Dual-recognition network anomaly prediction and compensation control method and system for submersible motor

Abstract

The double-recognition network abnormality prediction and submersion motor compensation control method and system provided by the application relate to the motor control technical field, establish an abnormality prediction model, perform equipment monitoring of underwater special power equipment, establish a time sequence monitoring data set, perform model self-checking, complete dynamic weight configuration of the first and second abnormality recognition networks according to the self-checking result, input the time sequence monitoring data set after preprocessing into the abnormality prediction model, generate an abnormality prediction result, perform abnormality evaluation on the result, establish a ladder compensation strategy, and thus perform optimized control, and complete optimized control through verification and identification of the optimized control of the verification window, thereby solving the technical problems that in the process of submersion motor control optimization, the equipment operation efficiency is low, the stability and adaptability are poor, and the abnormality prediction and optimization cannot be performed in time and accurately, and the technical effects of accurate abnormality prediction and optimized control of the submersion motor and improved equipment operation efficiency and safety are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of motor control, and particularly relates to a double-identification network abnormality prediction underwater motor compensation control method and system. BACKGROUND

[0002] Underwater special power equipment, especially submersible motors, play an important role in many fields such as ocean engineering, water resources development, and fisheries. As underwater special power equipment, submersible motors have complex and variable working environments, such as water temperature, water pressure, and water quality, which can affect their operation. At the same time, submersible motors also need to face challenges such as load changes and power fluctuations during operation. Therefore, the optimization of submersible motor control is of great significance to improve its operating efficiency, stability, and safety.

[0003] The prior art often has the technical problems of low operating efficiency, poor stability and adaptability of the equipment in the process of submersible motor control optimization, and cannot timely and accurately predict and optimize abnormalities, thereby increasing the failure rate and maintenance cost. SUMMARY

[0004] The present application provides a double-identification network abnormality prediction underwater motor compensation control method and system for solving the technical problems of low operating efficiency, poor stability and adaptability of the equipment in the process of submersible motor control optimization, and inability to timely and accurately predict and optimize abnormalities, thereby increasing the failure rate and maintenance cost.

[0005] In view of the above problems, the present application provides a double-identification network abnormality prediction underwater motor compensation control method and system.

[0006] In a first aspect, the present application provides a double-identification network abnormality prediction underwater motor compensation control method, which is applied to a double-identification network abnormality prediction underwater motor compensation control system, and the method comprises:

[0007] The abnormal prediction model is a model for performing abnormal prediction on the underwater special power equipment, and the abnormal prediction model comprises a first abnormal identification network and a second abnormal identification network, the first abnormal identification network is constructed through a use data set of the underwater special power equipment, and the second abnormal identification network is constructed through big data; device monitoring of the underwater special power equipment is performed, and a time sequence monitoring data set is established; model self-checking of the abnormal prediction model is performed, and dynamic weight configuration of the first abnormal identification network and the second abnormal identification network is completed according to a model self-checking result; after the time sequence monitoring data set is preprocessed, the time sequence monitoring data set is input into the abnormal prediction model after dynamic weight configuration, and an abnormal prediction result is generated; abnormal evaluation is performed on the abnormal prediction result, a ladder compensation strategy is established according to an abnormal evaluation result, optimization control is performed through the ladder compensation strategy, and a verification window is set; verification and identification of the optimization control are performed through the verification window, and optimization control is completed based on a verification and identification result.

[0008] In a second aspect, the application provides a double-identification network abnormal prediction compensation control system for a submersible motor, which comprises:

[0009] A model building module is configured to build an abnormal prediction model, the abnormal prediction model being a model for performing abnormal prediction on underwater special power equipment, the abnormal prediction model comprising a first abnormal identification network and a second abnormal identification network, the first abnormal identification network being constructed through a use data set of the underwater special power equipment, and the second abnormal identification network being constructed through big data; a device monitoring module is configured to perform device monitoring of the underwater special power equipment and establish a time sequence monitoring data set; a model self-checking module is configured to perform model self-checking of the abnormal prediction model and complete dynamic weight configuration of the first abnormal identification network and the second abnormal identification network according to a model self-checking result; an abnormal prediction module is configured to input the time sequence monitoring data set into the abnormal prediction model after dynamic weight configuration after preprocessing of the time sequence monitoring data set, and generate an abnormal prediction result; a strategy compensation module is configured to perform abnormal evaluation on the abnormal prediction result, establish a ladder compensation strategy according to an abnormal evaluation result, perform optimization control through the ladder compensation strategy, and set a verification window; and a verification and identification module is configured to perform verification and identification of the optimization control through the verification window, and complete optimization control based on a verification and identification result.

[0010] The one or more technical solutions provided in the application have at least the following technical effects or advantages:

[0011] The double-recognition network abnormality prediction submersion motor compensation control method provided by the application, by establishing an abnormality prediction model, the abnormality prediction model is a model for abnormality prediction of underwater special power equipment, the abnormality prediction model includes a first abnormality recognition network and a second abnormality recognition network, the first abnormality recognition network is constructed by using data set of underwater special power equipment, and the second abnormality recognition network is constructed by big data; device monitoring of the underwater special power equipment is performed, and a time sequence monitoring data set is established; model self-checking of the abnormality prediction model is performed, and dynamic weight configuration of the first abnormality recognition network and the second abnormality recognition network is completed according to the model self-checking result; after preprocessing the time sequence monitoring data set, the time sequence monitoring data set is input into the abnormality prediction model after dynamic weight configuration, and an abnormality prediction result is generated; the abnormality prediction result is evaluated, a ladder compensation strategy is established according to the abnormality evaluation result, optimization control is performed through the ladder compensation strategy, and a verification window is set; the optimization control is checked and identified through the verification window, and the optimization control is completed based on the checking and identifying result, which solves the technical problems of low equipment operation efficiency, poor stability and adaptability, and inability to timely and accurately perform abnormality prediction and optimization in the process of submersion motor control optimization, improves the failure rate and maintenance cost, achieves accurate abnormality prediction and optimization control of the submersion motor, improves the operation efficiency and safety of the equipment, and reduces the failure rate and maintenance cost. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 A double-recognition network abnormality prediction submersion motor compensation control method flowchart is provided for the application.

[0013] Figure 2 A double-recognition network abnormality prediction submersion motor compensation control system structure diagram is provided for the application.

[0014] The reference signs are explained: model building module 11, device monitoring module 12, model self-checking module 13, abnormality prediction module 14, strategy compensation module 15, and checking and identifying module 16. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical scheme and advantages of the application more clear, the application is further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the application and do not limit the application. EMBODIMENT

[0016] As shown in Figure 1 The double-recognition network abnormality prediction submersion motor compensation control method provided by the application, the method is applied to a double-recognition network abnormality prediction submersion motor compensation control system, and the method includes:

[0017] Step S100: establishing an anomaly prediction model, which is a model for anomaly prediction of the underwater special power equipment, the anomaly prediction model comprising a first anomaly identification network and a second anomaly identification network, the first anomaly identification network being constructed by a use data set of the underwater special power equipment, and the second anomaly identification network being constructed by big data.

[0018] The underwater special power equipment is an electrical equipment designed and optimized for underwater environment. These equipments usually need to operate in harsh marine environment, so they need to have key characteristics such as corrosion resistance, sealing, pressure resistance, waterproofness, etc. And since underwater power equipment is usually used in fields such as marine engineering and scientific research, it may also need to meet some special application requirements, such as high torque, high power output, high precision control, etc. Therefore, it is necessary to study the optimization control of underwater special power equipment. Underwater power equipment generally includes underwater cable, submersible motor, underwater transformer, frequency converter, etc. The method proposed in the embodiments of the present application is mainly used for control optimization of submersible motor, and provides a reliable control scheme for underwater kinetic energy application.

[0019] First, in order to make anomaly prediction of the submersible motor, an anomaly prediction model is established, which includes a first anomaly identification network and a second anomaly identification network, the first anomaly identification network being constructed by a use data set of the submersible motor, and the second anomaly identification network being constructed by big data. Specifically, the historical use data of the submersible motor is collected, including but not limited to the operating state, temperature, pressure, voltage, current and other parameters of the motor, as well as any related log information or fault records; the acquired data is removed for duplication, incompleteness or errors, the data is standardized to ensure the consistency and comparability of the data, and meaningful features such as statistical features, time series features, etc. are extracted from the original data, which will be used to train the first anomaly identification network. Similarly, big data related to the submersible motor is collected from various sources, including but not limited to historical data of other similar equipment, industry reports, public data sets, etc., and data preprocessing similar to the data set of the first anomaly identification network is performed to obtain the data set of the second anomaly identification network.

[0020] Next, the construction of the anomaly prediction model is performed. First, the first anomaly identification network is built, that is, based on the characteristics of the data set of the submersible motor, a suitable machine learning or deep learning algorithm is selected, such as support vector machine (SVM), random forest, neural network, etc., and the extracted features and corresponding labels (normal / abnormal) are used to train the model, which can be annotated by experts or automatically generated based on historical fault records. After training, the performance of the model is evaluated using cross-validation and other methods, and the parameters are adjusted as needed to obtain a model with a first anomaly identification network. Similarly, the second anomaly identification network is built in a similar way to the first anomaly identification network, but due to the complexity and diversity of big data, a more powerful deep learning algorithm such as convolutional neural network (CNN) and recurrent neural network (RNN) is selected. Pre-training is performed using a large data set to obtain a pre-trained model, which enables the model to learn more extensive submersible motor anomaly patterns. The parameters of the pre-trained model are migrated to the first anomaly identification network, and fine-tuning is performed using the submersible motor data set to improve the model's prediction ability for specific equipment. Through the above steps, the entire anomaly prediction model is built, and finally the trained anomaly prediction model can be deployed in the actual application environment to interact with the submersible motor in real time, thereby monitoring the running state of the submersible motor in real time, and inputting real-time data into the anomaly prediction model to obtain the anomaly prediction result of the equipment.

[0021] By building an anomaly prediction model from the use level and big data level, the individualized anomaly detection response needs of special equipment are met, and the deep analysis of big data is integrated, thereby ensuring the comprehensiveness and accuracy of model prediction.

[0022] Step S200: Perform device monitoring of the underwater special power equipment to establish a time series monitoring data set.

[0023] Optionally, the monitoring needs and monitoring targets are determined, that is, performance optimization is taken as the goal to determine the submersible motor parameters that need to be monitored, such as current, voltage, temperature, vibration, insulation performance, etc. According to the monitoring needs, appropriate sensors are selected, such as temperature sensors, vibration sensors, current sensors, etc., and a real-time online monitoring system is used to continuously and real-time monitor the submersible motor. A specific monitoring scheme is designed, including monitoring points, monitoring frequency, and data acquisition design. The specific positions of the sensors to be installed on the submersible motor are determined to ensure that the running state of the motor can be monitored comprehensively and accurately. According to the running condition of the motor and the monitoring needs, a suitable monitoring frequency is set, such as collecting data once a minute, once an hour, or once a day. In addition, a data acquisition system is designed to ensure that the data of the sensors can be collected in real time and accurately. When collecting data, a specific data format is designed, including time stamp, sensor type, sensor position, monitoring data, etc. The collected data is uploaded to the time series database in real time to form a time series monitoring data set.

[0024] By establishing the time series monitoring dataset, the change trend of each parameter of the submersible motor can be analyzed by time series analysis and other methods to predict possible problems. Meanwhile, abnormal values in the data can be detected by statistical methods or machine learning algorithms to identify possible faults or abnormalities of the submersible motor, so as to evaluate the performance status of the submersible motor, such as operating efficiency, energy consumption, etc., thereby laying a data analysis foundation for subsequent optimization control.

[0025] Step S300: performing model self-checking of the abnormal prediction model, and completing dynamic weight configuration of the first abnormal identification network and the second abnormal identification network according to a model self-checking result.

[0026] For example, in order to verify the accuracy, stability and generalization ability of the abnormal prediction model and ensure that the model can work effectively on new and unseen data, the model can be self-checked. Specifically, a part of the known labeled dataset (different from the training set) is used as a verification set to perform self-checking of the model, and the verification set should contain normal data and abnormal data; appropriate indexes for evaluating the abnormal detection model are selected, such as accuracy and recall rate, the model is evaluated using the verification set data, and the evaluation results are recorded. According to the model self-checking result, the performance of the model on different types of data is analyzed to find out possible problems and improvement space. The evaluation results of the model on the verification set are compared with the expected target to determine whether the model meets the requirements; the false positives (identifying normal data as abnormal) and the false negatives (identifying abnormal data as normal) of the model are observed and the reasons are analyzed; and the dependence of the model on each feature is analyzed to find out the features that have greater impact on the performance of the model. Finally, according to the model self-checking result, the weights of the first abnormal identification network and the second abnormal identification network are dynamically adjusted to make the overall model performance optimal. When adjusting the weights, if the first abnormal identification network performs better in the self-checking, its weight can be appropriately increased to make its contribution greater in the overall model; similarly, if the second abnormal identification network performs better on specific types of data, its weight can be increased on these data. Specifically, a weight distribution mechanism can be designed to collect data and evaluate the performance of the model in real time during the operation of the model, and then the weights of the two networks are dynamically adjusted according to the performance changes. In addition to adjusting the weights, an ensemble learning method can also be used to fuse the prediction results of the two networks to improve the performance of the overall model. The specific fusion method can be realized by simple weighted average or more complex ensemble learning algorithms (such as stacked ensemble).

[0027] By dynamically configuring the weights of the abnormal identification network based on model self-checking, the model can maintain optimal performance in actual application to improve the accuracy of optimization control.

[0028] Step S400: After preprocessing the time series monitoring dataset, input it into the abnormality prediction model with dynamic weight configuration to generate abnormality prediction results.

[0029] Further, before inputting the time series monitoring dataset into the abnormality prediction model, data preprocessing is usually required to improve the quality of data and the performance of the model. Data preprocessing includes data cleaning, data conversion, feature selection and feature engineering, i.e. first remove noise, missing values, outliers or duplicates in the data, and normalize, normalize or discretize the data as needed so that the model can better process the data; select the most relevant features from the original dataset to simplify the model and improve prediction accuracy; and create new features or modify existing features to capture hidden patterns or relationships in the data.

[0030] After data preprocessing is complete, the abnormality prediction model that has been dynamically weighted needs to be loaded and configured. Specifically, load the pre-trained abnormality prediction model from the storage location, set the weights of each component (such as the first abnormality identification network and the second abnormality identification network) in the model according to the previous dynamic weight configuration results, and finally check the parameters and status of the model to ensure that the model has been correctly loaded and is ready for prediction. When the model is loaded and configured, the preprocessed time series monitoring dataset can be input into the model to generate abnormality prediction results. Specifically, if the dataset is large, it may need to be divided into smaller batches for processing to improve computational efficiency; pass the preprocessed data as input to the abnormality prediction model, which uses the input data to calculate and generate abnormality prediction results, which may include abnormality scores, abnormality labels (normal / abnormal) or abnormality probabilities, etc.

[0031] Step S500: Evaluate the abnormality prediction results to establish a step compensation strategy based on the evaluation results, optimize control through the step compensation strategy, and set a verification window.

[0032] Furthermore, indicators for evaluating the abnormal prediction results are determined, including accuracy, recall rate, false positive rate, etc. The specific indicators depend on the actual demand and the characteristics of the prediction model. The selected evaluation indicators are used to evaluate the abnormal prediction results, i.e. comparing the prediction results with the true situation and calculating the corresponding indicator values. According to the evaluation results, the abnormal prediction results are classified according to the severity, for example, they can be classified into slight abnormality, medium abnormality and severe abnormality, etc. Through classification, it lays the foundation for subsequent development of different levels of compensation strategies. Next, according to the abnormal severity classification, the compensation measures for different levels of abnormality are determined, which can include adjusting control parameters, sending alarm notifications, starting backup devices, etc. The compensation measures are set in a ladder according to the severity of the abnormality, for example, for slight abnormality, only parameter fine-tuning may be needed; while for severe abnormality, backup devices may need to be started immediately and relevant personnel notified. At the same time, trigger conditions are set for each level of compensation measures, which can be based on the confidence of the prediction results, the duration of the abnormality, etc.

[0033] During the optimization control process, the system is monitored in real time to timely discover new abnormal prediction results, and when the trigger conditions of a certain level are met, the corresponding compensation measures are automatically triggered for optimization control. At the same time, the execution of the compensation measures is tracked and recorded for subsequent analysis and optimization. Finally, verification indicators for verifying the effect of optimization control are selected, which are closely related to the control objectives and system performance. According to the control requirements and data update frequency, the size of the verification window is determined. It should be noted that the verification window should be large enough to contain enough data to evaluate the effect of optimization control. Data is collected within the verification window, and the effect of optimization control is evaluated using the selected verification indicators. The difference between the performance indicators before and after optimization is compared to generate an evaluation result. Finally, according to the evaluation result, the ladder compensation strategy and optimization control parameters are adjusted and optimized to improve the performance of the system and the effect of abnormal handling.

[0034] Step S600: Verification and identification of optimization control through the verification window, and completion of optimization control based on the verification and identification results.

[0035] Specifically, data related to the optimization control, including system state, performance indicators, abnormality prediction results, and the execution of compensation measures, is collected in real time within the verification window. Based on the collected data, verification and identification are performed to verify whether the optimization control strategy works as expected and achieves the expected effect. Specifically, the data before and after the implementation of the optimization control strategy is compared and analyzed to observe whether the performance indicators have improved significantly; the trend of the data is analyzed to determine whether the optimization control strategy has a positive impact on the system; it is checked whether new abnormalities still occur after the implementation of the optimization control strategy and whether these abnormalities have been effectively handled. The results of the verification and identification are further evaluated according to the preset evaluation criteria (such as performance improvement percentage, abnormality reduction rate, etc.) to determine whether the optimization control strategy is successful. If the evaluation result meets the expectation, it means that the optimization control strategy is effective; otherwise, further analysis is needed to make adjustments. Further, based on the evaluation of the verification and identification results, the optimization control strategy is adjusted. If the strategy is successful, further optimization can be considered to improve performance; if the strategy fails, the reasons for the failure need to be analyzed and appropriate improvement measures are taken. The adjustment direction can include parameter adjustment, algorithm improvement, and compensation measure optimization, etc. Since the adjustment of the optimization control strategy is an iterative process, the verification window needs to be reset and data needs to be collected for verification and identification after each adjustment. Through continuous verification and adjustment, the optimization control strategy can be gradually improved to better meet the actual needs and achieve the best performance, thereby ensuring the effectiveness and reliability of the optimization control strategy in practical application, and improving the performance and stability of the system.

[0036] Further, the model self-check of the execution of the abnormality prediction model is performed, and the dynamic weight configuration of the first abnormality identification network and the second abnormality identification network is completed according to the model self-check result. The step S300 of the present application further comprises:

[0037] Step S310: calling a self-check processing layer, performing data richness evaluation of the first abnormality identification network, and constructing a first self-check result according to the data richness evaluation result, wherein the self-check processing layer is a sub-processing layer of the abnormality prediction model.

[0038] Step S320: calling the self-check processing layer, performing data adaptation evaluation of the second abnormality network, constructing a second self-check result according to the adaptation evaluation result, integrating the first self-check result and the second self-check result, and completing the dynamic weight configuration.

[0039] Optionally, in the dynamic weight configuration of the anomaly identification network based on model self-checking, the self-checking processing layer is first initialized, that is, it is confirmed that the self-checking processing layer is a defined and configured sub-processing layer in the anomaly prediction model, and the processing layer contains the algorithms and logic required to perform data richness evaluation and data adaptation evaluation. Next, load the data set required by the first anomaly identification network, call the data richness evaluation algorithm in the self-checking processing layer to evaluate the data set of the first anomaly identification network. The data richness evaluation includes the diversity, integrity, novelty and other aspects of the data. After the evaluation is completed, the evaluation result is recorded as part of the first self-checking result. Further, load the data set required by the second anomaly identification network, call the data adaptation evaluation algorithm in the self-checking processing layer to evaluate the data set of the second anomaly identification network. The data adaptation evaluation mainly focuses on the matching degree between the data and the network model, such as whether the data distribution is consistent with the model assumption, whether it contains enough abnormal samples, etc. Similarly, after the evaluation is completed, the evaluation result is recorded as part of the second self-checking result. Further, the data richness score, data adaptation score and other key information in the two self-checking results are extracted, and weighted average, maximum and minimum value and other strategies are used to integrate the two results to form a comprehensive self-checking report or score. Based on the integrated self-checking result, the weights of the first anomaly identification network and the second anomaly identification network in the final anomaly prediction model are determined, that is, if the data richness and data adaptation score of a certain network is higher, its weight in the model should be increased accordingly. Dynamic weight configuration can ensure that the model can fully utilize the advantages of each network during runtime and improve prediction performance.

[0040] Further, the anomaly prediction result is evaluated, and a ladder compensation strategy is established based on the anomaly evaluation result. The step S500 of the present application further comprises:

[0041] Step S510: Analyzing the anomaly evaluation result to obtain the maximum abnormal deviation value.

[0042] Step S520: Establishing a compensation extreme value of the ladder compensation strategy based on the maximum abnormal deviation value.

[0043] Step S530: Performing device stability analysis on the underwater special power equipment, and configuring the number of ladders and the length of ladders based on the device stability analysis result.

[0044] Step S540: Completing the establishment of the ladder compensation strategy through the compensation extreme value, the number of ladders and the length of ladders.

[0045] For example, first, the results of the abnormal evaluation (i.e., a set of one or more indicators or metrics) reflecting the performance of the submersible motor under abnormal conditions are analyzed. From these indicators or metrics, a parameter representing the degree of abnormal deviation is identified, and the maximum value thereof, i.e., the maximum abnormal deviation value, is determined, which reflects the maximum performance deviation or the degree of deviation from the expected operating range of the device under abnormal conditions. Further, a compensation extreme value of the step compensation strategy is established based on the maximum abnormal deviation value. The compensation extreme value refers to the maximum range or limit value that needs to be compensated, and in general, the compensation extreme value is set based on the maximum abnormal deviation value to ensure that the compensation strategy can cover the maximum deviation that may occur under abnormal conditions of the device.

[0046] At the same time, the device stability analysis of the submersible motor is performed. The device stability analysis is a process of evaluating the stability and reliability of the device under normal operating conditions and potential abnormal conditions, which includes the examination of the operating history data, performance parameters, fault records, etc. of the device. Through the device stability analysis, the performance of the device under different operating conditions, as well as the sensitivity and recovery ability of the device to abnormal conditions can be understood. Further, the number of steps and the step length are configured based on the results of the device stability analysis. The number of steps determines the degree of precision of the compensation strategy, i.e., how many steps or intervals the abnormal deviation range covered by the compensation strategy is divided into. The step length represents the size of each step or interval, i.e., the difference between adjacent steps. Through the results of the device stability analysis, appropriate number of steps and step length can be determined, for example, if the performance of the device under abnormal conditions is relatively stable, then fewer number of steps and larger step length can be configured; conversely, if the device is sensitive to abnormal conditions, then more number of steps and smaller step length need to be configured.

[0047] Finally, after the compensation extreme value, the number of steps and the step length are determined, a complete step compensation strategy can be established. This strategy is usually a table or function that defines the compensation measures or compensation values to be taken under different abnormal deviation values. The compensation measures or compensation values can be set according to actual needs, such as adjusting the operating parameters of the device, triggering a fault alarm, starting a backup device, etc. The establishment of this step compensation strategy should ensure that appropriate compensation measures can be taken in a timely and accurate manner when the device encounters abnormal conditions, in order to ensure the normal operation of the device and the stability of the system.

[0048] Through the implementation of the above steps, the abnormal conditions of the device can be accurately addressed, and potential device failures or system instability can be avoided; at the same time, this flexibility enables the compensation strategy to adapt to abnormal conditions of different severity, thereby providing more precise and effective compensation measures and improving the control optimization efficiency.

[0049] Further, the verification and identification of the optimization control through the verification window are based on the verification and identification results, and step S600 of the present application further includes:

[0050] Step S610: constructing node deviation based on the verification and identification results, the node deviation being the mapping deviation of the verification window and each step node in the step compensation strategy.

[0051] Step S620: unit deviation analysis through the node deviation for control, generating optimization compensation.

[0052] Step S630: reconstructing the step compensation strategy without compensation according to the node deviation, and continuing the optimization control based on the reconstruction results and the optimization compensation.

[0053] Further, in order to quantitatively evaluate the accuracy and effectiveness of the step compensation strategy in actual application, node deviation is constructed based on the verification and identification results. The node deviation refers to the mapping deviation of the verification window (i.e. actual device running data or performance indicators) and each step node in the step compensation strategy. By comparing the actual data in the verification window with the expected data or state in the step compensation strategy, the deviation value existing at each step node can be determined as a reference for subsequent optimization. Further, unit deviation analysis is performed through the node deviation for control. The unit deviation analysis refers to detailed analysis of the node deviation to determine the specific reasons, influence degree and possible solutions of the deviation. By analyzing the deviation of each unit (e.g. each step node), deficiencies or problems in the compensation strategy can be identified to provide direction for subsequent optimization. Finally, optimization compensation is generated based on the results of the unit deviation analysis. These optimization compensations may be fine-tuning of certain parameters in the step compensation strategy, or introduction of new compensation measures to make up for existing deficiencies. Through targeted optimization measures, the accuracy and effectiveness of the step compensation strategy are improved, further improving the stability of the device and the performance of the system.

[0054] If it is found in the node deviation analysis that there is a large deviation at certain step nodes and it cannot be eliminated by simple optimization compensation, the entire step compensation strategy needs to be reconstructed, and the reconstruction process includes redesigning the number, length and compensation measures of the steps. Through comprehensive reconstruction, the problems existing in the step compensation strategy are fundamentally solved, providing a more stable and reliable control strategy for the device. Finally, after the reconstruction of the step compensation strategy and the generation of the optimization compensation, these optimization measures are applied to the actual control system, and the control effect is continuously monitored and analyzed. If new problems are found or further optimization is needed, the above steps can be repeated for iterative improvement. Thus, a closed-loop optimization process is realized, which can continuously improve the performance of the control system and the stability of the device.

[0055] Further, the application also includes step S700, which further includes:

[0056] Step S710: Extract historical data of underwater special power equipment, and establish a historical data set after denoising.

[0057] Step S720: Configure comprehensive performance evaluation indicators, evaluate the performance of underwater special power equipment through the comprehensive performance evaluation indicators and the historical data set, and generate an optimized control strategy according to the performance evaluation result.

[0058] Step S730: Optimize the task response of underwater special power equipment through the optimized control strategy.

[0059] Optionally, in addition to considering the optimization strategy from the compensation angle, optimization can also be performed from the basic control optimization angle to ensure the accuracy of the optimization. Specifically, historical data is extracted from the operation records of the submersible motor, the monitoring system or other data sources, and these data include the operating parameters, performance indicators, fault records and the like of the equipment, and the extracted original historical data is denoised to eliminate noise and outliers in the data to obtain a final historical data set. Next, the performance of the submersible motor is evaluated. Before evaluating the performance of the equipment, a series of comprehensive performance evaluation indicators are configured, which can comprehensively reflect the performance of the equipment in various aspects, such as operating efficiency, stability, reliability, energy consumption, etc. Further, the performance of the submersible motor is evaluated using the configured comprehensive performance evaluation indicators and the denoised historical data set, and an evaluation result is generated. For example, the scores of various indicators of the submersible motor are calculated, performance curves are drawn, statistical analysis is performed, etc., and the performance evaluation result can objectively reflect the performance of the equipment, providing an important basis for formulating an optimized control strategy.

[0060] Further, according to the performance evaluation result of the equipment, the problems and deficiencies in the performance of the equipment are analyzed, and then corresponding optimized control strategies are formulated, which include adjusting the operating parameters of the equipment, improving the control algorithm, increasing maintenance measures, etc. The optimized control strategy can propose solutions to the specific problems of the equipment, which helps to improve the performance and stability of the equipment and prolong the service life of the equipment. Finally, the generated optimized control strategy is applied to the actual operation of the submersible motor, and the task response of the equipment is optimized, including improving the response speed of the task, reducing the execution time of the task, improving the completion quality of the task, etc. Task response optimization can improve the working efficiency and reliability of the equipment, so that the equipment can better meet the needs of actual work; at the same time, it can also reduce the operating cost and failure rate of the equipment, and improve the overall performance of the equipment.

[0061] Further, the application step S720 further includes:

[0062] The efficiency index is established, and the formula is as follows:

[0063] .

[0064] wherein, is the efficiency index, is the power efficiency, is the power stability, and are weight factors of the power efficiency and the power stability, respectively.

[0065] The reliability index is established, and the formula is as follows:

[0066] .

[0067] wherein, is the reliability index, represents the failure rate, represents the failure-free operation time, and are weight factors of the failure rate and the failure-free operation time, respectively.

[0068] The adaptability index is established, and the formula is as follows:

[0069] .

[0070] wherein, represents the adaptability index, is the adaptability of the underwater special power equipment to the underwater environment, is the adaptability of the underwater special power equipment to the load change, and are adaptability weight factors of the underwater environment and the load change.

[0071] The efficiency index, the reliability index and the adaptability index are integrated to establish a comprehensive performance evaluation index.

[0072] Specifically, the comprehensive performance evaluation index proposed in the embodiments of the present application includes the efficiency index, the reliability index and the adaptability index. The calculation formula of the efficiency index is ; wherein, is the efficiency index, is the power efficiency, is the power stability, and are weight factors of the power efficiency and the power stability, respectively. The calculation formula of the reliability index is ; wherein, is the reliability index, represents the failure rate, characterizing the failure-free operation time, and respectively are the weight factors of the failure rate and the failure-free operation time. The formula for calculating the adaptability index is ; wherein, characterizes the adaptability index, is the adaptability of the underwater special power equipment to the underwater environment, is the adaptability of the underwater special power equipment to the load change, and are the adaptability weight factors of the underwater environment and the load change. Finally, the efficiency index, the reliability index, and the adaptability index are integrated to establish a comprehensive performance evaluation index.

[0073] By constructing the comprehensive performance evaluation index including the efficiency index, the reliability index, and the adaptability index, the performance of the submersible motor can be comprehensively and objectively evaluated, thereby guiding the corresponding optimization decision, improving the operation efficiency, stability, and adaptability of the equipment, and further reducing the operation cost and risk.

[0074] Further, the step S400 of preprocessing the time series monitoring data set further comprises:

[0075] Step S410: performing correlation analysis on the time series monitoring data set to construct independent data identifiers.

[0076] Step S420: regarding the data corresponding to the independent data identifiers as sensitive data, performing global similarity verification within the time series monitoring data set.

[0077] Step S430: performing corresponding data elimination according to the global similarity verification result, and completing the preprocessing of the time series monitoring data set according to the elimination result.

[0078] Specifically, before preprocessing the time series monitoring data set, correlation analysis needs to be performed on the data in the time series monitoring data set. Correlation analysis aims to find out whether there is some kind of relationship or pattern between data items. In the time series data set, such correlation may manifest as temporal correlation, value trend, or cross-reference with other data sources, etc. Specifically, statistical methods (such as correlation analysis, regression analysis), data mining techniques (such as clustering analysis, association rule mining), or machine learning algorithms (such as deep learning models) can be used for correlation analysis. On the basis of correlation analysis, data items with unique characteristics or attributes are identified, and independent data identifiers are constructed for them. These independent data identifiers can be unique identifiers (such as ID numbers) of the data, specific time series patterns, or abnormal values above a certain threshold, etc. The construction of independent data identifiers helps the subsequent identification and management of sensitive data.

[0079] Further, the data with independent data identification is taken as sensitive data, and global similarity verification in the time series monitoring data set is performed. The global similarity verification aims to check whether there are data items similar or related to the sensitive data in the entire data set. The similarity verification process can be realized by calculating the similarity (such as cosine similarity, Euclidean distance, etc.) between data items or using a machine learning algorithm (such as K- nearest neighbor algorithm). The result of the global similarity verification will determine which data items are similar to the sensitive data enough to be treated as sensitive data. According to the result of the global similarity verification, data items similar or related to the sensitive data are identified, and according to the preset threshold or rule, it is determined which data items need to be removed or processed; finally, the data removal operation is performed to remove these data items from the data set or perform anonymization, encryption, etc. The preprocessed data set will no longer contain sensitive data or data items highly related to sensitive data, thereby reducing the risk of data leakage; at the same time, the preprocessed data set is also more suitable for subsequent data analysis, mining, etc.

[0080] Through the technical solutions of the above embodiments, the double-identification network anomaly prediction submersion motor compensation control method provided by the present application solves the technical problems of low equipment operation efficiency, poor stability and adaptability, inability to timely and accurately predict and optimize anomalies, and high failure rate and maintenance cost in the process of submersion motor control optimization, achieves accurate anomaly prediction and optimization control of the submersion motor, improves the operation efficiency and safety of the equipment, and reduces the failure rate and maintenance cost. Embodiment

[0081] Based on the same inventive concept as the double-identification network anomaly prediction submersion motor compensation control method in the foregoing embodiments, as shown in Figure 2 The present application provides a double-identification network anomaly prediction submersion motor compensation control system, which comprises:

[0082] A model building module 11 is configured to build an anomaly prediction model, wherein the anomaly prediction model is a model for anomaly prediction of underwater special power equipment, and the anomaly prediction model comprises a first anomaly identification network and a second anomaly identification network, the first anomaly identification network is constructed by using a data set of the underwater special power equipment, and the second anomaly identification network is constructed by using big data.

[0083] A device monitoring module 12 is configured to perform device monitoring of the underwater special power equipment and build a time series monitoring data set.

[0084] A model self-checking module 13 is configured to perform model self-checking of the anomaly prediction model and complete dynamic weight configuration of the first anomaly identification network and the second anomaly identification network according to the model self-checking result.

[0085] An anomaly prediction module 14 is configured to input the preprocessed time sequence monitoring data set into the anomaly prediction model with dynamic weight configuration to generate an anomaly prediction result.

[0086] A strategy compensation module 15 is configured to perform anomaly evaluation on the anomaly prediction result, establish a ladder compensation strategy based on the anomaly evaluation result, perform optimization control through the ladder compensation strategy, and set a verification window.

[0087] A verification identification module 16 is configured to perform verification identification of the optimization control through the verification window, and complete the optimization control based on the verification identification result.

[0088] Further, the model self-checking module 13 is further configured to perform the following steps:

[0089] A self-checking processing layer is called to perform data richness evaluation of the first anomaly identification network, and a first self-checking result is constructed according to the data richness evaluation result, wherein the self-checking processing layer is a sub-processing layer of the anomaly prediction model.

[0090] The self-checking processing layer is called to perform data adaptation evaluation of the second anomaly network, a second self-checking result is constructed according to the adaptation evaluation result, and the first self-checking result and the second self-checking result are integrated to complete dynamic weight configuration.

[0091] Further, the strategy compensation module 15 is further configured to perform the following steps:

[0092] The anomaly evaluation result is analyzed to obtain a maximum anomaly deviation value.

[0093] A compensation extreme value of the ladder compensation strategy is established based on the maximum anomaly deviation value.

[0094] The underwater special-purpose power equipment is subjected to equipment stability analysis, and the number of ladders and the length of ladders are configured based on the equipment stability analysis result.

[0095] The ladder compensation strategy is established through the compensation extreme value, the number of ladders, and the length of ladders.

[0096] Further, the verification identification module 16 is further configured to perform the following steps:

[0097] A node deviation is constructed based on the verification identification result, wherein the node deviation is a mapping deviation of each ladder node in the verification window and the ladder compensation strategy.

[0098] Unit deviation analysis of the control through the node deviation is performed to generate optimization compensation.

[0099] The ladder compensation strategy that is not compensated is reconstructed according to the node deviation, and the optimization control is continued based on the reconstruction result and the optimization compensation.

[0100] Further, the application also includes a performance evaluation module, which is further configured to perform the following steps:

[0101] Historical data of the underwater special power equipment is extracted, and a historical data set after denoising is established.

[0102] An integrated performance evaluation index is configured, and performance evaluation of the underwater special power equipment is performed through the integrated performance evaluation index and the historical data set. An optimized control strategy is generated according to the performance evaluation result.

[0103] Task response optimization of the underwater special power equipment is performed through the optimized control strategy.

[0104] Further, the performance evaluation module is further configured to perform the following steps:

[0105] An efficiency index is established, and the formula is as follows:

[0106] .

[0107] Wherein, is the efficiency index, is the power efficiency, is the power stability, and are weight factors of the power efficiency and the power stability, respectively.

[0108] A reliability index is established, and the formula is as follows:

[0109] .

[0110] Wherein, is the reliability index, represents the failure rate, represents the failure-free operation time, and are weight factors of the failure rate and the failure-free operation time, respectively.

[0111] An adaptability index is established, and the formula is as follows:

[0112] .

[0113] Wherein, represents the adaptability index, is the adaptability of the underwater special power equipment to the underwater environment, is the adaptability of the underwater special power equipment to load changes, and are adaptability weight factors of the underwater environment and the load changes.

[0114] Integrate the efficiency index, reliability index, adaptability index, and establish a comprehensive performance evaluation index.

[0115] Further, the anomaly prediction module 14 is further configured to perform the following steps:

[0116] Performing correlation analysis on the time series monitoring data set, and constructing independent data identification.

[0117] Taking the data corresponding to the independent data identification as sensitive data, performing global similarity verification in the time series monitoring data set.

[0118] According to the global similarity verification result, the corresponding data is removed, and the time series monitoring data set preprocessing is completed according to the removal result.

[0119] Through the foregoing detailed description of the double-identification network anomaly prediction submersion motor compensation control method, those skilled in the art can clearly understand the double-identification network anomaly prediction submersion motor compensation control system in the embodiments. For the device disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant part is referred to the method part description.

[0120] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A motor compensation control method for undersea drilling using dual-identification network anomaly prediction, characterized in that, The method includes: An anomaly prediction model is established, which is a model for predicting anomalies in underwater dedicated power equipment. The anomaly prediction model includes a first anomaly identification network and a second anomaly identification network. The first anomaly identification network is constructed using the usage dataset of underwater dedicated power equipment, and the second anomaly identification network is constructed using big data. Perform equipment monitoring of the underwater dedicated power equipment and establish a time-series monitoring dataset; Perform a model self-check of the anomaly prediction model, and complete the dynamic weight configuration of the first anomaly identification network and the second anomaly identification network based on the model self-check results; After preprocessing the time-series monitoring dataset, it is input into the anomaly prediction model with dynamic weight configuration to generate anomaly prediction results. Anomaly evaluation is performed on the anomaly prediction results, a tiered compensation strategy is established based on the anomaly evaluation results, optimization control is performed through the tiered compensation strategy, and a verification window is set. The verification window is used to perform verification and identification for optimization control, and optimization control is completed based on the verification and identification results. The abnormal prediction results are evaluated for anomalies, and a tiered compensation strategy is established based on the anomaly evaluation results. The method further includes: Analyze the anomaly evaluation results to obtain the maximum anomaly deviation value; The compensation extreme value of the step compensation strategy is established based on the maximum abnormal deviation value; A stability analysis was performed on the underwater dedicated power equipment, and the number and length of steps were configured based on the stability analysis results. The step compensation strategy is established by defining the compensation extreme value, the number of steps, and the step length.

2. The motor compensation control method for undersea drilling using dual-identification network anomaly prediction as described in claim 1, characterized in that, The step of performing a model self-check on the anomaly prediction model, and configuring the dynamic weights of the first and second anomaly detection networks based on the model self-check results, further includes: The self-checking processing layer is invoked to perform data richness evaluation of the first anomaly identification network, and the first self-checking result is constructed based on the data richness evaluation result. The self-checking processing layer is a sub-processing layer of the anomaly prediction model. The self-check processing layer is invoked to perform data adaptation evaluation of the second abnormal network. Based on the adaptation evaluation results, a second self-check result is constructed. The first and second self-check results are integrated to complete the dynamic weight configuration.

3. The motor compensation control method for undersea drilling using dual-identification network anomaly prediction as described in claim 1, characterized in that, The verification and identification process for optimization control via the verification window, and the completion of optimization control based on the verification and identification results, further includes: Based on the verification and identification results, a node deviation is constructed, which is the mapping deviation between the verification window and each step node in the step compensation strategy. Unit deviation analysis, which controls the node deviation, generates optimized compensation. Based on the node deviation, the uncompensated step compensation strategy is reconstructed, and the control is further optimized based on the reconstruction results and optimized compensation.

4. The motor compensation control method for undersea drilling using dual-identification network anomaly prediction as described in claim 1, characterized in that, The method further includes: Historical data of underwater specialized power equipment was extracted to establish a denoised historical dataset; Configure comprehensive performance evaluation indicators, evaluate the performance of underwater dedicated power equipment using the comprehensive performance evaluation indicators and the historical dataset, and generate an optimized control strategy based on the performance evaluation results. The optimized control strategy is used to optimize the task response of underwater dedicated power equipment.

5. The motor compensation control method for undersea drilling using dual-identification network anomaly prediction as described in claim 4, characterized in that, The overall performance evaluation indicators for the configuration also include: The efficiency index is established using the following formula: ； in, As an efficiency indicator, For power efficiency, For power stability, and These are the weighting factors for power efficiency and power stability, respectively. The reliability index is established using the following formula: ； in, As a reliability indicator, Characterizing the failure rate, Characterizing fault-free uptime, and These are the weighting factors for failure rate and fault-free uptime, respectively. Establish an adaptation index, with the following formula: ； in, Characteristic of adaptability To enhance the adaptability of underwater-specific power equipment to the underwater environment, To enhance the adaptability of underwater specialized power equipment to load changes, and For adaptive weighting factors of underwater environment and load changes; By integrating the aforementioned efficiency, reliability, and adaptability indicators, a comprehensive performance evaluation index is established.

6. The motor compensation control method for undersea drilling using dual-identification network anomaly prediction as described in claim 1, characterized in that, The preprocessing of the time-series monitoring dataset further includes: Perform data correlation analysis on the time-series monitoring dataset to construct independent data identifiers; The data corresponding to the independent data identifier is used as sensitive data, and global similarity verification is performed within the time-series monitoring dataset. Based on the global similarity verification results, corresponding data are removed, and the time series monitoring dataset is preprocessed based on the removal results.

7. A dual-identification network anomaly prediction and submersion motor compensation control system, characterized in that, The system for implementing the method according to any one of claims 1-6 comprises: The model building module is used to build an anomaly prediction model. The anomaly prediction model is a model for predicting anomalies in underwater special power equipment. The anomaly prediction model includes a first anomaly identification network and a second anomaly identification network. The first anomaly identification network is built using the usage dataset of underwater special power equipment, and the second anomaly identification network is built using big data. The equipment monitoring module is used to perform equipment monitoring of the underwater dedicated power equipment and establish a time-series monitoring dataset; The model self-checking module is used to perform a model self-check of the anomaly prediction model and to complete the dynamic weight configuration of the first anomaly identification network and the second anomaly identification network based on the model self-checking results. The anomaly prediction module is used to preprocess the time-series monitoring dataset and input it into the anomaly prediction model with dynamically configured weights to generate anomaly prediction results. The strategy compensation module is used to evaluate the anomaly prediction results, establish a tiered compensation strategy based on the anomaly evaluation results, optimize and control the system through the tiered compensation strategy, and set a verification window. The verification and identification module is used for verification and identification through the verification window to optimize control, and to complete the optimization control based on the verification and identification results.

Images