Model optimization method, device and system

Abstract

The invention discloses a model optimization method, device and system, the method, device and system are applied to a server, the server comprises a controlled object model and a controller model, and the method comprises the following steps: receiving operation data sent by mechanical equipment; data analysis is conducted on the operation data, a data analysis result is obtained, and the data analysis result comprises the actual working condition of the mechanical equipment and / or a predicted value output by the controlled object model; and under the condition that the data analysis result does not meet the preset condition, updating the controller model to obtain an accurate control strategy. According to the technical scheme, through prediction and optimization of the model, dependence on a physical model machine is reduced, and efficiency and quality of control strategy optimization are improved.

Description

Technical Field

[0001] The present application relates to the technical field of engineering machinery optimization control algorithm simulation, and in particular to a model optimization method, device and system. Background Art

[0002] In the field of engineering machinery, due to the complex and changeable working environment of mechanical equipment and the aging of equipment parts, traditional control strategies often rely on repeated testing of physical prototypes and empirical data, which is not only time-consuming and labor-intensive, but also difficult to ensure that the mechanical equipment is always in the best working condition. Therefore, how to generate a control strategy that can adapt to complex working conditions and take into account factors such as equipment aging is a technical problem that needs to be solved urgently. Summary of the invention

[0003] In view of this, the embodiments of the present application provide a method, device and system for model optimization, which can reduce the dependence on physical prototypes and improve the efficiency and quality of strategy optimization.

[0004] In a first aspect, an embodiment of the present application provides a model optimization method applied to a server, wherein the server includes a controlled object model and a controller model, wherein the method includes: receiving operating data sent by mechanical equipment; performing data analysis on the operating data to obtain data analysis results, wherein the data analysis results include the actual working conditions of the mechanical equipment, and / or the predicted values ​​output by the controlled object model; when the data analysis results do not meet preset conditions, updating the controller model to obtain an accurate control strategy.

[0005] In a second aspect, an embodiment of the present application provides a model optimization system, comprising: a mechanical device and a server, wherein the mechanical device and the server are communicatively connected, and the server is used to execute the model optimization method described in the first aspect above.

[0006] In the third aspect, an embodiment of the present application provides a model optimization device, which is applied to a server, including: a receiving module, used to receive operating data sent by mechanical equipment; a data analysis module, used to perform data analysis on the operating data to obtain data analysis results, wherein the data analysis results include the actual working conditions of the mechanical equipment, and / or the predicted values ​​output by the controlled object model; an updating module, used to update the controller model to obtain an accurate control strategy when the data analysis results do not meet preset conditions.

[0007] In a fourth aspect, an embodiment of the present application provides a computer program product, including a computer program, which implements the model optimization method described in the first aspect when the computer program is executed by a processor.

[0008] In a fifth aspect, an embodiment of the present application provides an electronic device, comprising: a processor; and a memory for storing processor executable instructions, wherein the processor is used to execute the model optimization method described in the first aspect above.

[0009] The embodiment of the present application provides a method, device and system for model optimization, which receives the operation data sent by the mechanical equipment and performs data analysis on the operation data to obtain the data analysis results, and then updates the controller model when the data analysis results do not meet the preset conditions to obtain an accurate control strategy, so that the embodiment of the present application can optimize the control strategy by updating the model in the server, reduce the dependence on the physical prototype, and improve the efficiency and quality of the control strategy optimization. In addition, the operation data of the mechanical equipment is responded to in real time to ensure the adaptability of the control strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation of the present disclosure. The above and other features and advantages will become more apparent to those skilled in the art by describing detailed example embodiments with reference to the accompanying drawings. In the accompanying drawings:

[0011] Figure 1 It is a schematic diagram of a system for model optimization provided by an exemplary embodiment of the present application.

[0012] Figure 2 It is a flowchart of a model optimization method provided by an exemplary embodiment of the present application.

[0013] Figure 3 It is a flowchart of a method for model optimization provided by another exemplary embodiment of the present application.

[0014] Figure 4 It is a flowchart of a method for model optimization provided by yet another exemplary embodiment of the present application.

[0015] Figure 5 It is a flowchart of a method for model optimization provided by yet another exemplary embodiment of the present application.

[0016] Figure 6 It is a flowchart of a method for model optimization provided by yet another exemplary embodiment of the present application.

[0017] Figure 7 It is a structural schematic diagram of a model optimization device provided by an exemplary embodiment of the present application.

[0018] Figure 8 It is a block diagram of an electronic device for model optimization provided by an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0020] In the field of engineering machinery, control strategies need to have a certain degree of adaptability due to the complex and changeable working environment of mechanical equipment and the aging of parts. However, existing control strategies usually rely on repeated testing of physical prototypes and empirical data. This method is not only time-consuming and labor-intensive, but also difficult to ensure that the mechanical equipment is always in the best condition and it is difficult to find the optimal control strategy. In addition, current control strategies usually focus on standard working conditions in the R&D and design stages, and cannot achieve optimal control in complex and changeable customer operating scenarios.

[0021] In view of the above problems, an embodiment of the present application provides a method for model optimization. Various non-limiting embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0022] Figure 1 1 is a schematic diagram of a model optimization system provided by an exemplary embodiment of the present application. The model optimization system includes a server 101 and a mechanical device 102, wherein the server 101 and the mechanical device 102 are in communication connection, that is, data can be exchanged between the server 101 and the mechanical device 102. In other words, the model optimization system can be understood as mainly including three modules: a physical space representing the mechanical device 102, data exchange, and a virtual space running in the server 101.

[0023] The mechanical equipment 102 may be an engineering machine (i.e., the physical space may refer to an engineering machine entity), for example, an excavator or a material grabber. The physical space may be divided into three levels, namely, the whole machine, (sub)systems, and parts. The various levels and systems cooperate with each other to form different types of engineering machines.

[0024] The data interaction module can establish a communication connection between the server 101 and the mechanical device 102 through the transmission base station. In other words, the data interaction module connects the physical space module (i.e., the mechanical device 102) and the virtual space module (i.e., the server 101), which may include data communication, data storage, and data preprocessing. Through the two-way interaction of data, the real-time operation data of the mechanical device 102 can be transmitted to the server 101 through the transmission base station for analysis, preview, and comparison, and the analysis structure, optimization conclusions, and control strategy of the server 101 can also be fed back to the mechanical device 102 in a timely manner.

[0025] The virtual space running in the server 101 may refer to various virtual models (e.g., controlled object model and controller model) and algorithms deployed in the server 101. The controlled object model and controller model can be used as accurate mappings of physical entities, which are continuously iterated under the drive of the real-time operation data and optimization algorithm of the mechanical device 102, maintaining synchronization with the mechanical device 102, and optimizing the control strategy through the intelligent optimization algorithm and sending it to the mechanical device 102.

[0026] In addition, the server 101 may also pre-store offline data such as experimental data and design data to facilitate subsequent updating of various virtual models running in the server 101 .

[0027] Figure 2 It is a flowchart of a model optimization method provided by an exemplary embodiment of the present application. Figure 2 The method is performed by a computing device, such as a server (ie, Figure 1 The model optimization method includes the following contents.

[0028] In one embodiment, the server includes a digital twin model, which includes a controlled object model (also called a physical model) and a controller model.

[0029] The controlled object model can be a virtual representation of a physical entity (e.g., mechanical equipment). It obtains the state data of the physical entity object in real time through sensors and other data sources, and then performs modeling and simulation in a digital environment.

[0030] A controller model can refer to a digital model used to control or manage a controlled object. It is usually used to simulate the control of physical entities in a digital twin environment. The controller model can simulate the control system or operator's thinking in reality, and optimize the control strategy, improve the system response and improve the overall performance through interaction with the controlled object model.

[0031] S210: receiving operation data sent by mechanical equipment.

[0032] Specifically, the server may receive the operation data sent by the mechanical equipment in real time or non-real time. The mechanical equipment may be an engineering machine, such as an excavator or a vehicle. The operation data may include data such as the working condition data and status information of the mechanical equipment, wherein the status information may be data collected by sensors on the mechanical equipment, for example, the speed data of the vehicle detected by the wheel speed sensor of the vehicle. The working condition data may be, for example, the China Heavy-duty Commercial Vehicle Test Cycle (CHTC).

[0033] S220: Analyze the operation data to obtain data analysis results.

[0034] In one embodiment, the data analysis result includes the actual working condition of the mechanical equipment and / or the predicted value of the controlled object model output.

[0035] Specifically, data analysis can analyze different types of data in the operating data at the same time, or can perform weighted analysis on different types of data, etc., which is not specifically limited in the embodiments of the present application. In addition, data analysis can be performed by statistical methods or by a controlled object model.

[0036] In one embodiment, data analysis is performed on the working condition data in the operation data to obtain the actual working condition of the mechanical equipment. And / or, the operation data is input into the controlled object model to obtain the predicted value output by the controlled object model.

[0037] S230: When the data analysis result does not meet the preset conditions, the controller model is updated to obtain an accurate control strategy.

[0038] Specifically, when the data analysis result is inconsistent with the preset target analysis result (e.g., target operating condition), the controller model is updated. And / or, when the data analysis result is greater than a preset target threshold (e.g., target prediction value), the controller model is updated. It should be noted that the preset conditions of the embodiment of the present application can be flexibly set according to actual conditions. Moreover, the preset conditions can also be understood as trigger conditions for updating the controller model, wherein the controller model can output a control strategy for controlling the mechanical equipment.

[0039] In one embodiment, data analysis is performed on the operating condition data in the operation data to obtain the actual operating condition of the mechanical equipment, and when the actual operating condition is inconsistent with the preset target operating condition, the controller model is updated.

[0040] In one embodiment, operating data is input into the controlled object model to obtain a predicted value output by the controlled object model. When the difference between the predicted value and the target predicted value exceeds a preset deviation threshold, the parameters of the controlled object model are updated to obtain updated parameters of the controlled object model, and then the controller model is updated based on the updated parameters.

[0041] It should be noted that the above two methods of triggering controller model update, i.e., control strategy optimization methods, can exist at the same time. For example, see Figure 3 ,The two triggering methods ensure that the control strategy of mechanical equipment can flexibly adapt to complex working conditions and take into account the influence of factors such as equipment aging.

[0042] It should also be noted that the interaction between the controlled object model and the controller model is the core of the digital twin. Through continuous feedback and adjustment, real-time monitoring and optimization of the physical world can be achieved.

[0043] It can be seen that the embodiment of the present application receives the operation data sent by the mechanical equipment and performs data analysis on the operation data to obtain the data analysis results, and then updates the controller model when the data analysis results do not meet the preset conditions to obtain an accurate control strategy, so that the embodiment of the present application can optimize the control strategy by updating the model in the server, reducing the dependence on the physical prototype and improving the efficiency and quality of the control strategy optimization. In addition, the operation data of the mechanical equipment is responded to in real time to ensure the adaptability of the control strategy.

[0044] In one embodiment of the present application, the operating data includes operating condition data, wherein data analysis is performed on the operating data to obtain data analysis results, including: data analysis is performed on the operating condition data to obtain the actual operating condition of the mechanical equipment, and when the data analysis result does not meet the preset conditions, the controller model is updated, including: when the actual operating condition is inconsistent with the preset target operating condition, the controller model is updated.

[0045] Specifically, the correspondence between the operating condition data and the target operating condition can be pre-stored in the server, that is, there is a one-to-one correspondence between the operating condition data and the target operating condition, wherein the target operating condition can be an operating condition designed based on prior knowledge or historical data.

[0046] In one embodiment, see Figure 3, the server 101 can analyze the working condition data in the operation data to obtain the actual working condition of the mechanical equipment 102 during the operation phase (i.e., step S310). The actual working condition obtained by analysis is judged to be different from the target working condition pre-stored in the server 101 (i.e., step S320), and when the actual working condition is inconsistent with the preset target working condition and the deviation is large, the controller model is optimized (i.e., step S330) to obtain the updated parameters of the controller model. Then, based on the updated parameters of the controller model, the controller model is updated (i.e., step S340).

[0047] It should be noted that in the embodiment of the present application, the controlled object model in the server is always driven by the actual working condition data. When the actual working condition of the mechanical equipment is significantly different from the target working condition preset in the R&D design stage, it will trigger the update of the controller model, that is, trigger the optimization of the control strategy.

[0048] It can be seen from this that the embodiment of the present application analyzes the actual working conditions and updates the controller model when there is a large deviation between the actual working conditions and the target working conditions, so that the controller model is always updated during the operation phase, ensuring that the controller model can quickly follow the changes in the mechanical equipment, and update and adjust parameters in real time, thereby improving the accuracy of the control strategy.

[0049] In one embodiment of the present application, a predictive evaluation is performed on the operating data to obtain a predictive evaluation result, including: inputting the operating data into a controlled object model to obtain a predicted value output by the controlled object model, wherein, when the data analysis result does not meet a preset condition, the controller model is updated, including: when the difference between the predicted value and the target predicted value exceeds a preset deviation threshold, updating the parameters of the controlled object model to obtain updated parameters of the controlled object model; and updating the controller model based on the updated parameters.

[0050] Specifically, the target prediction value can be obtained by actual measurement of the sensor of the mechanical equipment. For example, after the mechanical equipment obtains the target prediction value through its own sensor, the target prediction value is transmitted to the server. The preset deviation threshold can be flexibly set according to actual conditions, and the embodiments of the present application do not specifically limit this.

[0051] In one embodiment, see Figure 3, input the operation data into the controlled object model in the server 101, and obtain the predicted value output by the controlled object model (i.e., step S350). Then compare the predicted value with the target predicted value (i.e., step S380) measured by the mechanical equipment 102 through the sensor (i.e., step S360), wherein when the difference (which can be understood as the deviation) between the predicted value and the target predicted value exceeds the preset deviation threshold, update the model parameters of the controlled object model (i.e., step S370); optimize the parameters of the controller model in combination with the updated model parameters of the controlled object model (i.e., step S330), and obtain the updated parameters of the controller model. Then, based on the updated parameters of the controller model, update the controller model (i.e., step S340).

[0052] It should be understood that the situation where the difference exceeds the preset deviation threshold can be understood as a large deviation between the predicted value and the target predicted value.

[0053] It should be noted that in the embodiment of the present application, the controlled object model in the server is always driven by the actual working condition data. When the predicted value of the controlled object model deviates greatly from the actual collected target predicted value, the parameter update of the controlled object model is triggered, and the optimization of the controller model is triggered immediately after the update is completed.

[0054] It can be seen from this that the embodiment of the present application compares the predicted value of the model with the target predicted value actually measured by the sensor, and updates the controller model when there is a large deviation between the predicted value of the model and the target predicted value, so that the controller model is always updated during the operation phase, thereby ensuring that the control strategy of the mechanical equipment can flexibly adapt to complex working conditions and take into account the influence of factors such as equipment aging.

[0055] In one embodiment of the present application, a predictive evaluation is performed on the operating data to obtain a predictive evaluation result, including: performing data analysis on the operating condition data in the operating data to obtain the actual operating condition of the mechanical equipment, and inputting the operating data into a controlled object model to obtain a predicted value output by the controlled object model, wherein, when the data analysis result does not meet the preset conditions, the controller model is updated, including: when the actual operating condition is inconsistent with the preset target operating condition, and the difference between the predicted value and the target predicted value exceeds a preset deviation threshold, updating the parameters of the controlled object model to obtain updated parameters of the controlled object model; and updating the controller model based on the updated parameters.

[0056] It should be noted that, for example Figure 3 For details of the specific description of this step, please refer to the records of the above embodiment.

[0057] In one embodiment of the present application, parameters of a controlled object model are updated to obtain updated parameters of the controlled object model, including: when the controlled object model is a linear model, based on the least squares method, the linear function corresponding to the linear model and the operating data, obtaining the updated parameters of the controlled object model.

[0058] It should be noted that the controlled object model can be a linear model or a nonlinear model. Therefore, when updating the controlled object model, it is necessary to judge the model type of the controlled object model and update it according to the characteristics of the controlled object model.

[0059] Specifically, see Figure 4 Step S410: The controlled object model starts updating. Step S415: Select the latest m groups of data in the operation data. The m groups of data can be expressed as D = {(x 1 ,y 1 ),(x 2 ,y 2 ),……,(x m ,y m )}. Step S420: Determine whether the controlled object model is a linear model. If the controlled object model is a linear model, execute step S425: determine the parameters to be updated of the linear model based on the linear function of the linear model and m groups of data. The linear function can be expressed by the following formula (1):

[0060] y=θx (1)

[0061] Wherein, θ is the parameter to be updated; y and x are m sets of data in the running data. It should be noted that θ can be obtained by substituting m sets of data into formula (1).

[0062] Further, step S430: solve the parameters to be updated of the model by the least square method of the linear model. In addition to the least square method, the maximum likelihood estimation method, the gradient descent method, the random search method, the genetic algorithm and other algorithms can also be used to calculate the estimated values ​​of the parameters to be updated. Step S435: The estimated values ​​of the parameters to be updated solved by the least square method are used as the updated parameters of the controlled object model.

[0063] It can be seen from this that the embodiment of the present application updates the controlled object model as a linear model, realizes the preset optimal parameter update method for different models, reduces the complexity of building the model optimization system, and improves the uniformity.

[0064] In one embodiment of the present application, parameters of a controlled object model are updated to obtain updated parameters of the controlled object model, including: when the controlled object model is a nonlinear model, obtaining multiple parameters to be updated of the nonlinear model; calculating an error value between a predicted value and an expected output value of running data output by each parameter to be updated in the multiple parameters to be updated of the controlled object model; and when the error value is less than a preset error threshold, obtaining updated parameters of the controlled object model.

[0065] Specifically, see Figure 4 , step S440: when the controlled object model is a nonlinear model, the parameters to be updated are determined based on the nonlinear function of the nonlinear model, where the nonlinear function can be expressed by the following formula (2):

[0066] y=f(x / θ) (2)

[0067] Among them, f is a nonlinear function; θ is the parameter to be updated of the controlled object model; y and x are m groups of data in the operating data.

[0068] Further, step S445: the parameters to be updated currently applied by the controlled object model and the m parameters to be updated calculated based on the m groups of data are used as multiple parameters to be updated and as initial values ​​at the same time, for example, Latin hypercube sampling (LHS). Step 450: Calculate the error value of the controlled object model on the m groups of data (i.e., part of the data in the operating data) under each parameter value to be updated (i.e., the numerical value corresponding to the parameter to be updated). Determine whether the error is less than a preset error threshold (i.e., step S455), wherein if the error value is less than the preset error threshold, execute step S435: based on the multiple parameters to be updated corresponding to the error value less than the error threshold, obtain the updated parameters of the controlled object model.

[0069] It can be seen that the embodiment of the present application updates the controlled object model as a nonlinear model, realizes the preset optimal parameter update method for different models, reduces the complexity of building the model optimization system, and improves the uniformity.

[0070] In one embodiment of the present application, it also includes: when the error value is greater than a preset error threshold, determining whether the number of updates of the parameters to be updated of the controlled object model is greater than a first update number threshold; when the number of updates is greater than the first update number threshold, using a preset plurality of initial parameters to be updated as a plurality of parameters to be updated, wherein the plurality of initial parameters to be updated are determined based on a genetic algorithm or a random search algorithm.

[0071] Specifically, see Figure 4 , in the error value (Q loss) is greater than a preset error threshold, executing step S460: determining whether the update times of the parameters to be updated of the controlled object model are greater than a first update times threshold, wherein when the update times of the parameters to be updated are greater than the first update times threshold, executing step S465: performing preliminary iterations of optimizing the parameters of the controlled object model using a genetic algorithm or a random search algorithm, and obtaining a plurality of initial parameters to be updated, that is, a set of initial parameter values ​​to be updated.

[0072] The specific value of the first update number threshold can be flexibly set according to actual conditions. It should be noted that the determination of the update number in the embodiment of the present application can be understood as determining whether the number of Bayesian iterations exceeds the first update number threshold.

[0073] In addition, when the update times of the parameter to be updated is less than the first update times threshold, step S470 is performed: training the proxy model. Specifically, the embodiment of the present application trains the proxy model so that the error value can be expressed as the following formula (3):

[0074] Q loss =s(θ) (3)

[0075] Among them, Q loss is the error value; s is the proxy model; θ is the parameter to be updated of the controlled object model. The proxy model can be Gaussian Process Regression (GPR), Tree-structured Parzen Estimator (TPE), neural network model, etc.

[0076] Further, step S475: based on the expected improvement function corresponding to the proxy model, determine the next sampled parameter to be updated.

[0077] It can be seen that in order to ensure the smooth update of the model parameters of the controlled object, the embodiment of the present application reduces the risk of non-convergence of the Bayesian optimization process. When the number of parameter iterations exceeds the predetermined number and no updated parameters that meet the conditions are found, a rough iteration of the parameters can be performed through a basic genetic algorithm or a random search algorithm, and then the updated parameters after the iteration are used as the initial values ​​for a new round of Bayesian optimization to ensure the smooth updating of the model.

[0078] In one embodiment of the present application, the controller model is updated, including: obtaining target operating condition data and parameters to be updated in the controller model, wherein the target operating condition data is part of the historical operating data; based on the target operating condition data and the parameters to be updated in the controller model, obtaining a performance value of the controlled object model; when the performance value is greater than a preset performance threshold, obtaining update parameters of the controller model; and updating the controller model based on the update parameters of the controller model.

[0079] Specifically, see Figure 5 , step S510: the parameters of the controller model (also called the controller model) start to be updated. Step S515: select target operating condition data from the historical operating data, such as vehicle speed-time curve data. Step S520: determine the parameters to be updated in the controller model. Step S525: use the parameters to be updated in the current controller model and several groups of parameters to be updated obtained from the operating data as initial values, such as Latin hypercube sampling.

[0080] Further, step 530: calculate the performance value of the controlled object model in the control strategy under the target operating condition data and each parameter to be updated. The performance value can be a single-objective performance such as power and economy, or a weighted multi-objective performance. Step S535: determine whether the performance value is greater than the target performance value. Wherein, if the performance value is greater than the target performance value, execute step S540: obtain the update parameters of the controller model.

[0081] It can be seen from this that the embodiment of the present application sets a corresponding parameter updating method for the controller model, so that the embodiment of the present application can automatically adjust the model parameters of the controller model according to factors such as changes in operating conditions, thereby ensuring the optimality of the control strategy.

[0082] In one embodiment of the present application, it also includes: when the performance value is greater than a preset performance threshold, determining whether the number of updates of the parameters to be updated of the controller model exceeds a second update number threshold; when the number of updates is greater than the second update number threshold, using the preset initial parameters to be updated as the parameters to be updated.

[0083] Specifically, see Figure 5 , when the performance value is less than the target performance value, further executing step S550: determining whether the update times of the parameters to be updated of the controller model exceed the second update times threshold. Wherein, when the update times of the parameters to be updated of the controller model exceed the second update times threshold, executing step S560: performing preliminary iterations of optimizing the controller model parameters using a genetic algorithm or a random search algorithm to obtain initial parameters to be updated, such as a set of initial parameter values ​​to be updated.

[0084] The specific value of the second update number threshold can be flexibly set according to actual conditions. It should be noted that the determination of the update number in the embodiment of the present application can be understood as determining whether the number of Bayesian iterations exceeds the second update number threshold.

[0085] In addition, if the update times of the to-be-updated parameters of the controller model do not exceed the second update times threshold, step S570 is performed: training the proxy model. Specifically, the embodiment of the present application trains the proxy model, which can be expressed as the following formula (4):

[0086] P=g(c) (4)

[0087] Where P is the performance value, g is the proxy model, and c is the parameter to be updated of the controller model. The proxy model can be Gaussian Process Regression (GPR), Tree-structured Parzen Estimator (TPE), neural network model, etc.

[0088] Further, step S580: based on the expected improvement function corresponding to the proxy model, determine the next sampled parameter to be updated.

[0089] It can be seen from this that in the embodiment of the present application, when the number of parameter iterations exceeds the predetermined number and still no updated parameters that meet the conditions are found, a rough iteration of the parameters can be performed using a basic genetic algorithm or a random search algorithm, and then the updated parameters after the iteration are used as the initial values ​​for a new round of Bayesian optimization to ensure the smooth update of the model parameters of the controlled object and reduce the risk of non-convergence of the Bayesian optimization process.

[0090] In one embodiment of the present application, it also includes: obtaining a sample data set, wherein the sample data set includes experimental sample data, initial control parameter sample data, and operating condition sample data; performing constraint training for parameter calculation of an initial controlled object model based on the experimental sample data to obtain a controlled object model; performing constraint training for parameter calculation of an initial controller model based on the initial control parameter sample data, the operating condition sample data, and the controlled object model to obtain a controller model.

[0091] Specifically, see Figure 6The controlled object model and the controller model are established in the R&D and design stage of the mechanical equipment. They can be established in the server by building a data layer 610, an algorithm layer 620, and a model layer 630. The parameters of the controlled object model are calibrated 621 through the experimental sample data 611 to obtain the calibrated experimental sample data. Then, based on the calibrated experimental sample data, the initial controlled object model is trained for parameter calculation constraints to obtain the desired controlled object model 631.

[0092] Furthermore, the controller model parameter optimization 622 is performed using the initial control parameter sample data 612 and the operating condition sample data 613 in combination with the trained controlled object model 631. Then, based on the controller model parameter optimization 622, the initial controller model is trained for parameter calculation constraints to obtain a controller model 632 that meets the performance requirements.

[0093] It should be noted that the method of establishing a digital twin model can be understood as the establishment of a controlled object model and a controller model covering the R&D design stage and the operation stage, as well as the interaction mechanism between the two.

[0094] It can be seen that the embodiment of the present application lays a foundation for subsequently obtaining a model that meets performance requirements by building a model in the research and development design stage.

[0095] Figure 7 Schematic diagram of the structure of a model optimization device provided by an exemplary embodiment of the present application. Figure 7 As shown, the model optimization device 700 includes: a receiving module 710, a data analysis module 720 and an updating module 730.

[0096] The receiving module 710 is used to receive the operating data sent by the mechanical equipment; the data analysis module 720 is used to perform data analysis on the operating data to obtain data analysis results, wherein the data analysis results include the actual working conditions of the mechanical equipment and / or the predicted values ​​output by the controlled object model; the updating module 730 is used to update the controller model to obtain an accurate control strategy when the data analysis results do not meet the preset conditions.

[0097] The embodiment of the present application provides a model optimization device, which receives the operation data sent by the mechanical equipment and performs data analysis on the operation data to obtain the data analysis results, and then updates the controller model when the data analysis results do not meet the preset conditions to obtain an accurate control strategy, so that the embodiment of the present application can optimize the control strategy by updating the model in the server, reducing the dependence on the physical prototype, and improving the efficiency and quality of the control strategy optimization. In addition, the operation data of the mechanical equipment is responded to in real time to ensure the adaptability of the control strategy.

[0098] According to one embodiment of the present application, the operating data includes operating condition data, and the data analysis module 720 is used to perform data analysis on the operating condition data to obtain the actual operating condition of the mechanical equipment; the update module 730 is used to update the controller model when the actual operating condition is inconsistent with the preset target operating condition.

[0099] According to one embodiment of the present application, the data analysis module 720 is used to input the operating data into the controlled object model to obtain the predicted value output by the controlled object model; the update module 730 is used to update the parameters of the controlled object model when the difference between the predicted value and the target predicted value exceeds a preset deviation threshold to obtain the updated parameters of the controlled object model; and the controller model is updated based on the updated parameters.

[0100] According to one embodiment of the present application, the data analysis module 720 is used to perform data analysis on the operating condition data in the operating data to obtain the actual operating condition of the mechanical equipment, and to input the operating data into the controlled object model to obtain the predicted value output by the controlled object model; the update module 730 is used to update the parameters of the controlled object model when the actual operating condition is inconsistent with the preset target operating condition and the difference between the predicted value and the target predicted value exceeds the preset deviation threshold to obtain the updated parameters of the controlled object model; and the controller model is updated based on the updated parameters.

[0101] According to an embodiment of the present application, the updating module 730 is used to obtain the updated parameters of the controlled object model based on the least squares method, the linear function corresponding to the linear model and the operating data when the controlled object model is a linear model.

[0102] According to one embodiment of the present application, the update module 730 is used to obtain multiple parameters to be updated of the nonlinear model when the controlled object model is a nonlinear model; calculate the error value between the predicted value and the expected output value of the running data output of each parameter to be updated in the running of the controlled object model; and obtain the updated parameters of the controlled object model when the error value is less than a preset error threshold.

[0103] According to one embodiment of the present application, the update module 730 is used to determine whether the number of updates of the parameters to be updated of the controlled object model is greater than a first update number threshold when the error value is greater than a preset error threshold; when the number of updates is greater than the first update number threshold, the preset multiple initial parameters to be updated are used as multiple parameters to be updated, wherein the multiple initial parameters to be updated are determined based on a genetic algorithm or a random search algorithm.

[0104] According to one embodiment of the present application, the update module 730 is used to obtain target operating condition data and parameters to be updated in the controller model, wherein the target operating condition data is part of the historical operating data; based on the target operating condition data and the parameters to be updated in the controller model, the performance value of the controlled object model is obtained; when the performance value is greater than a preset performance threshold, the update parameters of the controller model are obtained; and based on the update parameters of the controller model, the controller model is updated.

[0105] According to one embodiment of the present application, the update module 730 is used to determine whether the number of updates of the parameters to be updated of the controller model exceeds a second update number threshold when the performance value is greater than a preset performance threshold; when the number of updates is greater than the second update number threshold, the preset initial parameters to be updated are used as the parameters to be updated.

[0106] According to one embodiment of the present application, the data analysis module 720 is used to obtain a sample data set, wherein the sample data set includes experimental sample data, initial control parameter sample data, and operating condition sample data; based on the experimental sample data, constraint training of parameter calculation is performed on the initial controlled object model to obtain the controlled object model; based on the initial control parameter sample data, the operating condition sample data, and the controlled object model, constraint training of parameter calculation is performed on the initial controller model to obtain the controller model.

[0107] It should be understood that the specific working process and functions of the receiving module 710, the data analysis module 720 and the updating module 730 in the above embodiment can refer to the above Figures 2 to 6 In order to avoid repetition, the description of the model optimization method provided in the embodiment will not be repeated here.

[0108] Figure 8 It is a block diagram of an electronic device for model optimization provided by an exemplary embodiment of the present application.

[0109] Reference Figure 8 The electronic device 800 includes a processing component 810, which further includes one or more processors, and a memory resource represented by a memory 820 for storing instructions executable by the processing component 810, such as an application. The application stored in the memory 820 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 810 is configured to execute instructions to perform the above-mentioned model optimization method.

[0110] The electronic device 800 may also include a power supply component configured to perform power management of the electronic device 800, a wired or wireless network interface configured to connect the electronic device 800 to a network, and an input / output (I / O) interface. The electronic device 800 may be operated based on an operating system stored in the memory 820, such as Windows Server 2000. TM , Mac OSX TM , Unix TM , Linux TM , FreeBSD TM or similar.

[0111] An embodiment of the present application also provides a computer program product, including a computer program, wherein the computer program is used to execute the steps of the model optimization method described in the above method embodiment. For details, please refer to the above method embodiment, which will not be repeated here.

[0112] The computer program product may be implemented in hardware, software or a combination thereof. In one optional embodiment, the computer program product is embodied as a computer storage medium, and in another optional embodiment, the computer program product is embodied as a software product, such as a software development kit (SDK) and the like.

[0113] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, and will not be described one by one here.

[0114] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0115] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0116] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0117] Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, which may be electrical, mechanical or other forms.

[0118] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0119] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0120] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program check codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0121] It should be noted that, in the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" is two or more.

[0122] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for model optimization, characterized in that: applied to a server, the server comprising a controlled object model and a controller model, Wherein, the method comprises: Receive operation data sent by mechanical equipment; Performing data analysis on the operating data to obtain data analysis results, wherein the data analysis results include actual operating conditions of the mechanical equipment and / or predicted values ​​output by the controlled object model; When the data analysis result does not meet the preset conditions, the controller model is updated to obtain an accurate control strategy.

2. The method according to claim 1, characterized in that: The operation data includes operating condition data, wherein the performing of data analysis on the operation data to obtain a data analysis result includes: Perform data analysis on the working condition data to obtain the actual working condition of the mechanical equipment. When the data analysis result does not meet the preset condition, updating the controller model includes: When the actual operating condition is inconsistent with the preset target operating condition, the controller model is updated.

3. The method according to claim 1, characterized in that The predictive evaluation of the operation data to obtain the predictive evaluation result includes: Inputting the operating data into the controlled object model to obtain a predicted value output by the controlled object model, Wherein, when the data analysis result does not meet the preset condition, updating the controller model includes: When the difference between the predicted value and the target predicted value exceeds a preset deviation threshold, updating the parameters of the controlled object model to obtain updated parameters of the controlled object model; The controller model is updated based on the update parameters.

4. The method according to claim 1, characterized in that The predictive evaluation of the operation data to obtain the predictive evaluation result includes: Performing data analysis on the working condition data in the operating data to obtain the actual working condition of the mechanical equipment, and inputting the operating data into the controlled object model to obtain a predicted value output by the controlled object model, Wherein, when the data analysis result does not meet the preset condition, updating the controller model includes: When the actual working condition is inconsistent with the preset target working condition, and the difference between the predicted value and the target predicted value exceeds a preset deviation threshold, updating the parameters of the controlled object model to obtain updated parameters of the controlled object model; The controller model is updated based on the update parameters.

5. The method according to claim 3, characterized in that: The step of updating the parameters of the controlled object model to obtain updated parameters of the controlled object model includes: In the case where the controlled object model is a linear model, an update parameter of the controlled object model is obtained based on a least square method, a linear function corresponding to the linear model and the operating data.

6. The method according to claim 3, characterized in that: The step of updating the parameters of the controlled object model to obtain updated parameters of the controlled object model includes: When the controlled object model is a nonlinear model, obtaining a plurality of parameters to be updated of the nonlinear model; Calculating an error value between a predicted value and an expected output value of the operation data outputted by each of the plurality of parameters to be updated when the controlled object model is running; When the error value is less than a preset error threshold, an updated parameter of the controlled object model is obtained.

7. The method according to claim 6, characterized in that Also includes: In the case where the error value is greater than a preset error threshold, determining whether the update times of the to-be-updated parameters of the controlled object model are greater than a first update times threshold; In a case where the number of updates is greater than the first update number threshold, a plurality of preset initial parameters to be updated are used as the plurality of parameters to be updated, wherein the plurality of initial parameters to be updated are determined based on a genetic algorithm or a random search algorithm.

8. The method according to claim 1, characterized in that The updating of the controller model comprises: Acquire target operating condition data and parameters to be updated in the controller model, wherein the target operating condition data is part of the historical operating data; Based on the target operating condition data and the parameters to be updated in the controller model, obtaining a performance value of the controlled object model; When the performance value is greater than a preset performance threshold, obtaining update parameters of the controller model; The controller model is updated based on the update parameters of the controller model.

9. The method according to claim 8, characterized in that Also includes: In the case where the performance value is greater than a preset performance threshold, determining whether the update times of the to-be-updated parameters of the controller model exceed a second update times threshold; When the update number is greater than the second update number threshold, the preset initial parameter to be updated is used as the parameter to be updated.

10. The method according to any one of claims 1 to 9, characterized in that Also includes: Acquire a sample data set, wherein the sample data set includes experimental sample data, initial control parameter sample data, and operating condition sample data; Performing constraint training of parameter calculation on the initial controlled object model based on the experimental sample data to obtain the controlled object model; Based on the initial control parameter sample data, the operating condition sample data and the controlled object model, constraint training of parameter calculation is performed on the initial controller model to obtain the controller model.

11. A system for model optimization, characterized in that: include: A mechanical device and a server, wherein the mechanical device and the server are communicatively connected, and the server is used to execute the method described in any one of claims 1 to 10 above.

12. A device for model optimization, characterized in that: Applicable to servers, including: A receiving module, used for receiving operation data sent by mechanical equipment; A data analysis module, configured to perform data analysis on the operating data to obtain data analysis results, wherein the data analysis results include the actual working conditions of the mechanical equipment and / or the predicted values ​​output by the controlled object model; The updating module is used to update the controller model to obtain an accurate control strategy when the data analysis result does not meet the preset conditions.

13. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

14. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor, The processor is used to execute the method described in any one of claims 1 to 10.

Images