Precision-improved equipment model adaptive training method and system, storage medium and electronic equipment

Abstract

The invention relates to a precision-improved equipment model adaptive training method and system, a storage medium and electronic equipment, and the method comprises the steps: obtaining factory data and real-time operation data of equipment, carrying out the preprocessing of the factory data and the real-time operation data, and obtaining a dynamic training data set; carrying out equipment model training based on the dynamic training data set to obtain a trained equipment model; verifying the trained equipment model to obtain an equipment model passing the verification; judging whether the precision of the equipment model passing the verification meets the precision requirement or not; if the precision requirement is not met, retraining the equipment model; if the precision requirement is met, deploying the equipment model for online use; continuously monitoring the precision; and if the precision does not meet the precision requirement, updating the equipment model on line. The method can effectively cope with the influence of the dynamic change of data on the performance of the model, improves the adaptability and robustness of the model, also can improve the prediction precision and reliability of the model, and reduces the operation cost.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing and machine learning, and more specifically, to a method, system, storage medium and electronic device for adaptive training of a device model with improved accuracy. Background Art

[0002] In the current field of AI applications and scientific research, the accuracy optimization of the underlying device model has always been a hot topic and difficulty in research. With the rapid development of technologies such as big data and artificial intelligence, higher requirements are placed on the prediction and computing capabilities of device models. However, existing device model training methods often seem to be unable to cope with complex and changing application scenarios, which is specifically manifested in the following aspects:

[0003] Limitations of static training strategies: Most traditional equipment model training methods adopt static strategies, that is, one-time or limited number of iterative training based on fixed data sets and preset algorithm parameters. This method ignores the impact of dynamic changes in data on model performance, resulting in the model's prediction or calculation accuracy being difficult to guarantee when facing new data or changes in data distribution. Accuracy improvement bottleneck: Due to the limitations of static training strategies, the model may quickly converge to a certain level of accuracy in the early stages of training, but then fall into a "plateau" and it is difficult to improve further. This is mainly because the existing methods lack an effective adaptive mechanism to continuously introduce new information and adjust the training strategy to cope with data changes. Low resource utilization efficiency: In pursuit of high accuracy, some existing methods do not hesitate to invest a lot of computing resources and time costs for training. However, this "brute force" way to improve accuracy not only increases costs, but also causes resource waste, especially in actual production environments with limited resources. Lack of flexibility: Traditional equipment model training methods are often optimized for specific tasks or data sets, lacking versatility and flexibility. When the application scenario or data set changes, the training strategy needs to be redesigned, which increases the difficulty and cost of development. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a device model adaptive training method, system, storage medium and electronic device with improved accuracy in response to the problems existing in the prior art.

[0005] The technical solution adopted by the present invention to solve the technical problem is: constructing an adaptive training method for a device model with improved accuracy, comprising the following steps:

[0006] Acquire factory data and real-time operation data of the equipment, and pre-process the factory data and real-time operation data to obtain a dynamic training data set;

[0007] Performing device model training based on the dynamic training data set to obtain a trained device model;

[0008] Verifying the trained device model to obtain a verified device model;

[0009] Determining whether the accuracy of the verified device model meets the accuracy requirements;

[0010] If the accuracy requirement is not met, retraining the device model;

[0011] If the accuracy requirement is met, the device model is deployed and put into use;

[0012] After the device model is deployed and put into use, continue to monitor the accuracy of the device model;

[0013] If the accuracy of the device model does not meet the accuracy requirement, the device model is updated online.

[0014] In the method for adaptively training a device model with improved accuracy of the present invention, retraining the device model includes:

[0015] Determining whether the samples of the dynamic training data set have reached a maximum threshold;

[0016] If the maximum threshold is not reached, the amount of running data in the preset time period is added to obtain an updated training data set;

[0017] The device model is trained based on the updated training data set.

[0018] In the method for adaptively training a device model with improved accuracy of the present invention, if the accuracy requirement is not met, retraining the device model includes:

[0019] If the accuracy requirement is not met, a preset method is used to add new data to the dynamic training data set to obtain a dynamically updated training data set;

[0020] The device model is retrained based on the dynamically updated training data set.

[0021] In the device model adaptive training method with improved accuracy described in the present invention, the preset method is a sliding window data update method.

[0022] In the device model adaptive training method with improved accuracy described in the present invention, the method further includes:

[0023] After adding a new amount of data to the dynamic training data set;

[0024] Determine whether the sample of the dynamically updated training data set is greater than a threshold;

[0025] If not, retraining the device model based on the dynamically updated training data set;

[0026] If so, old data exceeding the threshold is removed.

[0027] In the method for adaptively training a device model with improved accuracy of the present invention, determining whether the accuracy of the verified device model meets the accuracy requirement includes:

[0028] calculating a predicted value of the device model;

[0029] Calculating according to the real-time data of the predicted value to obtain the accuracy of the device model;

[0030] Determining whether the accuracy of the device model is continuously greater than an accuracy threshold;

[0031] If so, it is determined that the accuracy requirement is met; if not, it is determined that the accuracy requirement is not met.

[0032] In the method for adaptive training of equipment models with improved accuracy of the present invention, the acquisition of equipment factory data and real-time operation data, and preprocessing of the factory data and real-time operation data to obtain a dynamic training data set includes:

[0033] Obtaining factory data of the device to obtain an original training data set;

[0034] Using a data quality screening method to screen the original training data set to obtain primary screening data;

[0035] Screening the primary screening data according to the business logic screening method to obtain secondary screening data;

[0036] Perform model pre-training based on the secondary screening data to obtain an initialization data sample;

[0037] Acquiring the real-time operation data;

[0038] The real-time operation data is added to the initialization data sample to obtain the dynamic training data set; the real-time operation data is the operation data collected in real time after the equipment model is pre-trained and put into the online collection system.

[0039] The present invention also provides a device model adaptive training system with improved accuracy, comprising:

[0040] A data collection and processing unit is used to obtain the factory data and real-time operation data of the equipment, and pre-process the factory data and real-time operation data to obtain a dynamic training data set;

[0041] A model training unit, used to perform device model training based on the dynamic training data set to obtain a trained device model;

[0042] A model verification unit, used to verify the trained device model and obtain a verified device model;

[0043] An accuracy evaluation unit, used to determine whether the accuracy of the verified device model meets the accuracy requirements; if it does not meet the accuracy requirements, retrain the device model; if it meets the accuracy requirements, deploy the device model for online use;

[0044] The model updating unit is used to continue monitoring the accuracy of the device model after the device model is deployed and put into use online; if the accuracy of the device model does not meet the accuracy requirement, the device model is updated online.

[0045] The present invention also provides a storage medium storing a computer program, wherein the computer program is suitable for loading by a processor to execute the steps of the device model adaptive training method with improved accuracy as described above.

[0046] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the steps of the above-mentioned method for adaptive training of a device model with improved accuracy by calling the computer program stored in the memory.

[0047] The device model adaptive training method, system, storage medium and electronic device for improving the accuracy of the present invention have the following beneficial effects: including obtaining the factory data and real-time operation data of the device, and preprocessing the factory data and real-time operation data to obtain a dynamic training data set; training the device model based on the dynamic training data set to obtain a trained device model; verifying the trained device model to obtain a verified device model; judging whether the accuracy of the verified device model meets the accuracy requirements; if it does not meet the accuracy requirements, retraining the device model; if it meets the accuracy requirements, deploying the device model for online use; continuing to monitor the accuracy; if the accuracy does not meet the accuracy requirements, updating the device model online. The present invention can effectively deal with the impact of dynamic changes in data on model performance, improve the adaptability and robustness of the model, and can also improve the prediction accuracy and reliability of the model and reduce operating costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0049] Figure 1 It is a flow chart of the method for adaptive training of device models with improved accuracy provided by the present invention;

[0050] Figure 2 It is a flowchart of a preferred embodiment of a method for adaptively training a device model with improved accuracy provided by the present invention;

[0051] Figure 3 It is a principle block diagram of the device model adaptive training system with improved accuracy provided by the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] The present invention aims to solve the problems of static training strategy limitations, accuracy improvement bottlenecks, low resource utilization efficiency and lack of flexibility in existing equipment model training methods and systems, and provides an adaptive training method for equipment models with improved accuracy. Through adaptive training and gradual accuracy improvement, the prediction or calculation accuracy of the equipment model is improved, resource consumption is reduced, and the versatility and flexibility of the method are enhanced.

[0054] refer to Figure 1 , Figure 1 A flowchart of the device model adaptive training method with improved accuracy provided by the present invention.

[0055] Specifically, Figure 1 As shown, the device model adaptive training method with improved accuracy includes the following steps:

[0056] Step S101: Obtain equipment factory data and real-time operation data, and pre-process the factory data and real-time operation data to obtain a dynamic training data set.

[0057] In an embodiment of the present invention, the acquisition of equipment factory data and real-time operation data, and preprocessing of the factory data and real-time operation data to obtain a dynamic training data set includes: acquiring the equipment factory data to obtain an original training data set; filtering the original training data set using a data quality screening method to obtain primary screening data; filtering the primary screening data according to a business logic screening method to obtain secondary screening data; pre-training a model based on the secondary screening data to obtain an initialization data sample; acquiring the real-time operation data; adding the real-time operation data to the initialization data sample to obtain the dynamic training data set. The real-time operation data is the operation data collected in real time after the equipment model pre-training is put on line in the collection system.

[0058] Step S102: Performing device model training based on the dynamic training data set to obtain a trained device model.

[0059] Step S103: verifying the trained device model to obtain a verified device model.

[0060] Step S104: Determine whether the accuracy of the verified device model meets the accuracy requirement.

[0061] In the embodiment of the present invention, the step of determining whether the accuracy of the verified device model meets the accuracy requirement includes: calculating the predicted value of the device model; performing calculations based on the real-time data of the predicted value to obtain the accuracy of the device model; determining whether the accuracy of the device model is continuously greater than the accuracy threshold; if so, determining that the accuracy requirement is met; if not, determining that the accuracy requirement is not met. If the accuracy requirement is not met, retraining the device model.

[0062] In an embodiment of the present invention, the retraining of the device model includes: determining whether the samples of the dynamic training data set have reached a maximum threshold (such as 3 years of hourly samples); if the maximum threshold has not been reached, adding the amount of operating data for a preset time period to obtain an updated training data set; and training the device model based on the updated training data set.

[0063] In an embodiment of the present invention, if the accuracy requirement is not met, retraining the device model includes: if the accuracy requirement is not met, adding a new amount of data to the dynamic training data set using a preset method to obtain a dynamically updated training data set; and retraining the device model based on the dynamically updated training data set.

[0064] Optionally, the preset method is a sliding window data update method.

[0065] In an embodiment of the present invention, after adding a new amount of data to the dynamically updated training data set, it is determined whether the sample of the dynamically updated training data set is greater than a threshold; if not, the device model is retrained based on the dynamically updated training data set; if so, the old data exceeding the threshold is discarded.

[0066] Step S105: If the accuracy requirement is met, the device model is deployed and put into use.

[0067] Step S106: After the device model is deployed and put into use, the accuracy of the device model continues to be monitored; if the accuracy of the device model does not meet the accuracy requirement, the device model is updated online.

[0068] Specifically, in a preferred embodiment, Figure 2 As shown, first obtain the factory data of the equipment or the existing historical data (such as the factory data of chillers, refrigeration pumps, cooling pumps, cooling towers and other equipment), and then use the data quality screening method to remove duplicate, missing, and stuck samples from the factory data (that is, clean the missing values, duplicate values, and out-of-limit values ​​of the data to ensure data quality), obtain the first-screened data, and use the business logic screening method to filter it to obtain the second-screened data, that is, remove the maximum and minimum values ​​and obvious abnormal values ​​according to the normal range determined by the business logic. For example, the outdoor temperature data of a certain area removes values ​​less than -10°C and greater than 40°C, and the data before and after the refrigerator is switched on. The model pre-training screening method is used to screen the secondary screening data. That is, after confirming the input variables and output variables according to the sample data, the model is quickly trained to eliminate data that obviously contributes little to the model (that is, samples with extremely large deviations between the predicted values ​​and the actual values ​​after the model is trained). The initialization data sample is obtained. Finally, after the equipment model is put into the collection system, data is collected in real time to obtain real-time operation data, and the real-time operation data is accumulated into the initialization data sample. Data is accumulated gradually until the maximum threshold of the trigger data sample is reached (such as 3 years of hourly samples) to obtain a dynamic training data set.

[0069] In an embodiment of the present invention, after obtaining a dynamic training data set, device model training can be performed based on the dynamic training data set. The dynamic training data triggers device model training at preset intervals, that is, there is a minimum sample number requirement for device model training, and the training of the device model is triggered when the minimum sample requirement is reached. Among them, the device model may include but is not limited to a regression model, a mechanism formula model, a machine learning model, etc. When the accuracy of the trained device model is continuously greater than the accuracy threshold (wherein the accuracy threshold can be set to 0.95, the closer the accuracy threshold, the better), such as a device model that has been trained for 10 times, and its model accuracy R2 is greater than 0.95, then it is judged that the accuracy of the device model meets the accuracy requirements. When the accuracy requirements are met, it can be deployed to the project for operation.

[0070] like Figure 2As shown in the figure, when the trained equipment model meets the accuracy requirements and is deployed online, the real-time data input by the model is collected, and the model prediction value is calculated. The accuracy of the model is calculated by the output real-time value and the model prediction value. For example, the chiller power model needs to input the chilled water outlet temperature, cooling water return temperature, and cooling capacity, and then the equipment model calculates the predicted power according to the input prediction, and then calculates the collected real-time value, such as the collected real-time power, with the predicted power output by the equipment model to obtain the accuracy of the equipment model. Among them, for the calculation of the accuracy of the equipment model, in order to ensure the accuracy and reliability of the accuracy calculation, it is necessary to calculate the R2 value of a recent period of time (such as the R2 value of the past 7 days). If the R2 values ​​of the past 7 days are all greater than 0.95, it is judged that the accuracy of the equipment model meets the accuracy requirements, and the equipment model can continue to be used. If the R2 value of the past 7 days is less than 0.95, it is judged that the accuracy of the equipment model is insufficient and needs to be retrained (such as parameter optimization, structural adjustment, etc.). The present invention sets a model accuracy threshold as a criterion for evaluating model performance. When the accuracy of the device model (i.e., prediction accuracy) is lower than the accuracy threshold, the model retraining process is automatically triggered. During the retraining process, the model is adaptively adjusted (such as parameter optimization, structural adjustment, etc.) according to real-time operating data, and the performance of the retrained model is evaluated through a validation set, thereby achieving dynamic optimization and continuous improvement of model performance.

[0071] like Figure 2 As shown, the device model deployed and put into use needs to continue to monitor its prediction accuracy. When the prediction accuracy does not meet the requirements, the device model needs to be updated online. Specifically, when the device model deployed and put into use collects the R2 accuracy of a recent period of time (such as the past 7 days), and it is lower than the benchmark threshold for a period of time (such as lower than 0.90 for 24 hours), it triggers the retraining of the model. If the samples of the original training model are less than the maximum threshold, the number of recent periods of time (such as the past 7 days) is added, and the model is retrained; if the accuracy of the device model is not enough, the new data volume is added forward in a sliding window manner for 7 days until the accuracy R2 of the trained device model is greater than the threshold (that is, the accuracy R2 of the device model trained for 10 times is greater than 0.95). In other words, the device model deployed to the project still needs to continue to execute the above logic. It should be noted that after adding the new data volume, if the sample of the original training model is greater than the maximum threshold, the old data that exceeds the threshold needs to be eliminated, which can be specifically: the pre-training model calculation error is removed in sequence. The sample with a large contribution to the error. The present invention adopts a data update strategy based on a sliding window, which can continuously track data changes and optimize model performance. Among them, the data update strategy based on the sliding window sets a fixed-size time window (such as the past 7 days) to regularly collect new data and eliminate old data to maintain the timeliness and representativeness of the training data set. At the same time, through the continuous sliding of the sliding window, continuous data updating and continuous training of the model are achieved.

[0072] refer to Figure 3 , Figure 3 The invention provides an adaptive training system for device models with improved accuracy.

[0073] Specifically, Figure 3 As shown, the device model adaptive training system with improved accuracy includes:

[0074] The data collection and processing unit 301 is used to obtain the factory data and real-time operation data of the equipment, and pre-process the factory data and real-time operation data to obtain a dynamic training data set.

[0075] The model training unit 302 is used to perform device model training based on the dynamic training data set to obtain a trained device model.

[0076] The model verification unit 303 is used to verify the trained device model to obtain a verified device model.

[0077] The accuracy evaluation unit 304 is used to determine whether the accuracy of the verified device model meets the accuracy requirements; if it does not meet the accuracy requirements, the device model is retrained; if it meets the accuracy requirements, the device model is deployed and put into use.

[0078] The model updating unit 305 is used to continue monitoring the accuracy of the device model after the device model is deployed online for use; if the accuracy of the device model does not meet the accuracy requirement, update the device model online.

[0079] Specifically, the specific coordination operation process between the units in the precision-enhanced device model adaptive training system here can refer to the precision-enhanced device model adaptive training method mentioned above, which will not be repeated here.

[0080] The present invention can improve the adaptability and prediction accuracy of the equipment model and reduce the operating cost. Specifically, the present invention can effectively deal with the impact of dynamic data changes on model performance and improve the adaptability and robustness of the model through dynamic data preprocessing and selection mechanism and sliding window-based update strategy. At the same time, the adaptive model retraining and evaluation mechanism is adopted to ensure that the model can perform performance optimization and accuracy improvement in a timely manner when facing new data, thereby improving the prediction accuracy and reliability of the model. The design of the integrated training system simplifies the operation process, improves the training efficiency, and reduces the user's operating cost and time cost during use.

[0081] In addition, an electronic device of the present invention includes a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program to implement a device model adaptive training method with improved accuracy as any one of the above. Specifically, according to an embodiment of the present invention, the process described with reference to the flowchart above can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed by an electronic device and when executed, it performs the above functions defined in the method of the embodiment of the present invention. The electronic device in the present invention can be a terminal such as a notebook, a desktop, a tablet computer, a smart phone, or a server.

[0082] In addition, a storage medium of the present invention stores a computer program thereon, and when the computer program is executed by a processor, an adaptive training method for a device model with improved accuracy of any of the above items is implemented. Specifically, it should be noted that the storage medium of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer readable signal media may also be any computer readable medium other than computer readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0083] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0084] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0085] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may 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 the present invention.

[0086] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0087] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot limit the scope of protection of the present invention. All equivalent changes and modifications made to the scope of the claims of the present invention should fall within the scope of the claims of the present invention.

Claims

1. A method for adaptive training of device models with improved accuracy, characterized in that: The following steps are involved: Acquire factory data and real-time operation data of the equipment, and pre-process the factory data and real-time operation data to obtain a dynamic training data set; Performing device model training based on the dynamic training data set to obtain a trained device model; Verifying the trained device model to obtain a verified device model; Determining whether the accuracy of the verified device model meets the accuracy requirements; If the accuracy requirement is not met, retraining the device model; If the accuracy requirement is met, the device model is deployed and put into use; After the device model is deployed and put into use, continue to monitor the accuracy of the device model; If the accuracy of the device model does not meet the accuracy requirement, the device model is updated online.

2. The method for adaptively training a device model with improved accuracy according to claim 1, characterized in that: The retraining of the device model comprises: Determining whether the samples of the dynamic training data set have reached a maximum threshold; If the maximum threshold is not reached, the amount of running data in the preset time period is added to obtain an updated training data set; The device model is trained based on the updated training data set.

3. The method for adaptively training a device model with improved accuracy according to claim 1, characterized in that: If the accuracy requirement is not met, retraining the device model includes: If the accuracy requirement is not met, a preset method is used to add new data to the dynamic training data set to obtain a dynamically updated training data set; The device model is retrained based on the dynamically updated training data set.

4. The method for adaptively training a device model with improved accuracy according to claim 3, characterized in that: The preset method is a data update method of a sliding window.

5. The method for adaptively training a device model with improved accuracy according to claim 3, characterized in that: The method further comprises: After adding a new amount of data to the dynamic training data set; Determine whether the sample of the dynamically updated training data set is greater than a threshold; If not, retraining the device model based on the dynamically updated training data set; If so, old data exceeding the threshold is removed.

6. The method for adaptively training a device model with improved accuracy according to claim 1, characterized in that: The determining whether the accuracy of the verified device model meets the accuracy requirement includes: calculating a predicted value of the device model; Calculating according to the real-time data of the predicted value to obtain the accuracy of the device model; Determining whether the accuracy of the device model is continuously greater than an accuracy threshold; If so, it is determined that the accuracy requirement is met; if not, it is determined that the accuracy requirement is not met.

7. The method for adaptively training a device model with improved accuracy according to claim 1, characterized in that: The obtaining of equipment factory data and real-time operation data, and preprocessing the factory data and real-time operation data to obtain a dynamic training data set includes: Obtaining factory data of the device to obtain an original training data set; Using a data quality screening method to screen the original training data set to obtain primary screening data; Screening the primary screening data according to the business logic screening method to obtain secondary screening data; Perform model pre-training based on the secondary screening data to obtain an initialization data sample; Acquiring the real-time operation data; The real-time operation data is added to the initialization data sample to obtain the dynamic training data set; the real-time operation data is the operation data collected in real time after the equipment model is pre-trained and put into the online collection system.

8. A device model adaptive training system with improved accuracy, characterized in that: include: A data collection and processing unit is used to obtain the factory data and real-time operation data of the equipment, and pre-process the factory data and real-time operation data to obtain a dynamic training data set; A model training unit, used to perform device model training based on the dynamic training data set to obtain a trained device model; A model verification unit, used to verify the trained device model and obtain a verified device model; An accuracy evaluation unit, used to determine whether the accuracy of the verified device model meets the accuracy requirements; if it does not meet the accuracy requirements, retrain the device model; if it meets the accuracy requirements, deploy the device model for online use; The model updating unit is used to continue monitoring the accuracy of the device model after the device model is deployed and put into use online; if the accuracy of the device model does not meet the accuracy requirement, the device model is updated online.

9. A storage medium, characterized in that: The storage medium stores a computer program, which is suitable for being loaded by a processor to execute the steps of the device model adaptive training method with improved accuracy as described in any one of claims 1 to 7.

10. An electronic device, characterized in that: It includes a memory and a processor, wherein a computer program is stored in the memory, and the processor executes the steps of the device model adaptive training method with improved accuracy as described in any one of claims 1 to 7 by calling the computer program stored in the memory.

Images