A fault diagnosis method and system for large-scale phase regulator

Abstract

The present invention discloses a fault diagnosis method and system for a large-scale phase regulator, which relates to the field of phase regulator fault diagnosis technology, including establishing an Internet of Things monitoring network, monitoring key performance data, and uploading the data to a cloud platform in real time; pre-processing and extracting features based on the cloud platform combined with mechanical, electrical, and thermal sensing sensor data, using a deep learning algorithm to establish a multi-modal fault prediction model, and comprehensively predicting potential faults and diagnosing the source of the fault based on sensor information. The method of the present invention lays a solid foundation for subsequent deep learning analysis by establishing an Internet of Things monitoring network, combining cloud platform processing and use, and through a multi-modal fault prediction model, integrating key sensor data collected from the cloud platform, it can effectively predict and diagnose potential sources of faults, and evaluate the abnormal contribution of each sensor data at a specific time point, and output the possibility of the source of the fault accordingly.

Description

Technical Field

[0001] The invention relates to the technical field of phase regulator fault diagnosis, and in particular to a fault diagnosis method and system for a large-scale phase regulator. Background Art

[0002] Large phase regulators are usually called large phase adjusters or phase regulators. They are a type of equipment used in power systems. They are mainly used to adjust the power factor of transmission lines, optimize the voltage quality of the power grid and reduce energy losses. This type of equipment helps the power system maintain stable and efficient operation by adjusting the phase angle between current and voltage.

[0003] Traditional large-scale phase regulator failures are mostly divided into mechanical failures, electrical failures and thermal failures. Thermal failures are further divided into overheating caused by overload and poor heat dissipation. Different failures involve multiple components in large-scale phase regulators. Although multi-dimensional performance data can be obtained through monitoring equipment, the processing and analysis of these data are often not in-depth and systematic enough, and lack effective data fusion and intelligent processing. This also leads to the fact that when the operating efficiency of large-scale phase regulators is reduced or failures occur, maintenance personnel need to spend a lot of time to investigate before they can diagnose the source of the fault, which greatly increases maintenance costs and affects the efficiency of the use of large-scale phase regulators. Summary of the invention

[0004] In view of the problems existing in the above-mentioned existing fault diagnosis method and system for large-scale phase regulators, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is that although multi-dimensional performance data can be obtained through monitoring equipment, the processing and analysis of these data are often not in-depth and systematic enough, and lack effective data fusion and intelligent processing. This also leads to the fact that when the operating efficiency of the large-scale phase regulator is reduced or a failure occurs, maintenance personnel need to spend a lot of time to investigate before they can diagnose the source of the fault, which greatly increases the maintenance cost and affects the utilization efficiency of the large-scale phase regulator.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a fault diagnosis method for a large-scale phase regulator, comprising:

[0007] Establish an IoT monitoring network to monitor key performance data and upload key performance data to the cloud platform in real time;

[0008] Preprocess key performance data and extract features based on the cloud platform, use deep learning algorithms to build a multimodal fault prediction model, predict potential faults based on sensor data and model-predicted fault indexes, and diagnose fault sources;

[0009] Collect data based on the operating temperature of the condenser, predict the operating temperature data of the condenser through trend analysis, and execute corresponding maintenance strategies based on the predicted value of the operating temperature of the condenser;

[0010] Use the visualization platform to display the collected data of the cloud platform, the predicted data of the fault prediction model, and the operating temperature prediction data of the phase regulator.

[0011] As a preferred solution of the fault diagnosis method for large-scale phase regulators of the present invention, the method of establishing an Internet of Things monitoring network, monitoring key performance data, and uploading the key performance data to a cloud platform in real time includes:

[0012] Install sensors at key locations of the phase regulator to obtain monitoring data and set the monitoring data as key performance data;

[0013] Connect each sensor to local data storage for collection backup;

[0014] Using secure wireless communication protocols, the collected data is transmitted to the cloud platform in real time.

[0015] As a preferred solution of the fault diagnosis method for large-scale phase shifter of the present invention, the preprocessing and feature extraction of key performance data based on the cloud platform includes:

[0016] Acquire the detection data of accelerometers, Hall effect sensors and thermocouples based on the key performance data of the cloud platform, and perform data denoising and normalization;

[0017] The vibration data is converted to frequency domain representation by fast Fourier transform, and the amplitude of the main frequency and the corresponding frequency value are extracted as the extracted features. A fixed time window is set for the monitoring data of the Hall effect sensor and thermocouple, and the mean, standard deviation and peak statistics in each time window are calculated and used as the characteristic data of the corresponding sensor.

[0018] As a preferred solution of the fault diagnosis method for large-scale phase-shifting machine of the present invention, the multi-modal fault prediction model is established by using a deep learning algorithm, and potential faults are predicted and the fault source is diagnosed based on the fault index predicted by the sensor data and the model, including:

[0019] Construct a fault prediction model, including input layer, LSTM layer, diagnosis layer and output layer;

[0020] When the loss of the validation set data decreases too low in consecutive iterations, the iteration is stopped and the corresponding θ is obtained. k 、w k and w j parameter;

[0021] Substitute the validation set data into the fault prediction model, divide the validation set data into fault results and normal operation results, and set the index threshold according to the validation output of the fault prediction model;

[0022] The sensor feature data collected in real time is input into the fault prediction model to calculate the fault index F(t). When F(t) is greater than or equal to the index threshold, it is judged to be abnormal, and the fault contribution C of each sensor is used to predict the fault. t,k , and the sensor with the largest contribution is the fault source.

[0023] As a preferred solution of the fault diagnosis method for large-scale phase regulators of the present invention, wherein: the data collection according to the phase regulator operating temperature and the phase regulator operating temperature data prediction by trend analysis include:

[0024] Collect the internal temperature data of the condenser in real time during operation, and perform data cleaning and normalization on the collected temperature data;

[0025] Use the temperature training set to train the temperature trend analysis model and calculate the optimal parameter value by minimizing the prediction error;

[0026] The temperature parameters β0, β1 and β2 are calculated by gradient descent method respectively, and updated according to the gradient and learning rate in each iteration;

[0027] Repeat the update steps until the calculated value of the loss function decreases significantly, then stop iterating;

[0028] The collected new temperature data is substituted into the model to calculate the predicted operating temperature. The operating temperature threshold is set based on the operation manual. When the predicted operating temperature is outside the operating temperature threshold range, it is determined to be a temperature abnormality.

[0029] As a preferred solution of the fault diagnosis method for large-scale phase regulator of the present invention, the corresponding maintenance strategy is executed according to the predicted value of the phase regulator operating temperature, including:

[0030] When the operating temperature is determined to be abnormal, if the predicted value is lower than the temperature threshold range, the heating device is started; if the predicted value is higher than the temperature threshold range, the cooling device is started;

[0031] Based on the determination of temperature anomalies, an alarm is sent to the operator to remind the operator to take maintenance measures.

[0032] As a preferred solution of the fault diagnosis method for large-scale phase shifters of the present invention, the method of using a visualization platform to display the collected data of the cloud platform, the predicted data of the fault prediction model and the operating temperature prediction data of the phase shifters includes:

[0033] Select a visualization platform and obtain the cloud platform collection data, the fault prediction model prediction data, and the temperature trend prediction data for data visualization:

[0034] Displays real-time readings of current operating temperatures, as well as predicted values;

[0035] Show trend analysis of various sensor data and abnormal patterns that may indicate failure;

[0036] Lists recent historical fault events and corresponding alarms, including fault type, occurrence time, handling measures, and status.

[0037] Another object of the present invention is to provide a system for a fault diagnosis method of a large-scale phase regulator, which comprises:

[0038] Data acquisition module, which collects real-time data from various sensors and uses IoT devices to send the collected data to the cloud platform in real time;

[0039] The cloud platform data processing module cleans, normalizes and extracts features of the received data in the cloud;

[0040] The fault prediction module uses deep learning technology to analyze data, establish a fault prediction model, predict potential faults and diagnose the source of the fault;

[0041] Temperature trend prediction module, which analyzes the time series of temperature data, predicts future temperature changes, and adjusts maintenance strategies based on the predictions;

[0042] Visualization interface module, which displays real-time data, forecast results, system status and provides graphical representation of historical data and trend analysis.

[0043] A computer device comprises: a memory and a processor; the memory stores a computer program, and the processor implements the steps of the above-mentioned large-scale phase regulator fault diagnosis method when executing the computer program.

[0044] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned large-scale phase regulator fault diagnosis method.

[0045] The beneficial effects of the present invention are as follows: by establishing an Internet of Things monitoring network, combined with cloud platform processing and use, a solid foundation is laid for subsequent deep learning analysis. Through a multimodal fault prediction model, key sensor data collected from the cloud platform are integrated, and potential fault sources can be effectively predicted and diagnosed. The abnormal contribution of each sensor data at a specific time point is evaluated, and the possibility of the fault source is output accordingly. The information of the comprehensive diagnostic layer is used to reflect whether the current system status is normal or not, so that maintenance personnel can quickly judge the system status and analyze the fault source, improve maintenance efficiency, and predict temperature changes through trend analysis technology. Trend analysis is performed on abnormalities caused by poor heat dissipation during phase adjustment operation. By implementing a maintenance strategy based on predicted temperature data, the maintenance and operation efficiency of the equipment is further optimized. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0047] Figure 1 The figure is a flow chart of the fault diagnosis method for large phase regulator.

[0048] Figure 2 This is a schematic diagram of the structure of the fault diagnosis system for large-scale phase regulators. DETAILED DESCRIPTION

[0049] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0050] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0051] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0052] Example 1

[0053] Reference Figure 1, which is the first embodiment of the present invention, and which provides a fault diagnosis method for a large-scale phase regulator. The fault diagnosis method for a large-scale phase regulator includes:

[0054] S1, establish an IoT monitoring network, monitor key performance data, and upload key performance data to the cloud platform in real time;

[0055] Preferably, an IoT monitoring network is established to monitor key performance data and upload the key performance data to a cloud platform in real time, including:

[0056] Install sensors at key locations of the phase regulator to obtain monitoring data and set the monitoring data as key performance data;

[0057] Connect each sensor to local data storage for collection backup;

[0058] Using secure wireless communication protocols, the low-power wireless communication protocol Zigbee can be used, and AES-128 encryption is adopted to ensure the security of data transmission. The collected data is transmitted to the cloud platform in real time based on the low-power wide area network technology of the cellular network.

[0059] By establishing an IoT monitoring network, real-time monitoring of key performance data of large-scale phase regulators can be achieved. This step uses various sensors to collect real-time data from equipment and uploads the data to the cloud platform through IoT technology. This type of monitoring not only improves the speed and efficiency of data collection, but also ensures the accuracy and timeliness of the data. Using the cloud platform for data storage and processing can significantly improve data processing capabilities and achieve cost-efficiency in big data analysis and storage. Uploading data to the cloud platform in real time allows data to be accessed remotely, supports remote monitoring and management, and provides a basis for fault diagnosis and maintenance. In addition, the establishment of the IoT monitoring network lays a data foundation for subsequent fault prediction and the implementation of maintenance strategies, thereby effectively preventing and reducing the occurrence of faults and improving the reliability and safety of equipment operation.

[0060] S2, preprocesses key performance data and extracts features based on the cloud platform, uses deep learning algorithms to establish a multimodal fault prediction model, predicts potential faults based on sensor data and model-predicted fault indexes, and diagnoses fault sources;

[0061] Preferably, the key performance data is preprocessed and features are extracted based on the cloud platform, including:

[0062] Acquire key performance data from accelerometers, Hall effect sensors, and thermocouples based on the cloud platform, and perform data denoising and normalization. Accelerometers are used to capture vibration data, Hall effect sensors are used to monitor current in the circuit, and thermocouples are used to measure the temperature of key components such as motors and bearings.

[0063] The vibration data is converted to frequency domain representation by fast Fourier transform, and the amplitude of the main frequency and the corresponding frequency value are extracted as the extracted features. A fixed time window is set for the monitoring data of the Hall effect sensor and the thermocouple. The fixed time window refers to analyzing and processing data within a fixed time interval, allowing data within a specific time period to be extracted from a continuous data stream for statistical analysis or feature extraction. The mean, standard deviation and peak statistics within each time window are calculated and used as the characteristic data of the corresponding sensor.

[0064] The key sensor data of large-scale phase regulators are collected through the cloud platform. Using the powerful processing power of cloud computing, this step denoises and normalizes the sensor data to ensure the quality and consistency of the input data, laying a solid foundation for subsequent deep learning analysis. The vibration data collected by the accelerometer is converted to the frequency domain through the fast Fourier transform (FFT). This transformation can not only reveal the main frequency components in the vibration data, but also extract key features reflecting the mechanical state, such as the amplitude of the main frequency and the corresponding frequency value. These features are crucial for identifying faults caused by mechanical wear, imbalance or alignment problems. Hall effect sensors The data provided by the device is mainly used to monitor changes in the electromagnetic field, which is very effective in detecting electrical faults such as open circuits and short circuits. By analyzing these data in fixed time windows and calculating the average, standard deviation and peak value in each window, it is possible to effectively capture slight changes in the state of the electrical system, thereby discovering potential electrical problems in advance. The temperature monitoring data of the thermocouple is used to analyze whether the phase regulator is overheated and determine whether it is overloaded. Through this step, the predictability and accuracy of equipment operation and maintenance can be greatly improved, and the downtime and maintenance costs caused by faults can be significantly reduced, ensuring that the phase regulator operates in the optimal state and improving the stability and efficiency of the entire power system.

[0065] Furthermore, a multimodal fault prediction model is established using a deep learning algorithm to predict potential faults and diagnose the source of faults based on sensor data and the fault index predicted by the model, including:

[0066] Construct a fault prediction model, including input layer, LSTM layer, diagnosis layer and output layer;

[0067] The sensor feature data is input into the LSTM layer and expressed as:

[0068] h t,k =LSTM(x t,k ,h t-1,k ,θ k )

[0069] where x t,k represents the input data of the kth sensor at time t, h t-1,krepresents the output data of the previous time step, h t,k represents the output data of the kth sensor at time t, θ k Represents all weights and biases of the LSTM network;

[0070] Diagnostic layer evaluation output data h t,k The abnormal contribution of the output fault source is expressed as:

[0071]

[0072] Among them C t,k represents the fault contribution of the kth sensor at time t, w k represents the weight of the kth sensor, w j represents the weight of the jth sensor, h t,k represents the LSTM layer output of the kth sensor at time t, h t,j represents the LSTM layer output of the jth sensor at time t, and K represents the total number of sensors;

[0073] The information of the comprehensive diagnosis layer is used to output the fault index, which is expressed as:

[0074]

[0075] Where F(t) represents the fault index at time t, sig represents the activation function sigmoid;

[0076] Initialize θ of the LSTM layer k and the diagnostic layer w k With w j , for each training sample, the LSTM layer uses the current θ k Processing input feature data x t,k Generate h t,k , using the current w k With w j and h t,k Calculate the fault contribution C of each sensor t,k , complete the forward propagation;

[0077] The cross entropy loss function is used to calculate the difference between the model output and the actual label, expressed as:

[0078]

[0079] Where N represents the total number of training samples, y i,k represents the true label of the i-th sample for the k-th sensor, p i,k represents the probability that the prediction result of the i-th sample belongs to the k-th sensor, and L represents the loss function value;

[0080] According to the calculation of the loss function, the parameter θ is calculated by the back propagation algorithm k 、w k and w j The gradient of , using the gradient descent algorithm to update the model parameters to minimize the loss function, is expressed as:

[0081]

[0082]

[0083] Where η is the learning rate, represents the updated θ k parameter, represents the updated w k parameter, represents the updated w j parameter;

[0084] When the loss of the validation set data decreases too low in consecutive iterations, the iteration is stopped and the corresponding θ is obtained. k 、w k and w j parameter;

[0085] Substitute the validation set data into the fault prediction model, divide the validation set data into fault results and normal operation results, and set the index threshold according to the validation output of the fault prediction model. If the output data of F(t) of the normal operation validation set data are all lower than 0.3, and the output data of F(t) of the fault result validation set data are all higher than 0.5, then the threshold can be set to 0.4.

[0086] The sensor feature data collected in real time is input into the fault prediction model to calculate the fault index F(t). When F(t) is greater than or equal to the index threshold, it is judged to be abnormal, and the fault contribution C of each sensor is used to predict the fault. t,k , and the sensor with the largest contribution is the fault source.

[0087] Through deep learning algorithms, a multimodal fault prediction model is established. This model integrates data from mechanical, electrical and thermal sensing sensors, and can effectively predict and diagnose potential fault sources. The application of deep learning algorithms improves the accuracy and efficiency of fault diagnosis. In the LSTM layer, the feature data extracted by the sensor from step 2 is input. The characteristic of the LSTM network is that it can process and memorize the time series information of the input data, which is particularly important for dynamic and complex system fault pattern recognition. Through time series analysis, LSTM can capture early signals of fault development, even if these signals are very weak or delayed in time. The design of the diagnosis layer enables the model to evaluate the abnormal contribution of each sensor data at a specific time point, and output the possibility of the fault source accordingly. The calculation of this layer is based on the output of the LSTM layer, by comparing the differences in output data from different sensors. The model accurately locates the location and cause of the fault by integrating the information of the diagnostic layer and outputs a fault index. This index is calculated by the sigmoid activation function and reflects whether the current system status is normal or not. By setting a reasonable fault index threshold, when the index monitored by the system in real time exceeds this threshold, it can be judged as an abnormal state and corresponding maintenance measures need to be taken. The training process of the entire model is optimized using the cross entropy loss function to ensure the accuracy of the prediction. The model parameters are continuously adjusted through the back propagation algorithm. The model can gradually improve its fault diagnosis accuracy and robustness during the training process, which facilitates and quickly determines the specific source of the abnormality, allowing the maintenance team to perform necessary repairs or adjustments. This not only reduces the downtime of the equipment, but also helps to avoid a wider range of equipment damage caused by fault expansion, reducing maintenance costs and potential safety risks.

[0088] S3, collect data based on the operating temperature of the condenser, predict the operating temperature data of the condenser through trend analysis, and execute corresponding maintenance strategies according to the predicted value of the operating temperature of the condenser;

[0089] Preferably, data collection is performed based on the phase regulator operating temperature, and the phase regulator operating temperature data is predicted through trend analysis, including:

[0090] Collect the internal temperature data of the condenser in real time during operation, and perform data cleaning and normalization on the collected temperature data;

[0091] A temperature trend analysis model is constructed to predict the operating temperature of large phase condensers, which can be expressed as:

[0092]

[0093] Where ΔT t Indicates the temperature difference between the current moment and the previous moment. Indicates the predicted temperature at the next moment, T trepresents the temperature at the current moment, β0 represents the basic coefficient of the model, representing the baseline temperature prediction when there is no temperature change, β1 represents the prediction coefficient of the model, representing the direct impact of the current temperature on the future temperature prediction, and β2 represents the influence coefficient of the model change rate, reflecting the impact of the temperature change trend on the future temperature;

[0094] The temperature training set is used to train the temperature trend analysis model. The optimal parameter value is calculated by minimizing the prediction error. The error function is expressed as:

[0095]

[0096] Where n represents the total number of data in the temperature training set, T t+1 They represent the measured values ​​of the temperature at the next moment in the training set respectively;

[0097] The temperature parameters β0, β1 and β2 are calculated by gradient descent method respectively, and updated according to the gradient and learning rate in each iteration, which can be expressed as:

[0098]

[0099] Where j represents different parameters and a represents the learning rate;

[0100] Repeat the update steps until the calculated value of the loss function decreases significantly, then stop iterating;

[0101] The collected new temperature data is substituted into the model to calculate the predicted operating temperature. The operating temperature threshold is set based on the operation manual. When the predicted operating temperature is outside the operating temperature threshold range, it is determined to be a temperature abnormality.

[0102] By monitoring the internal temperature of the phase regulator during operation and collecting data, and using trend analysis technology to predict future temperature changes, trend analysis can be performed on abnormalities caused by poor heat dissipation during phase regulator operation. The trend of abnormal temperature rise can be discovered in time, so that measures can be taken before the problem develops to a serious stage. This method makes the maintenance strategy more proactive, reduces equipment failures and performance degradation caused by heat dissipation temperature problems, ensures that the equipment operates in the best condition, extends equipment service life, and reduces energy waste.

[0103] Furthermore, a corresponding maintenance strategy is executed according to the predicted value of the phase-converter operating temperature, including:

[0104] When the operating temperature is determined to be abnormal, if the predicted value is lower than the temperature threshold range, the heating device is started; if the predicted value is higher than the temperature threshold range, the cooling device is started;

[0105] Based on the determination of temperature anomalies, an alarm is sent to the operator to remind the operator to take maintenance measures.

[0106] By implementing a maintenance strategy based on predicted temperature data, the maintenance and operating efficiency of the equipment can be further optimized. According to the predicted value of the operating temperature, the cooling system can be scheduled for maintenance or optimization in advance, the workload can be adjusted to avoid overload, or the cooling system can be adjusted to cope with high temperature environments. This predictive maintenance strategy not only improves the reliability and efficiency of the equipment, but also significantly reduces the maintenance and downtime costs caused by failures or unsuitable working environments. Compared with the abnormalities caused by overload, which require maintenance personnel to investigate, the abnormal internal operating temperature can be quickly resolved through the cooperation of the cooling device to achieve automatic cooling and improve the heat dissipation effect, thereby reducing the workload of maintenance personnel.

[0107] S4, using the visualization platform to display the collected data of the cloud platform, the predicted data of the fault prediction model, and the operating temperature prediction data of the phase regulator;

[0108] Preferably, a visualization platform is used to display the collected data of the cloud platform, the predicted data of the fault prediction model and the operating temperature prediction data of the phase regulator, including:

[0109] Select a visualization platform and obtain the cloud platform collection data, the fault prediction model prediction data, and the temperature trend prediction data for data visualization:

[0110] Displays real-time readings of current operating temperatures, as well as predicted values;

[0111] Show trend analysis of various sensor data and abnormal patterns that may indicate failure;

[0112] Lists recent historical fault events and corresponding alarms, including fault type, occurrence time, handling measures, and status.

[0113] By using a visualization platform, the data collected by the cloud platform, the predicted data of the fault prediction model, and the temperature prediction data are intuitively displayed. This not only enables operators to easily understand and monitor the status of equipment, but also enhances the transparency and accessibility of the data. The visual data display helps operators make decisions quickly, effectively respond to potential fault warnings, and adjust operating strategies in a timely manner, thereby improving the operating efficiency and safety of the entire power system.

[0114] Example 2

[0115] Reference Figure 2 , which is the second embodiment of the present invention, and which is different from the previous embodiment, provides a fault diagnosis system for a large-scale phase regulator, including:

[0116] Data acquisition module, which collects real-time data from various sensors and uses IoT devices to send the collected data to the cloud platform in real time;

[0117] The cloud platform data processing module cleans, normalizes and extracts features of the received data in the cloud;

[0118] The fault prediction module uses deep learning technology to analyze data, establish a fault prediction model, predict potential faults and diagnose the source of the fault;

[0119] Temperature trend prediction module, which analyzes the time series of temperature data, predicts future temperature changes, and adjusts maintenance strategies based on the predictions;

[0120] Visualization interface module, which displays real-time data, forecast results, system status and provides graphical representation of historical data and trend analysis.

[0121] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0122] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0123] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0124] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0125] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A fault diagnosis method for a large-scale phase regulator, characterized in that: include, Establish an IoT monitoring network to monitor key performance data and upload key performance data to the cloud platform in real time; Preprocess key performance data and extract features based on the cloud platform, and use deep learning algorithms to establish a multi-modal fault prediction model; Predict potential faults and diagnose fault sources based on sensor data and model-predicted fault indices; Collect data based on the operating temperature of the condenser, predict the operating temperature data of the condenser through trend analysis, and execute corresponding maintenance strategies based on the predicted value of the operating temperature of the condenser; Use the visualization platform to display the collected data of the cloud platform, the predicted data of the fault prediction model, and the operating temperature prediction data of the phase regulator; The cloud platform is used to pre-process key performance data and extract features. include, Acquire the detection data of accelerometers, Hall effect sensors and thermocouples based on the key performance data of the cloud platform, and perform data denoising and normalization; The vibration data is converted to frequency domain representation by fast Fourier transform, and the amplitude of the main frequency and the corresponding frequency value are extracted as the extracted features. A fixed time window is set for the monitoring data of the Hall effect sensor and the thermocouple, and the mean, standard deviation and peak statistics in each time window are calculated and used as the feature data of the corresponding sensor; The multi-modal fault prediction model is established using a deep learning algorithm to predict potential faults and diagnose the source of the fault based on sensor data and the fault index predicted by the model, including: Construct a fault prediction model, including input layer, LSTM layer, diagnosis layer and output layer; The sensor feature data is input into the LSTM layer and expressed as: h t,k =LSTM(x t,k ,h t-1,k ,θ k ) where x t,k represents the input data of the kth sensor at time t, h t-1,k represents the output data of the previous time step, h t,k represents the output data of the kth sensor at time t, θ k Represents all weights and biases of the LSTM network; Diagnostic layer evaluation output data h t,k The abnormal contribution of the output fault source is expressed as: Among them C t,k represents the fault contribution of the kth sensor at time t, w k represents the weight of the kth sensor, w j represents the weight of the jth sensor, h t,k represents the LSTM layer output of the kth sensor at time t, h t,j represents the LSTM layer output of the jth sensor at time t, and K represents the total number of sensors; The information of the comprehensive diagnosis layer is used to output the fault index, which is expressed as: Where F(t) represents the fault index at time t, sig represents the activation function sigmoid; Initialize θ of the LSTM layer k and the diagnostic layer w k With w j , for each training sample, the LSTM layer uses the current θ k Processing input feature data x t,k Generate h t,k , using the current w k With w j and h t,k Calculate the fault contribution C of each sensor t,k , complete the forward propagation; The cross entropy loss function is used to calculate the difference between the model output and the actual label, expressed as: Where N represents the total number of training samples, y i,k represents the true label of the i-th sample for the k-th sensor, p i,k represents the probability that the prediction result of the i-th sample belongs to the k-th sensor, and L represents the loss function value; According to the calculation of the loss function, the parameter θ is calculated by the back propagation algorithm k 、w k and w j The gradient of , using the gradient descent algorithm to update the model parameters to minimize the loss function, is expressed as: Where η is the learning rate, represents the updated θ k parameter, represents the updated w k parameter, represents the updated w j parameter; When the validation set loss decreases too much in consecutive iterations, the iteration is stopped and the corresponding θ is obtained. k 、w k and w j parameter; Substitute the validation set data into the model, divide the validation set data into failure results and normal operation results, and set the index threshold based on the validation output of the model; The sensor feature data collected in real time is input into the model to calculate the fault index F(t). When F(t) is greater than or equal to the index threshold, it is judged to be abnormal, and the fault contribution C of each sensor is calculated. t,k , and the sensor with the largest contribution is the fault source; The data collection is performed based on the phase regulator operating temperature, and the phase regulator operating temperature data is predicted through trend analysis, including: Collect the internal temperature data of the condenser in real time during operation, and perform data cleaning and normalization on the collected temperature data; A temperature trend analysis model is constructed to predict the operating temperature of large phase condensers, which can be expressed as: Where ΔT t Indicates the temperature difference between the current moment and the previous moment. Indicates the predicted temperature at the next moment, T t represents the temperature at the current moment, β0 represents the basic coefficient of the model, β1 represents the prediction coefficient of the model, and β2 represents the influence coefficient of the model change rate; The temperature training set is used to train the temperature trend analysis model. The optimal parameter value is calculated by minimizing the prediction error. The error function is expressed as: Where n represents the total number of data in the temperature training set, T t+1 They represent the measured values ​​of the temperature at the next moment in the training set respectively; The temperature parameters β0, β1 and β2 are calculated by gradient descent method respectively, and updated according to the gradient and learning rate in each iteration, which can be expressed as: Where j represents different parameters and a represents the learning rate; Repeat the update steps until the calculated value of the loss function decreases significantly, then stop iterating; The collected new temperature data is substituted into the model to calculate the predicted operating temperature. The operating temperature threshold is set based on the operation manual. When the predicted operating temperature is outside the operating temperature threshold range, it is determined to be a temperature abnormality.

2. The fault diagnosis method for a large-scale phase regulator according to claim 1, characterized in that: The establishment of an IoT monitoring network, monitoring key performance data, and uploading the key performance data to a cloud platform in real time includes: Install sensors at key locations of the phase regulator to obtain monitoring data and set the monitoring data as key performance data; Connect each sensor to local data storage for collection backup; Using secure wireless communication protocols, the collected data is transmitted to the cloud platform in real time.

3. The fault diagnosis method for a large-scale phase regulator according to claim 2, characterized in that: The corresponding maintenance strategy is executed according to the predicted value of the phase regulator operating temperature, including: When the operating temperature is determined to be abnormal, if the predicted value is lower than the temperature threshold range, the heating device is started; If the predicted value is higher than the temperature threshold range, the cooling device is activated; Based on the determination of temperature anomalies, an alarm is sent to the operator to remind the operator to take maintenance measures.

4. The fault diagnosis method for a large-scale phase regulator according to claim 1, characterized in that: The visualization platform is used to display the collected data of the cloud platform, the predicted data of the fault prediction model and the operating temperature prediction data of the phase regulator, including: Select a visualization platform and obtain the cloud platform collection data, the fault prediction model prediction data, and the temperature trend prediction data for data visualization: Displays real-time readings of current operating temperatures, as well as predicted values; Show trend analysis of various sensor data and abnormal patterns that may indicate failure; Lists recent historical fault events and corresponding alarms, including fault type, occurrence time, handling measures, and status.

5. A system based on the fault diagnosis method of a large-scale phase regulator according to any one of claims 1 to 4, characterized in that: include, Data acquisition module, which collects real-time data from various sensors and uses IoT devices to send the collected data to the cloud platform in real time; The cloud platform data processing module cleans, normalizes and extracts features of the received data in the cloud; The fault prediction module uses deep learning technology to analyze data, establish a fault prediction model, predict potential faults and diagnose the source of the fault; Temperature trend prediction module, which analyzes the time series of temperature data, predicts future temperature changes, and adjusts maintenance strategies based on the predictions; Visualization interface module, which displays real-time data, forecast results, system status and provides graphical representation of historical data and trend analysis.

6. A computer device comprising: A memory and a processor; the memory stores a computer program, wherein the processor implements the steps of the fault diagnosis method for a large-scale phase regulator according to any one of claims 1 to 4 when executing the computer program.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the fault diagnosis method for a large-scale phase regulator according to any one of claims 1 to 4 are implemented.

Images