulse width, thereby generating multiple candidate pulse width states, providing more options for subsequently selecting the optimal pulse width state.

[0064] For example, a pulse width ensemble configurator is built and trained based on a convolutional neural network.

[0065] First, data preparation involves collecting multiple sample load data sets to construct a sample load dataset. The input node of the pulse width integrated configurator is the sample load dataset, which includes historical motor operating characteristics, historical operating signal codes, and corresponding historical load requirements.

[0066] Secondly, model construction involves building a convolutional neural network model, including convolutional layers, pooling layers, and fully connected layers. Convolutional layers extract features from the data, pooling layers reduce the size of the feature data, and fully connected layers convert the feature data into labels for predicting load demand. The input layer has a node count equal to the dimension of the input features. For example, if there are 10 features (historical motor operation features, historical operation signal encoding, and corresponding historical load demand), the input layer contains 10 nodes. One to three hidden layers are set, with the number of nodes in each layer adjusted experimentally (e.g., 64, 32, etc.). The ReLU activation function is used. The output layer has a node count equal to the predicted load demand. For example, predicting only time consumption requires one node, while predicting both time consumption and energy consumption requires two nodes. The output layer generally does not use an activation function and directly outputs continuous values.

[0067] Next, model training uses the predicted load demand as the output. The training framework is constructed using the Adam optimizer and mean squared error loss function, with a sample pulse width annotation set as the supervised objective. The batch size is set to 32, the total training epochs to 50, and an early stopping mechanism (patience = 5) is introduced. If the validation set loss does not decrease for five consecutive epochs, the training process is automatically terminated, resulting in a trained model for identifying modeling element types. This effectively avoids overfitting while ensuring the model reaches convergence. Multiple configuration unit architectures are trained simultaneously to obtain multiple pulse width configuration units.

[0068] Finally, the multiple pulse width configuration units obtained from training are integrated to obtain a pulse width integrated configurator. For example, a voting method is used to integrate the multiple pulse width configuration units.

[0069] Each pulse width configuration unit outputs one or more candidate pulse width states based on the input predicted load demand. The candidate pulse width states output by all pulse width configuration units are aggregated, and the number of votes for each candidate pulse width state is counted. The candidate pulse width state with the most votes is considered the most representative and reliable and is included in the final integration result. Meanwhile, to prevent individual abnormal pulse width configuration unit outputs from having an excessive impact on the final result, a vote threshold is set; only candidate pulse width states with more than this threshold are included in the final integration result.

[0070] S30: Perform operational response analysis and motor influence analysis on the multiple candidate pulse width states to obtain multiple operational response degrees and multiple motor influence degrees;

[0071] In this embodiment, operational response analysis and motor impact analysis are performed on multiple candidate pulse width states. Operational response analysis mainly focuses on the impact of candidate pulse width states on the motor's operational response, such as the motor's response speed and sensitivity to operator operation signals. Motor impact analysis focuses on the impact of candidate pulse width states on the performance and state of the target electric drive system, such as the motor's efficiency, heat generation, and lifespan.

[0072] In terms of operational response analysis, each candidate pulse width state is applied to the control of the target electric drive system. This simulates different operator input signals for the electric drive system, such as start-up and shutdown, and records indicators such as motor response time and response amplitude to measure operational responsiveness. For example, when a start signal is input, the time required for the motor to accelerate from its current stationary state to the target operating state is observed; a shorter time indicates a higher operational responsiveness.

[0073] In terms of motor impact analysis, the degree of motor impact is assessed by monitoring changes in motor operating parameters under different candidate pulse width states, such as motor speed fluctuations, current changes, and temperature increases. For example, if a candidate pulse width state causes excessive motor current, it may lead to severe motor overheating and affect the motor's lifespan; in this case, the motor impact of that candidate pulse width state is relatively high.

[0074] Specifically, step S30 in the method includes:

[0075] Select the first candidate pulse width state from the plurality of candidate pulse width states;

[0076] The first candidate pulse width state and the predicted load demand are input into the operation response evaluator to obtain the first operation response degree;

[0077] The current pulse width is extracted from the motor operating parameters, and the first motor influence degree is obtained based on the current pulse width and the state of the first candidate pulse width.

[0078] Following the method of obtaining the first operational response degree and the first motor influence degree of the first candidate pulse width state, the operational response degree and motor influence degree of the remaining candidate pulse width states are obtained, resulting in multiple operational response degrees and multiple motor influence degrees.

[0079] In this embodiment, firstly, one of multiple candidate pulse width states is randomly selected as the first candidate pulse width state. The operation response estimator, trained with experimental data, can simulate the motor's response to different operation signals during actual operation based on the input first candidate pulse width state and predicted load demand. During the simulation, the operation response estimator considers various factors such as the motor's power system characteristics, transmission system characteristics, and operator habits, and calculates using algorithms and models to ultimately output the first operation response degree. The first operation response degree can be represented by a specific numerical value or level; a higher value or level indicates a better motor operation response under that candidate pulse width state.

[0080] Secondly, for obtaining the motor influence degree, the current pulse width is the pulse width currently being used by the motor. By comparing the current pulse width with the first candidate pulse width state, the impact of the difference between the two on the motor operating parameters is analyzed. For example, if the first candidate pulse width state is significantly larger than the current pulse width, it may lead to a sudden increase in motor current, aggravated speed fluctuations, and a faster temperature rise. Based on the changes in motor operating parameters, the first motor influence degree is calculated using pre-set evaluation rules and algorithms.

[0081] Finally, after obtaining the operational response and motor impact of the first candidate pulse width state, the remaining candidate pulse width states are processed in the same manner. That is, each time, one of the remaining candidate pulse width states is selected, and its operational response is obtained by inputting the predicted load demand into the operational response evaluator. Simultaneously, the current pulse width is compared with that candidate pulse width state to obtain the motor impact. Through this iterative process, multiple operational response and motor impact values ​​corresponding to all candidate pulse width states are ultimately obtained.

[0082] For example, suppose there are 3 candidate pulse width states, namely pulse width A, pulse width B and pulse width C. The maximum score for operation responsiveness is 100 points. The higher the score, the better the operation response of the motor under the candidate pulse width state.

[0083] First, pulse width A is selected as the first candidate pulse width state. Pulse width A and the predicted load demand are input into the operation response evaluator. The operation response evaluator simulates the motor's response under different operation signals, and the operation response score corresponding to pulse width A is determined to be 80 points. Currently, the motor's operating pulse width is 20%, and pulse width A is 30%. By analyzing the impact of the difference between the two on the motor's operating parameters, it is found that the motor current increases slightly, the speed fluctuation is within an acceptable range, and the temperature rise is not significant. Based on the evaluation rules, the motor impact of pulse width A is determined.

[0084] Select pulse width B and input it along with the predicted load demand into the operation response evaluator to obtain an operation response score of 70. With pulse width B at 15%, compared to the current pulse width of 20%, the motor current decreases, speed fluctuations are smaller, and the temperature decreases. The motor impact of pulse width B can be queried.

[0085] Selecting pulse width C, the operation response score is 90 points, and pulse width C is 35%. Compared with the current pulse width, the motor current increases significantly, the speed fluctuates significantly, and the temperature rises rapidly. The motor influence of C can be obtained by querying.

[0086] The above calculations of operational responsiveness and motor impact data provide important reference for selecting the most suitable pulse width, so as to maximize the protection of the target electric drive system, extend the service life of the motor, and improve the operating efficiency of the motor while meeting the motor's operational responsiveness requirements.

[0087] The construction steps of the operation response evaluator include:

[0088] Based on historical control data of motors of the same model, sample pulse width parameter sets and sample load demand sets were collected, and the operation response effects obtained under different sample pulse width parameters and sample load demands were collected. The sample operation response degree set was obtained by labeling the operation response effects.

[0089] Construct an operation response evaluator architecture, wherein the input features of the operation response evaluator architecture are pulse width parameters and load requirements, and the output feature is operation response degree;

[0090] The operational response estimator architecture is trained under supervision using the sample pulse width parameter set, sample load requirement set, and sample operational response degree set until convergence, thus obtaining the trained operational response estimator.

[0091] In this embodiment, data acquisition is first performed, collecting sample pulse width parameter sets and sample load demand sets, and then collecting the operational response effects obtained from tests under different sample pulse width parameters and sample load demands. Historical control data for the same model of motor contains information about the motor under various actual operating scenarios. The sample pulse width parameter set covers different pulse width settings, while the sample load demand set reflects the load size borne by the motor under different operating conditions. By testing under different sample pulse width parameters and sample load demands, the operational response effects of the motor are recorded, such as the response time and response amplitude of the motor to operations like loading, unloading, acceleration, and deceleration.

[0092] Secondly, an operational response evaluator architecture is constructed. This architecture is designed to handle input pulse width parameters and load requirements, and output the corresponding operational response degree. The input features are the pulse width parameters and load requirements, and the output feature is the operational response degree, thus quantifying the motor's operational response under different input conditions.

[0093] Finally, after the architecture was built, the operational response estimator architecture was trained under supervision using a sample pulse width parameter set, a sample load requirement set, and a sample operational response set. During training, the model continuously adjusted its parameters to make the predicted operational response as close as possible to the labeled values ​​in the sample operational response set. Through multiple iterations of training, the model reached convergence, meaning that the model's performance no longer showed significant improvement. At this point, the trained operational response estimator was obtained.

[0094] For example, an operational response evaluator is built and trained based on a convolutional neural network.

[0095] First, data preparation is performed, including a sample pulse width parameter set and a sample load requirement set. Operational response results are collected under different sample pulse width parameters and load requirements. The input nodes for the operational response evaluator are the pulse width parameter set and the load requirement set.

[0096] Secondly, model construction involves building a convolutional neural network model, including convolutional layers, pooling layers, and fully connected layers. Convolutional layers extract features from the data, pooling layers reduce the size of the feature data, and fully connected layers convert the feature data into labels for operational response. The input layer has a node count equal to the dimension of the input features. For example, if there are 10 features, such as the sample pulse width parameter set, the sample load requirement set, and the operational response results obtained under different sample pulse width parameters and sample load requirements, then the input layer contains 10 nodes. One to three hidden layers are set, with the number of nodes in each layer adjusted experimentally, such as 64 or 32. The activation function used is ReLU. The output layer has a node count equal to the number of current operational responses. For example, if only time consumption is evaluated, there is one node; if both time consumption and energy consumption are evaluated, there are two nodes. The output layer generally does not use an activation function and directly outputs continuous values.

[0097] Next, the model is trained, and the operational response evaluation is used as the output. Using the sample pulse width parameter set, sample load demand set, and sample operational response set as supervision, a training framework is constructed using the Adam optimizer and mean squared error loss function. The batch size is set to 32, the total number of training epochs is 50, and an early stopping mechanism (patience = 5) is introduced. When the validation set loss does not decrease for 5 consecutive epochs, the training process is automatically terminated, resulting in a trained operational response evaluator. This effectively avoids model overfitting while ensuring the model reaches convergence.

[0098] Once trained, the operation response estimator can accurately predict the motor's operation response based on the input pulse width parameters and load requirements.

[0099] Further, the first motor influence degree is obtained based on the current pulse width and the first candidate pulse width state, including:

[0100] Calculate the deviation between the current pulse width and the first candidate pulse width state to obtain the first pulse width deviation;

[0101] Retrieve a preset pulse width influence table, which records the mapping relationship between different pulse width deviation ranges and the corresponding motor influence.

[0102] Based on the first pulse width deviation, the matching motor influence degree is retrieved from the pulse width influence degree correspondence table to obtain the first motor influence degree.

[0103] In this embodiment of the application, firstly, the deviation between the current pulse width and the first candidate pulse width state is calculated. The value of the current pulse width is subtracted from the value of the first candidate pulse width state, and the absolute value of the difference is the first pulse width deviation. The deviation value reflects the magnitude of the pulse width change.

[0104] Secondly, a pre-defined pulse width influence table is retrieved. This table, derived through extensive experiments and data analysis, records the mapping relationship between different pulse width deviation ranges and their corresponding motor influence. For example, the motor's pulse width is changed, and various operating parameters, such as temperature, current, and speed fluctuations, are monitored. Based on the changes in these parameters, different pulse width deviation ranges are associated with their corresponding motor influence, forming the pulse width influence table. This table covers various scenarios, from minute to large pulse width deviations.

[0105] When querying the corresponding motor influence degree in the pulse width influence degree correspondence table based on the first pulse width deviation, the interval in which the first pulse width deviation is located is first determined.

[0106] For example, if the current pulse width of a motor is 20% and the first candidate pulse width state is 30%, the absolute value of the difference between the current pulse width value of 20% and the first candidate pulse width state value of 30% is 10%, which is the first pulse width deviation. When the first pulse width deviation is 10%, and the pulse width influence degree corresponds to a motor influence degree of 20 points for the 0-15% pulse width deviation range specified in the pulse width influence degree correspondence table, then the first motor influence degree of this motor can be determined to be 20 points.

[0107] The influence of the first motor is obtained by consulting the pulse width influence table, which intuitively reflects the degree of influence of the first candidate pulse width state on the motor's operating state relative to the current pulse width. If the first motor influence is high, it means that the candidate pulse width state may cause significant damage to the motor, and careful consideration is needed when selecting the pulse width; conversely, if the first motor influence is low, the candidate pulse width state may be a more suitable choice, meeting the motor's operational response requirements while also providing good protection for the motor.

[0108] S40: Obtain multiple pulse width adaptability values ​​based on the multiple operation responsiveness values ​​and the multiple motor influence values, select the optimal pulse width state from multiple candidate pulse width states based on the multiple pulse width adaptability values, and control the target electric drive system based on the optimal pulse width state.

[0109] In this embodiment, multiple pulse width fitness values ​​are obtained based on multiple operational responsiveness values ​​and multiple motor influence values. First, the weights of the operational responsiveness values ​​and motor influence values ​​are determined. After calculating the pulse width fitness values ​​of all candidate pulse width states through a weighted calculation method, the candidate pulse width state with the highest pulse width fitness value is the optimal pulse width state, and the target electric drive system is controlled using the optimal pulse width state.

[0110] Specifically, step S40 in the method includes:

[0111] Obtain preset weighting coefficients, which include operation response weights and motor influence weights;

[0112] Based on the operation response weight, the motor influence weight, the multiple operation response degrees, and the multiple motor influence degrees, an adaptive evaluation is performed on multiple candidate pulse width states to obtain multiple pulse width fitness values.

[0113] Select the candidate pulse width state corresponding to the largest pulse width fitness among multiple pulse width fitness values ​​as the optimal pulse width state.

[0114] In this embodiment, firstly, preset weighting coefficients are obtained, including operation response weight and motor influence weight. For scenarios where motor operation response speed is important, the weight of operation response is appropriately increased; for scenarios requiring motor protection and extended motor lifespan, the weight of motor influence can be increased. For example, assuming the weight of operation response is α and the weight of motor influence is β, and α + β = 1. Since a lower motor influence is better, the pulse width fitness for each candidate pulse width state can be calculated using the formula "Pulse Width Fitness = α × Operation Response + β × (1 / Motor Influence)".

[0115] The candidate pulse width state corresponding to the largest pulse width fitness among multiple pulse width fitness values ​​is selected as the optimal pulse width state. This optimal pulse width state is then applied to the control of the target electric drive system. The motor control system converts this optimal pulse width state into a specific control signal and sends it to the motor driver. During motor operation, the motor's operational response and operating status are continuously monitored. If a decrease in motor operational responsiveness is detected, such as a longer response time during acceleration, or abnormal motor operating parameters such as excessively high motor temperature or excessive current fluctuations, the pulse width fitness is recalculated. Based on the new operational responsiveness and motor impact data, combined with preset weighting coefficients, the pulse width fitness of each candidate pulse width state is recalculated, and a new optimal pulse width state is selected. This ensures that the motor can always meet operational response requirements while maximizing motor protection and maintaining optimal operating conditions.

[0116] Further, based on the operation response weight, the motor influence weight, the multiple operation response degrees, and the multiple motor influence degrees, an adaptive evaluation is performed on multiple candidate pulse width states to obtain multiple pulse width fitness values, including:

[0117] Select a target candidate pulse width state from the plurality of candidate pulse width states, and obtain the corresponding target operation response and target motor influence from the plurality of operation response and the plurality of motor influence;

[0118] Based on the operation response weight, the motor influence weight, the target operation response degree, and the target motor influence degree, the target pulse width fitness degree of the target candidate pulse width state is obtained;

[0119] Following the method of obtaining the target pulse width fitness of the target candidate pulse width state, the pulse width fitness of the remaining candidate pulse width states is obtained, resulting in multiple pulse width fitness values.

[0120] In this embodiment of the application, firstly, one of the multiple candidate pulse width states is selected as the target candidate pulse width state. At the same time, the target operation response degree and target motor influence degree corresponding to the target candidate pulse width state are found from the multiple calculated operation response degrees and multiple motor influence degrees.

[0121] Next, based on the obtained operation response weight, motor influence weight, target operation response degree and target motor influence degree, the target pulse width fitness degree of the target candidate pulse width state is calculated using the weighted calculation formula, namely "pulse width fitness = operation response weight × operation response degree + motor influence weight × (1 / motor influence degree)".

[0122] For example, suppose there are three candidate pulse width states, with an operational responsiveness weight of 0.6 and a motor influence weight of 0.4. The operational responsiveness of the first candidate pulse width state is 80 points and the motor influence is 20; the operational responsiveness of the second candidate pulse width state is 70 points and the motor influence is 10; and the operational responsiveness of the third candidate pulse width state is 90 points and the motor influence is 30.

[0123] The pulse width fitness of the first candidate pulse width state is calculated according to the formula as follows: 0.6×80+0.4×(1 / 20)=48+0.02=48.02; the pulse width fitness of the second candidate pulse width state is: 0.6×70+0.4×(1 / 10)=42+0.04=42.04; the pulse width fitness of the third candidate pulse width state is: 0.6×90+0.4×(1 / 30)=54+0.012=54.012.

[0124] Comparing the three pulse width fitness values, 54.012 > 48.02 > 42.04, so the third candidate pulse width state has the largest pulse width fitness, and the second candidate pulse width state has the smallest pulse width fitness. Therefore, the pulse width of the third candidate pulse width state is taken as the optimal pulse width state.

[0125] Then, following the same procedure, one candidate pulse width state is selected each time, and its corresponding operational responsiveness and motor influence are obtained. Then, the pulse width fitness of the candidate pulse width state is calculated by combining the operational responsiveness weight and the motor influence weight. By repeating this process, until all candidate pulse width states have been calculated, multiple pulse width fitness values ​​corresponding to multiple candidate pulse width states are finally obtained.

[0126] In summary, compared to existing technologies, this application first analyzes the operating signals of the electric drive system and the motor operating parameters, and predicts load demand based on historical data to achieve an active control strategy. Then, it uses a pulse width integrated configurator to generate multiple candidate pulse width state schemes, providing rich control options for different operating conditions. Next, the candidate schemes are evaluated from two dimensions: operational response effect and motor impact, ensuring both control performance and equipment protection are considered. Finally, the optimal control parameters are selected through weighted fitness calculation, achieving intelligent dynamic optimization control.

[0127] In summary, the embodiments of this application have at least the following technical effects:

[0128] This application provides a pulse width modulation (PWM) dynamic optimization method and system that integrates artificial intelligence (AI). By incorporating AI technology into the PWM process, it can respond accurately and in real-time to dynamic changes in motor operation requirements. In the load prediction stage, which uses motor operating parameters and power transmission system operating signals, historical data and real-time parameters are used for precise analysis, improving prediction accuracy and avoiding the limitations of fixed parameters in traditional methods. The generation of multiple candidate pulse width states and the selection mechanisms for multiple operation response analyses and motor impact analyses enable the system to achieve a good balance between control response effectiveness and motor protection. Operation response analysis and motor impact analysis evaluate candidate pulse width states from different perspectives, ensuring that the selected optimal pulse width state meets the motor's operational response requirements while minimizing adverse effects on the motor. Through these technical solutions, not only is the motor's operating efficiency and performance improved, but its service life is also extended, and maintenance costs are reduced. Predictive load analysis and the generation of multiple candidate schemes improve the control response speed and accuracy, enabling proactive adaptation to changes in motor operation requirements. This ensures effective protection of the motor equipment while pursuing efficient control response, avoiding equipment damage caused by aggressive control strategies. This reduces reliance on manual parameter adjustment and improves the intelligence and adaptability of the control system.

[0129] Example 2, as Figure 2 As shown, based on the pulse width modulation dynamic optimization method combined with artificial intelligence provided in Embodiment 1 and the same inventive concept, this application also provides a pulse width modulation dynamic optimization system combined with artificial intelligence, including:

[0130] The operating parameter acquisition module 11 is used to respond to the operation signal of the electric drive system and acquire the motor operating parameters of the target electric drive system.

[0131] The candidate pulse width state generation module 12 is used to perform load prediction based on the motor operating parameters and the power transmission system operation signal, obtain the predicted load demand, and generate multiple candidate pulse width states according to the predicted load demand.

[0132] The candidate pulse width state analysis module 13 is used to perform operation response analysis and motor influence analysis on the multiple candidate pulse width states, and obtain multiple operation response degrees and multiple motor influence degrees.

[0133] The optimal pulse width state selection module 14 is used to obtain multiple pulse width fitness values ​​based on the multiple operation responsiveness values ​​and the multiple motor influence values, and select the optimal pulse width state from multiple candidate pulse width states based on the multiple pulse width fitness values, and control the target electric drive system based on the optimal pulse width state.

[0134] In one embodiment, the candidate pulse width state generation module 12 is specifically used for:

[0135] Obtain a preset feature template, and extract features from the motor operating parameters based on the preset feature template to obtain the current motor operating features;

[0136] Retrieve the signal encoding table, input the operation signal of the electric drive system into the signal encoding table, and obtain the current operation signal encoding;

[0137] The current motor operating characteristics are used as the first search condition, and the current operation signal encoding is used as the second search condition to jointly retrieve multiple motor load response records;

[0138] Each of the motor load response records includes historical motor operating characteristics, historical operation signal codes, and corresponding historical load requirements. The matching degree between the historical motor operating characteristics and the current motor operating characteristics is greater than a preset feature matching threshold, and the historical operation signal codes are consistent with the current operation signal codes.

[0139] Extract multiple historical load demands from multiple motor load response records, perform aggregated value statistics on the multiple historical load demands, and obtain the predicted load demand.

[0140] Furthermore, in one embodiment of the application, multiple candidate pulse width states are generated based on the predicted load demand, including:

[0141] Activate the pulse width integration configurator, which includes multiple pulse width configuration units;

[0142] The predicted load demand is input into the pulse width integration configurator, which calls multiple pulse width configuration units to configure the pulse width for the predicted load demand, thereby obtaining the multiple candidate pulse width states.

[0143] The construction steps of the pulse width integrated configurator include:

[0144] Collect multiple sample load data, construct a sample load dataset, and annotate the pulse width of each sample load data in the sample load dataset to obtain a sample pulse width annotation set;

[0145] Build a multi-configuration unit architecture;

[0146] Using the sample load dataset as input and the sample pulse width annotation set as the supervision target, the multiple configuration unit architectures are trained synchronously to obtain multiple pulse width configuration units.

[0147] The multiple pulse width configuration units are integrated to obtain the pulse width integrated configurator.

[0148] In one embodiment, the candidate pulse width state analysis module 13 is specifically used for:

[0149] Select the first candidate pulse width state from the plurality of candidate pulse width states;

[0150] The first candidate pulse width state and the predicted load demand are input into the operation response evaluator to obtain the first operation response degree;

[0151] The current pulse width is extracted from the motor operating parameters, and the first motor influence degree is obtained based on the current pulse width and the state of the first candidate pulse width.

[0152] Following the method of obtaining the first operational response degree and the first motor influence degree of the first candidate pulse width state, the operational response degree and motor influence degree of the remaining candidate pulse width states are obtained, resulting in multiple operational response degrees and multiple motor influence degrees.

[0153] Furthermore, in one embodiment of the application, the construction step of the operational response evaluator includes:

[0154] Based on historical control data of motors of the same model, sample pulse width parameter sets and sample load demand sets were collected, and the operation response effects obtained under different sample pulse width parameters and sample load demands were collected. The sample operation response degree set was obtained by labeling the operation response effects.

[0155] Construct an operation response evaluator architecture, wherein the input features of the operation response evaluator architecture are pulse width parameters and load requirements, and the output feature is operation response degree;

[0156] The operational response estimator architecture is trained under supervision using the sample pulse width parameter set, sample load requirement set, and sample operational response degree set until convergence, thus obtaining the trained operational response estimator.

[0157] The first motor influence degree is obtained based on the current pulse width and the state of the first candidate pulse width, including:

[0158] Calculate the deviation between the current pulse width and the first candidate pulse width state to obtain the first pulse width deviation;

[0159] Retrieve a preset pulse width influence table, which records the mapping relationship between different pulse width deviation ranges and the corresponding motor influence.

[0160] Based on the first pulse width deviation, the matching motor influence degree is retrieved from the pulse width influence degree correspondence table to obtain the first motor influence degree.

[0161] In one embodiment, the optimal pulse width state selection module 14 is specifically used for:

[0162] Obtain preset weighting coefficients, which include operation response weights and motor influence weights;

[0163] Based on the operation response weight, the motor influence weight, the multiple operation response degrees, and the multiple motor influence degrees, an adaptive evaluation is performed on multiple candidate pulse width states to obtain multiple pulse width fitness values.

[0164] Select the candidate pulse width state corresponding to the largest pulse width fitness among multiple pulse width fitness values ​​as the optimal pulse width state.

[0165] Further, in one embodiment, based on the operation response weight, the motor influence weight, the plurality of operation response degrees, and the plurality of motor influence degrees, an adaptive evaluation is performed on the plurality of candidate pulse width states to obtain a plurality of pulse width fitness values, including:

[0166] Select a target candidate pulse width state from the plurality of candidate pulse width states, and obtain the corresponding target operation response and target motor influence from the plurality of operation response and the plurality of motor influence;

[0167] Based on the operation response weight, the motor influence weight, the target operation response degree, and the target motor influence degree, the target pulse width fitness degree of the target candidate pulse width state is obtained;

[0168] Following the method of obtaining the target pulse width fitness of the target candidate pulse width state, the pulse width fitness of the remaining candidate pulse width states is obtained, resulting in multiple pulse width fitness values.

[0169] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0170] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0171] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A pulse width modulation dynamic optimization method combining artificial intelligence, characterized in that, The method includes: Responding to the operation signals of the electric drive system, the operating parameters of the target electric drive system's motor are obtained; Load prediction is performed based on the motor operating parameters and the power drive system operation signals to obtain the predicted load demand. Multiple candidate pulse width states are then generated based on the predicted load demand, including: Obtain a preset feature template, and extract features from the motor operating parameters based on the preset feature template to obtain the current motor operating features; Retrieve the signal encoding table, input the operation signal of the electric drive system into the signal encoding table, and obtain the current operation signal encoding; The current motor operating characteristics are used as the first search condition, and the current operation signal encoding is used as the second search condition to jointly retrieve multiple motor load response records; Each of the motor load response records includes historical motor operating characteristics, historical operation signal codes, and corresponding historical load requirements. The matching degree between the historical motor operating characteristics and the current motor operating characteristics is greater than a preset feature matching threshold, and the historical operation signal codes are consistent with the current operation signal codes. Extract multiple historical load demands from multiple motor load response records, perform mass value statistics on the multiple historical load demands, and obtain the predicted load demand. Operational response analysis and motor influence analysis are performed on the multiple candidate pulse width states to obtain multiple operational response degrees and multiple motor influence degrees, including: Select the first candidate pulse width state from the plurality of candidate pulse width states; The first candidate pulse width state and the predicted load demand are input into the operation response evaluator to obtain the first operation response degree; The current pulse width is extracted from the motor operating parameters, and the first motor influence degree is obtained based on the current pulse width and the state of the first candidate pulse width. Following the method of obtaining the first operational response and the first motor influence of the first candidate pulse width state, the operational response and motor influence of the remaining candidate pulse width states are obtained, resulting in multiple operational response and multiple motor influence. The construction steps of the operation response evaluator include: Based on historical control data of motors of the same model, sample pulse width parameter sets and sample load demand sets were collected, and the operation response effects obtained under different sample pulse width parameters and sample load demands were collected. The sample operation response degree set was obtained by labeling the operation response effects. Construct an operation response evaluator architecture, wherein the input features of the operation response evaluator architecture are pulse width parameters and load requirements, and the output feature is operation response degree; The operational response estimator architecture is supervised and trained using the sample pulse width parameter set, sample load requirement set, and sample operational response degree set until convergence, thus obtaining the trained operational response estimator. Multiple pulse width fitness values ​​are obtained based on the multiple operational responsiveness values ​​and the multiple motor influence values. An optimal pulse width state is selected from multiple candidate pulse width states based on these multiple pulse width fitness values. The target electric drive system is then controlled based on the optimal pulse width state, including: Obtain preset weighting coefficients, which include operation response weights and motor influence weights; Based on the operation response weight, the motor influence weight, the multiple operation response degrees, and the multiple motor influence degrees, an adaptive evaluation is performed on multiple candidate pulse width states to obtain multiple pulse width fitness values. Select the candidate pulse width state corresponding to the largest pulse width fitness among multiple pulse width fitness values ​​as the optimal pulse width state.

2. The method according to claim 1, characterized in that, Multiple candidate pulse width states are generated based on the predicted load demand, including: Activate the pulse width integration configurator, which includes multiple pulse width configuration units; The predicted load demand is input into the pulse width integration configurator, which calls multiple pulse width configuration units to configure the pulse width for the predicted load demand, thereby obtaining the multiple candidate pulse width states.

3. The method according to claim 2, characterized in that, The steps for constructing the pulse width integrated configurator include: Collect multiple sample load data, construct a sample load dataset, and annotate the pulse width of each sample load data in the sample load dataset to obtain a sample pulse width annotation set; Build a multi-configuration unit architecture; Using the sample load dataset as input and the sample pulse width annotation set as the supervision target, the multiple configuration unit architectures are trained synchronously to obtain multiple pulse width configuration units. The multiple pulse width configuration units are integrated to obtain the pulse width integrated configurator.

4. The method according to claim 1, characterized in that, The first motor influence degree is obtained based on the current pulse width and the state of the first candidate pulse width, including: Calculate the deviation between the current pulse width and the first candidate pulse width state to obtain the first pulse width deviation; Retrieve a preset pulse width influence table, which records the mapping relationship between different pulse width deviation ranges and the corresponding motor influence. Based on the first pulse width deviation, the matching motor influence degree is retrieved from the pulse width influence degree correspondence table to obtain the first motor influence degree.

5. The method according to claim 1, characterized in that, Based on the operation response weight, the motor influence weight, the multiple operation response degrees, and the multiple motor influence degrees, an adaptive evaluation is performed on multiple candidate pulse width states to obtain multiple pulse width fitness values, including: Select a target candidate pulse width state from the plurality of candidate pulse width states, and obtain the corresponding target operation response and target motor influence from the plurality of operation response and the plurality of motor influence; Based on the operation response weight, the motor influence weight, the target operation response degree, and the target motor influence degree, the target pulse width fitness degree of the target candidate pulse width state is obtained; Following the method of obtaining the target pulse width fitness of the target candidate pulse width state, the pulse width fitness of the remaining candidate pulse width states is obtained, resulting in multiple pulse width fitness values.

6. A pulse width modulation dynamic optimization system incorporating artificial intelligence, characterized in that, For performing the method according to any one of claims 1-5, comprising: The operating parameter acquisition module is used to respond to the operating signals of the electric drive system and acquire the motor operating parameters of the target electric drive system. The candidate pulse width state generation module is used to perform load prediction based on the motor operating parameters and the power transmission system operation signal to obtain the predicted load demand, and generate multiple candidate pulse width states according to the predicted load demand. The candidate pulse width state analysis module is used to perform operation response analysis and motor influence analysis on the multiple candidate pulse width states, and obtain multiple operation response degrees and multiple motor influence degrees. The optimal pulse width state selection module is used to obtain multiple pulse width fitness values ​​based on the multiple operation responsiveness values ​​and the multiple motor influence values, and select the optimal pulse width state from multiple candidate pulse width states based on the multiple pulse width fitness values, and control the target electric drive system based on the optimal pulse width state.

Images