Movable transformer operation monitoring method and device, medium and equipment

Abstract

The invention relates to the technical field of power equipment monitoring, in particular to a movable transformer operation monitoring method and device, a medium and equipment. The method comprises the following steps: acquiring an ultrasonic signal generated in the movable transformer in a monitoring period; performing fast Fourier transform on the ultrasonic signals, and extracting spectrum features of the ultrasonic signals; inputting the frequency spectrum characteristics into a partial discharge prediction model to obtain a predicted partial discharge result of the movable transformer at a first future time point corresponding to the monitoring period; acquiring monitoring data of the movable transformer in the monitoring period based on the predicted partial discharge result; inputting the monitoring data into a temperature prediction model to obtain a predicted hot spot temperature of the movable transformer at a second future time point corresponding to the monitoring period; and performing heat dissipation processing and risk early warning on the movable transformer based on the predicted partial discharge result and the predicted hot spot temperature. According to the invention, accurate prediction of partial discharge and thermal runaway of the movable transformer is realized.

Description

Technical Field

[0001] This application relates to the field of power equipment monitoring technology, and in particular to a method, device, medium and equipment for monitoring the operation of a portable transformer. Background Technology

[0002] Portable transformers are increasingly used in emergency power supply scenarios, but they still have significant shortcomings in insulation condition monitoring and thermal runaway early warning under high overload or complex operating conditions.

[0003] In related technologies, the monitoring of mobile transformers relies on static data, which cannot adapt to dynamic changes under complex operating conditions, resulting in large prediction errors, delayed response, and difficulty in achieving early warning and forward-looking control. Summary of the Invention

[0004] In view of this, this application provides a method, device, medium and equipment for monitoring the operation of a portable transformer. By acquiring ultrasonic signals and monitoring data inside the portable transformer, and combining a long short-term memory network model and a physical information neural network model, it can achieve accurate prediction of partial discharge and thermal runaway.

[0005] According to one aspect of this application, a method for monitoring the operation of a portable transformer is provided, comprising: Acquire ultrasonic signals generated inside the portable transformer during the monitoring period; Perform a Fast Fourier Transform on the ultrasonic signal to extract its spectral features; The spectral features are input into the partial discharge prediction model to obtain the predicted partial discharge result of the mobile transformer at the first future time point corresponding to the monitoring period. The partial discharge prediction model is trained based on historical ultrasonic signals and a long short-term memory network model. Based on the predicted partial discharge results, the monitoring data of the mobile transformer during the monitoring period is obtained. The monitoring data includes the operating data of the mobile transformer and the external environment data. The monitoring data is input into the temperature prediction model to obtain the predicted hot spot temperature of the portable transformer at the second future time point corresponding to the monitoring cycle. The temperature prediction model is trained based on historical monitoring data and a physical information neural network model. Based on the predicted partial discharge results and the predicted hot spot temperature, heat dissipation treatment and risk warning are carried out on the portable transformer.

[0006] According to another aspect of this application, a portable transformer operation monitoring device is provided, comprising: The first acquisition unit is used to acquire ultrasonic signals generated inside the movable transformer during the monitoring period; The first prediction unit is used to perform a fast Fourier transform on the ultrasonic signal to extract the spectral features of the ultrasonic signal; and to input the spectral features into a partial discharge prediction model to obtain the predicted partial discharge result of the portable transformer at the first future time point corresponding to the monitoring period. The partial discharge prediction model is trained based on historical ultrasonic signals and a long short-term memory network model. The second acquisition unit is used to acquire monitoring data of the mobile transformer within the monitoring period based on the predicted partial discharge result. The monitoring data includes the operation data of the mobile transformer and the external environment data. The second prediction unit is used to input the monitoring data into a temperature prediction model to obtain the predicted hot spot temperature of the portable transformer at the second future time point corresponding to the monitoring cycle. The temperature prediction model is trained based on historical monitoring data and a physical information neural network model. Based on the predicted partial discharge result and the predicted hot spot temperature, the unit performs heat dissipation treatment and risk warning for the portable transformer.

[0007] According to another aspect of this application, a readable storage medium is provided having a program or instructions stored thereon, which, when executed by a processor, implement the steps of the above-described portable transformer operation monitoring method.

[0008] According to another aspect of this application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the steps of the above-described portable transformer operation monitoring method.

[0009] By employing the above technical solutions, this application provides a method, device, medium, and equipment for monitoring the operation of a portable transformer. By acquiring ultrasonic signals and monitoring data inside the portable transformer, and combining a long short-term memory network model and a physical information neural network model, it achieves accurate prediction of partial discharge and hot spot temperature. Based on the prediction results, it performs heat dissipation treatment and risk warning, significantly improving the insulation status monitoring and thermal runaway warning capabilities of portable transformers under high overload or complex operating conditions, and meeting the diverse needs of emergency power supply scenarios.

[0010] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0011] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating the mobile transformer operation monitoring method provided in an embodiment of this application is shown. Figure 2 A flowchart illustrating a mobile transformer operation monitoring method according to another embodiment of this application is shown; Figure 3 A structural block diagram of a portable transformer provided in another embodiment of this application is shown; Figure 4 A structural block diagram of the portable transformer operation monitoring device provided in an embodiment of this application is shown. Detailed Implementation

[0012] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0013] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0014] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “attached” to another element, it can be directly connected or attached to the other element, or there may be intermediate elements. Furthermore, “connected” or “attached” as used herein can include wireless connections or wireless interconnections. The term “and / or” as used herein includes all or any unit and all combinations of one or more associated listed items.

[0015] Exemplary embodiments according to this application will now be described in more detail with reference to the accompanying drawings. However, these exemplary embodiments may be implemented in many different forms and should not be construed as being limited to the embodiments set forth herein. It should be understood that these embodiments are provided so that the disclosure of this application is thorough and complete, and that the concept of these exemplary embodiments is fully conveyed to those skilled in the art.

[0016] One embodiment of this application provides a method for monitoring the operation of a portable transformer, such as... Figure 1 As shown, the method includes: Step 101: Acquire the ultrasonic signals generated inside the movable transformer during the monitoring period.

[0017] It should be noted that a portable transformer is a type of transformer that can travel on roads. It possesses the towing capability of a vehicle, allowing it to move freely and change location at any time, thus meeting the power supply needs of engineering sites.

[0018] In this step, the structure of the portable transformer is designed in advance. The portable transformer includes a transformer module, a heat dissipation module, a connection module, a high and low voltage switchgear module, a synchronization and grid connection module, a mobile carrier module, and a relay protection module. The transformer module is used for the conversion and transmission of electrical energy. The heat dissipation module is located externally to the transformer module, dissipating the heat generated by the transformer module to the external environment in a timely manner, maintaining the normal operating temperature of the transformer module. The connection module enables a fast and reliable connection between the transformer module and the power grid. The high and low voltage switchgear module controls the connection between the transformer module and the power grid. The synchronization and grid connection module controls the synchronization of voltage, frequency, and phase between the transformer module and the power grid. The relay protection module quickly disconnects the circuit in the event of a fault in the portable transformer, protecting the equipment. The mobile carrier module carries the transformer module, heat dissipation module, connection module, high and low voltage switchgear module, synchronization and grid connection module, and relay protection module, enabling the portable transformer to be moved and transported between different locations.

[0019] Furthermore, for real-time monitoring of portable transformers in practical application scenarios, ultrasonic signals generated inside the portable transformer are collected by a surface acoustic wave (SAW) gas sensor embedded inside the portable transformer during different monitoring cycles. The collected ultrasonic signals reflect the partial discharge and insulation state changes inside the portable transformer.

[0020] Step 102: Perform a fast Fourier transform on the ultrasonic signal to extract the spectral features of the ultrasonic signal.

[0021] In this step, the acquired ultrasonic signal is subjected to a fast Fourier transform to convert the time-domain signal into a frequency-domain signal, and the spectral characteristics of the ultrasonic signal are extracted to intuitively reflect the partial discharge and insulation degradation inside the portable transformer through the spectral characteristics.

[0022] Step 103: Input the spectral features into the partial discharge prediction model to obtain the predicted partial discharge results of the portable transformer at the first future time point corresponding to the monitoring period.

[0023] The partial discharge prediction model is trained based on historical ultrasonic signals and a long short-term memory network model.

[0024] In this step, the spectral features are arranged chronologically and input into a pre-trained partial discharge prediction model as a feature sequence. This allows the model to output the predicted partial discharge result for the portable transformer at the first future time point corresponding to the monitoring period. The predicted partial discharge result includes both normal and partial discharge. If the predicted partial discharge result is partial discharge, it indicates that continued operation of the portable transformer in its current state will lead to partial discharge in the future. Therefore, a first warning message is generated based on the first future time point corresponding to the monitoring period and the predicted partial discharge result to remind staff to check promptly and prevent insulation failure caused by partial discharge, thus achieving early warning of insulation failure.

[0025] Here, the monitoring cycle refers to the time period for continuous monitoring of the portable transformer, which is a fixed time interval, such as every 5 minutes or every 10 minutes. This embodiment does not impose a specific limitation. The first future time point corresponding to the monitoring cycle refers to a future time point corresponding to a fixed first time interval counted backward from the end of the current monitoring cycle. It is used to predict whether partial discharge will occur at that future time point. For example, this time interval can be from a few minutes to tens of minutes. This embodiment does not impose a specific limitation.

[0026] It should be noted that the partial discharge prediction model is pre-trained based on historical ultrasonic signals and a Long Short-Term Memory (LSTM) network model. It can predict the partial discharge result of the portable transformer at the first future time point corresponding to the monitoring period based on the ultrasonic signals within the monitoring period.

[0027] Step 104: Based on the predicted partial discharge results, obtain the monitoring data of the portable transformer within the monitoring period.

[0028] The monitoring data includes operational data of the portable transformer and external environmental data.

[0029] In this step, based on the predicted partial discharge results, monitoring data of the portable transformer within the monitoring period is further acquired. This monitoring data is used to comprehensively assess the transformer's operating status. Specifically, if partial discharge is predicted to occur in the portable transformer during future operation, monitoring data of the portable transformer within the monitoring period will continue to be acquired. The monitoring data includes operational data of the portable transformer and external environmental data. Specifically, operational data may include the portable transformer's load current, hot spot temperature, magnetic flux density, voltage frequency, etc., while external environmental data may include the portable transformer's ambient temperature, etc.

[0030] Step 105: Input the monitoring data into the temperature prediction model to obtain the predicted hot spot temperature of the portable transformer at the second future time point corresponding to the monitoring cycle.

[0031] The temperature prediction model is trained based on historical monitoring data and a physical information neural network model.

[0032] In this step, the acquired monitoring data is input into the temperature prediction model to obtain the predicted hot spot temperature of the mobile transformer at the second future time point corresponding to the monitoring cycle, so as to predict whether the mobile transformer will experience thermal runaway.

[0033] The second future time point corresponding to the monitoring cycle refers to another future time point corresponding to another fixed second time interval counting backward from the end of the current monitoring cycle. It is used to predict whether thermal runaway will occur at that future time point. For example, this time interval can be from a few minutes to tens of minutes, and this embodiment does not impose specific limitations.

[0034] Here, the temperature prediction model will output the corresponding future predicted hot spot temperature for the monitoring data at different monitoring time points within the monitoring period. In this step, the highest predicted hot spot temperature corresponding to different monitoring time points within the monitoring period will be used as the predicted hot spot temperature of the portable transformer at the second future time point corresponding to the monitoring period.

[0035] It should be noted that the temperature prediction model is pre-trained based on historical monitoring data and a Physics-Informed Neural Networks (PINN) model. It can predict the hot spot temperature of the portable transformer at the second future time point corresponding to the monitoring period based on the monitoring data within the monitoring period.

[0036] It is worth mentioning that the time interval between the monitoring period and its corresponding first future time point is less than the time interval between the monitoring period and its corresponding second future time point. In other words, the second future time point corresponding to the monitoring period is later than its corresponding first future time point, in order to conform to the actual situation where partial discharge occurs first and then thermal runaway occurs.

[0037] Step 106: Based on the predicted partial discharge results and predicted hot spot temperature, perform heat dissipation treatment and risk warning for the portable transformer.

[0038] In this step, if the predicted hot spot temperature is greater than the temperature threshold of the portable transformer, it indicates that the portable transformer will experience thermal runaway during future operation. Based on the predicted partial discharge results and hot spot temperature, corresponding heat dissipation measures and risk warnings will be taken to prevent the portable transformer from being damaged due to partial discharge or overheating.

[0039] This embodiment acquires ultrasonic signals and monitoring data from inside the portable transformer in real time, and combines a long short-term memory network model and a physical information neural network model to achieve accurate prediction of partial discharge and hot spot temperature. Based on the prediction results, heat dissipation treatment and risk warning are carried out, which significantly improves the ability of portable transformers to monitor insulation status and provide early warning of thermal runaway under high overload or complex operating conditions. At the same time, through modular design and heat dissipation treatment, the stable operation and rapid deployment of portable transformers in extreme environments are ensured, meeting the diverse needs of emergency power supply scenarios.

[0040] Another embodiment of this application provides a method for monitoring the operation of a portable transformer, such as... Figure 2 As shown, the method includes: Step 201: Perform structural design for the portable transformer.

[0041] It should be noted that a portable transformer is a type of transformer that can travel on roads. It possesses the towing capability of a vehicle, allowing it to move freely and change location at any time, thus meeting the power supply needs of engineering sites.

[0042] In this step, such as Figure 3As shown, the portable transformer includes a transformer module, a heat dissipation module, a connection module, a high and low voltage switchgear module, a synchronization and grid connection module, a mobile carrier module, and a relay protection module. The transformer module is used for the conversion and transmission of electrical energy. The heat dissipation module, located externally to the transformer module, dissipates the heat generated by the transformer module to the external environment in a timely manner, maintaining the normal operating temperature of the transformer module. The connection module enables a fast and reliable connection between the transformer module and the power grid. The high and low voltage switchgear module controls the connection between the transformer module and the power grid. The synchronization and grid connection module controls the synchronization of voltage, frequency, and phase between the transformer module and the power grid. The relay protection module quickly disconnects the circuit to protect the equipment in the event of a fault in the portable transformer. The mobile carrier module carries the transformer module, heat dissipation module, connection module, high and low voltage switchgear module, synchronization and grid connection module, and relay protection module, enabling the portable transformer to be moved and transported between different locations.

[0043] Specifically, the transformer module includes an oil tank and a transformer body composed of windings and a core. The transformer body is immersed in the oil tank, which contains an insulating mixed heat dissipation medium. This mixed heat dissipation medium carries away the heat generated by the transformer body through its flow. The flow rate of the mixed heat dissipation medium can be adjusted. The modules automatically attach to each other via a magnetic-mechanical locking composite interface, ensuring a secure connection.

[0044] For a specific example, in this step, the core uses an orthogonal laminated structure of 0.23mm silicon steel sheets and 0.02mm nanocrystalline alloy strips. The nanocrystalline alloy strips serve as a supplementary material to the silicon steel sheets, optimizing the core's magnetic properties to reduce localized losses. The nanocrystalline alloy strips comprise 10% of the core, ensuring overall core performance while fully utilizing their high permeability and low loss characteristics to further reduce transformer iron losses and temperature rise. The core lamination factor is less than 0.97. The lamination factor measures the tightness of the core packing and is the ratio of the actual area of ​​the silicon steel sheets to the total cross-sectional area of ​​the core. A lamination factor closer to 1 indicates smaller air gaps, lower magnetic reluctance, and higher magnetic flux transmission efficiency. Setting the lamination factor to 0.97 ensures a certain proportion of air gaps in the core, helping to reduce hysteresis and eddy current losses. In the iron core, the seams of adjacent silicon steel sheets are staggered to form a stepped air gap, which reduces magnetic flux leakage and noise, reduces magnetic flux concentration at the seams, lowers the peak magnetic flux density, and thus reduces noise and vibration caused by magnetostriction. Setting the stepped seam air gap to less than or equal to 0.05 mm improves the magnetic flux transmission efficiency of the iron core and reduces losses.

[0045] The oil tank is equipped with a pressure relief valve with an opening pressure of 50 kPa and a closing pressure of 20 kPa. When the internal pressure of the oil tank becomes too high, the valve automatically opens to release the pressure and prevent the tank from rupturing. A combustible gas sensor is built into the outer shell of the oil tank, which can monitor the leakage of combustible gases inside the portable transformer in real time, ensuring operational safety.

[0046] The windings are vacuum-cast with epoxy resin to form a dense insulating layer, maintaining good insulation performance and mechanical strength at high temperatures. This effectively isolates different parts of the windings, preventing leakage and ensuring stable operation of the portable transformer under varying temperature and humidity conditions. The high-voltage and low-voltage windings are wound separately on the iron core. The high-voltage winding is responsible for inputting electrical energy, while the low-voltage winding is responsible for outputting electrical energy. Energy transfer is achieved through magnetic flux coupling between the windings within the iron core. In this step, the high-voltage winding uses a twisted winding method to reduce electromagnetic interference between windings. The inter-turn capacitance of the high-voltage winding is set to less than or equal to 1000 picofarads (pF) to reduce uneven voltage distribution and partial discharge. The high-voltage winding conductors use high-conductivity copper with a purity of less than or equal to 99.97% to improve conductivity and reduce losses. The lead connectors connecting the high-voltage winding to the external circuit contain a Bi-2223 superconducting transition section to reduce contact resistance and heat loss. The low-voltage winding adopts a copper foil structure with a copper foil thickness between 0.1 and 0.2 mm. The copper foil surface is tin-plated to improve corrosion resistance and solderability.

[0047] The hybrid heat dissipation medium uses a cooling liquid and a magnetically sensitive nanofluid. In this step, heat dissipation pipes can be installed within the heat dissipation module. The hybrid heat dissipation medium flows from the bottom of the oil tank via an oil pump into the heat dissipation pipes within the heat dissipation module. Within the heat dissipation module, heat is dissipated to the external environment through forced air cooling. The cooled hybrid heat dissipation medium then flows back to the oil tank through the heat dissipation pipes, forming a circulation. The flow rate of the hybrid heat dissipation medium can be controlled by the oil pump.

[0048] The heat dissipation module includes a heat sink, which can be an aluminum alloy finned tube and a cooling fan. The fins of the aluminum alloy finned tube heat sink have a fin height of 15-20mm and a fin spacing of 3-5mm. The aluminum alloy finned tube incorporates a paraffin-based phase change material with a phase change temperature of 60℃ and a latent heat of less than or equal to 200kJ / kg. This allows it to store and release heat energy under overload conditions by utilizing the property of the phase change material to absorb or release a large amount of heat during the phase change process, thus reducing temperature rise. Simultaneously, the aluminum alloy finned tube heat sink undergoes anodizing treatment to improve its oxidation and corrosion resistance, ensuring timely heat dissipation during high overload operation and maintaining the normal operating temperature of the portable transformer. The cooling fan is an axial flow fan, and the fan motor uses variable frequency speed control to reduce energy consumption.

[0049] The mobile carrier module uses a full trailer chassis and is equipped with a hydraulic support system. On unpaved roads or complex terrain, the mobile carrier module can be equipped with tracked modules to improve its mobility.

[0050] The access module is equipped with a 35kV infrared alignment quick-connect head on the high-voltage side of the portable transformer for quick and safe connection to the external high-voltage power grid. The access module is also equipped with a busbar interface on the low-voltage side of the portable transformer for connecting the low-voltage output of the transformer to the load equipment.

[0051] The synchronous grid-connection module uses virtual synchronous machine technology to simulate the characteristics of a synchronous generator in order to achieve seamless grid connection.

[0052] The relay protection module integrates multiple protection functions such as instantaneous overcurrent, zero-sequence, and over-temperature protection. A zinc oxide surge arrester is installed on the high-voltage side of the portable transformer to effectively suppress lightning overvoltage and switching overvoltage. Simultaneously, a surge protector is installed on the low-voltage side to prevent low-voltage electrical equipment from being subjected to surge voltage impacts.

[0053] The various modules of the portable transformer are integrated into a closed integrated compartment. The compartment body adopts a honeycomb-truss composite biomimetic structure, and the skeleton of the honeycomb-truss composite biomimetic structure is made of high-strength, lightweight carbon fiber reinforced resin-based material. The honeycomb-truss composite biomimetic structure has an aluminum foil reflective outer shell, which has heat insulation and reflective properties, and can reduce the influence of the external environment on the internal temperature.

[0054] In this embodiment, the structural design of the portable transformer integrates magnetic and thermal properties to improve its high overload capacity and ensure that the temperature rise of the portable transformer under overload conditions is controlled within a safe range.

[0055] Step 202: Obtain historical ultrasonic signals generated inside the portable transformer during the historical monitoring period; based on the historical partial discharge results of the portable transformer at the first future time point corresponding to the historical monitoring period, label the historical ultrasonic signals, and generate a first sample set based on the labeled historical ultrasonic signals.

[0056] The historical partial discharge results include both normal and partial discharges.

[0057] It should be noted that in portable transformers, the accumulation of interface charge and early discharge are difficult to detect. Space charge regions are easily formed at the interface between solid insulation and liquid insulation. Under the combined action of overload electric field and temperature field, irreversible insulation degradation can occur.

[0058] In this step, several surface acoustic wave (SAW) gas sensors are embedded within the inter-turn insulation material of the high-voltage winding to capture ultrasonic signals generated by partial discharge. This step uses SAW gas sensors to collect historical ultrasonic signals generated inside the portable transformer during different historical monitoring periods at a fixed sampling frequency. For any historical ultrasonic signal within a historical monitoring period, a Fast Fourier Transform (FFT) is performed to convert the historical ultrasonic signal from a time-domain signal to a frequency-domain signal. Key spectral features of the historical ultrasonic signal are then precisely quantified from the frequency-domain signal to capture the shift in the historical ultrasonic signal caused by partial discharge. Furthermore, based on the historical partial discharge results at a first future time point after the portable transformer has operated for a first time interval within that historical monitoring period, the spectral features corresponding to each historical ultrasonic signal generated within that historical monitoring period are labeled. Further, the labeled spectral features within that historical monitoring period are arranged chronologically to generate a first sample in the form of a feature sequence. Thus, a first sample set is generated based on the first samples corresponding to different historical monitoring periods. The temporal characteristics of the first sample set thus reflect the dynamic and time-varying process of insulation degradation in the portable transformer caused by partial discharge.

[0059] Here, historical partial discharge results include both normal and partial discharges, which can be further divided into early discharges and severe discharges. Frequency domain characteristics include the center frequency shift and phase shift of historical ultrasonic signals. Specifically, the micro-strain generated within the insulating material due to partial discharge alters the physical properties of the ultrasonic signal propagation path, resulting in an extremely small but detectable shift in its resonant frequency. Furthermore, changes in mechanical stress or density caused by partial discharge also affect the propagation speed of the ultrasonic signal, manifesting as a phase shift.

[0060] This step incorporates historical ultrasonic signals, including those from the mobile transformer under normal conditions as well as those from before and after partial discharge, to enhance data comprehensiveness. Furthermore, labeling the data with historical partial discharge results from a first future time point at a time interval from the historical monitoring cycle enables the subsequent partial discharge prediction model to provide early warning capabilities, allowing time for operational adjustments to the mobile transformer and ensuring its safe and reliable operation.

[0061] Step 203: Input the first sample set into the long short-term memory network model, and perform backpropagation and gradient calculation on the model parameters of the long short-term memory network model based on the preset loss function. Use the accelerated gradient adaptive moment estimation algorithm and the variable parameter neurodynamic algorithm to iteratively update the model parameters of the long short-term memory network model. When the change of the preset loss function is less than the preset threshold or the number of iterations is equal to the maximum number of iterations, output the long short-term memory network model as the partial discharge prediction model.

[0062] In this step, the initial long short-term memory network model is trained using the first sample set. Through the training process, the long short-term memory network model is made to have a certain partial discharge prediction capability, and the trained long short-term memory network model is used as a partial discharge prediction model.

[0063] It should be noted that Long Short-Term Memory (LSTM) networks are time-centric recurrent neural networks, mainly composed of forget gates, memory gates, and output gates. The forget gate determines how much of the previous time step's cell state is retained in the current time step's cell state; the input gate determines how much of the current time step's input is retained in the current time step's cell state; and the output gate determines how much of the current time step's cell state is output.

[0064] Specifically, the first sample set is input into the Long Short-Term Memory (LSTM) network model. The LTM network model processes the first sample set step by step, continuously updating its internal state and learning the dynamic patterns contained in each feature sequence. The LTM network model ultimately outputs a prediction result, which determines the discharge state of each input feature sequence at its corresponding first future time point: normal, early discharge, or severe discharge. During the training process of the LTM network model, backpropagation and gradient calculation can be performed on the model parameters based on common preset loss functions such as the binary cross-entropy loss function. The model parameters are then iteratively updated using accelerated gradient adaptive moment estimation algorithms and variable parameter neurodynamic algorithms until the change in the preset loss function is less than a preset threshold or the number of iterations equals the maximum number of iterations. At this point, the LTM network model is output as a partial discharge prediction model.

[0065] Understandably, before partial discharge occurs, charge begins to slowly accumulate at the interface of the insulating material inside the portable transformer, generating a continuous electrostatic force. This force causes micron-level, continuous deformation of the insulating material. This is reflected in changes in the spectral characteristics of the ultrasonic signal acquired by the surface acoustic wave gas sensor; for example, the center frequency of the ultrasonic signal drifts slowly but continuously, rather than fluctuating randomly. The partial discharge prediction model trained in this step can identify continuous, trend-like changes in the feature sequence, thereby combining the feature sequence with its future development trend to obtain the future partial discharge results of the feature sequence, achieving early warning of partial discharge.

[0066] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to fully illustrate the specific implementation process of this embodiment, the model parameters of the Long Short-Term Memory (LSTM) network model are iteratively updated using an accelerated gradient adaptive moment estimation algorithm and a variable-parameter neurodynamics algorithm. Specifically, this includes: updating the learning rate of the LSTM network model during the current iteration of training based on a preset convergence rate constant and a preset mapping function to obtain the learning rate of the LSTM network model during the next iteration of training; calculating the first and second moment estimates of the gradient during the current iteration of training based on the gradient of the LSTM network model during the current iteration of training; correcting the first and second moment estimates of the gradient based on the correction bias of the first and second moment estimates of the gradient during the current iteration of training; and updating the model parameters of the LSTM network model during the current iteration of training based on the learning rate of the LSTM network model during the next iteration of training and the corrected first and second moment estimates to obtain the model parameters of the LSTM network model during the next iteration of training.

[0067] In this step, during the training of the Long Short-Term Memory (LSTM) network model, based on updating the model parameters of the LSM network model using the Nesterov-accelerated Adaptive Moment Estimation (Nadam) algorithm, a variable-parameter neurodynamics algorithm is used to control the learning rate in the Nesterov-accelerated Adaptive Moment Estimation algorithm, so that it is no longer a constant, but dynamically adjusted over time.

[0068] It should be noted that Nadam combines Adaptive Moment Estimation (Adam) and Nesterov Accelerated Gradient (NAG) algorithms, integrating the advantages of adaptive learning rate and predictive momentum to improve the performance of the optimization algorithm. Variable-parameter neurodynamics is a method based on neurodynamics, which uses differential equations to describe the evolution of the error or loss function. In variable-parameter neurodynamics, the convergence rate constant and mapping function can be dynamically adjusted or selected as needed, rather than remaining fixed.

[0069] Specifically, during the training of the Long Short-Term Memory (LSTM) network model, the dynamic learning rate is controlled by a variable-parameter neurodynamics algorithm, characterized as follows: .in, , representing a continuous-time variable, is discretized into the number of iterations during the actual training of a Long Short-Term Memory (LSTM) network model. k ; d The differential symbol; λ is the learning rate of the long short-term memory network model that varies over time; λ is a preset constant that controls the convergence speed (i.e., the preset convergence speed constant), which determines the magnitude of the learning rate adjustment. w ( t For Long Short-Term Memory network models in t Model parameters under time variables; L ( w ( t () represents the preset loss function, indicating the loss function of the Long Short-Term Memory network model in terms of model parameters. w ( t Prediction error under ( ) The preset mapping function is a monotonically increasing odd function. This function maps the loss value of the Long Short-Term Memory (LSTM) network model to an appropriate adjustment level, ensuring that as the prediction error increases, the adjustment of the learning rate also increases, thereby accelerating convergence. For example, the preset mapping function can be set to... λ is 0.1.

[0070] In this step, the dynamic learning rate formula described above ensures that the adjustment of the learning rate is positively correlated with the magnitude of the loss value. When the loss value is large, the dynamic learning rate formula can significantly reduce the learning rate to improve the stability and reliability of the convergence process, avoiding oscillations and divergences during training, thus allowing the loss value to converge to zero more smoothly and quickly.

[0071] Furthermore, by discretizing the above continuous-time dynamic learning rate formula using the forward Euler method of numerical integration, it can be transformed into: .in, For the first k The learning rate of the short-term memory network model during each iteration of training; For the first k +1 iteration training duration: learning rate of the short-term memory network model; , The strength coefficient adjusted for the learning rate. The time interval for updating the learning rate; For the first k The loss value of the Long Short-Term Memory (LSTM) network model during each iteration of training represents the performance of the LTM network model. k Model parameters during the next iteration of training w k The following performance.

[0072] Therefore, after calculating the learning rate for the next iteration of the SMART network model based on the learning rate of the current iteration, the Nadam update is performed using this new learning rate to obtain the model parameters for the next iteration of the SMART network model. Specifically, the first-order moment estimation is initialized. m 0 and second moment estimation v 0, initialization For example, initializing the first-order moment estimate. m 0 and second moment estimation v 0 is 0, initialization The value is 0.01. Furthermore, in each training iteration, the learning rate is updated. Calculate the gradient Next, based on the first k Gradient of the Short-Term Memory Network Model during Sub-Iteration Training Update # k First and second moment estimation of gradient during subsequent training iterations , , is represented as: , ,in, and The exponential decay rates estimated by the first and second moments, respectively. and The first The first and second moments of the gradient are estimated during the nth iteration of training. Then, the first and second moments of the gradient are calculated. k First and second moment estimates after bias correction during the next iteration of training. , , is represented as: , ,in, and The first kThe correction bias of the first and second momentum estimates of the gradient during the nth iteration of training. Further, the calculation of the... k Momentum correction during the next iteration of training , Therefore, the model parameters are updated to obtain the first... k Model parameters at +1 training iteration w k+1 , is represented as: ,in, It is a constant.

[0073] Step 204: Construct a second sample set of the portable transformer based on the historical monitoring data and simulated monitoring data of the portable transformer.

[0074] The second sample in the second sample set includes the first sample monitoring data of the portable transformer at the sample time point, and the second sample monitoring data at the second future time point corresponding to the sample time point.

[0075] In this step, to increase the richness of the sample, a finite element model of the portable transformer is constructed, and simulated monitoring data of the portable transformer is obtained using finite element software. Simultaneously, historical monitoring data of the portable transformer within the historical monitoring period is acquired. Therefore, a second sample set of the portable transformer is constructed using the simulated monitoring data and the historical monitoring data. Each second sample in the second sample set includes the first sample monitoring data of the portable transformer at the sample time point, and the second sample monitoring data at the corresponding second future time point. The first and second sample monitoring data are determined based on the simulated monitoring data and the historical monitoring data.

[0076] It should be noted that the sample time point corresponds to the first sample monitoring data. If the first sample monitoring data is historical monitoring data, the sample time point is a historical time point within the historical monitoring period; similarly, if the first sample monitoring data is simulated monitoring data, the sample time point is the simulated time point during finite element simulation. In this step, the second sample uses the second sample monitoring data at the second future time point corresponding to the sample time point as the label of the first sample monitoring data, so that the physical information neural network model trained using the second sample can have a certain temperature prediction capability. It is worth mentioning that the time interval between the sample time point and its corresponding second future time point in the second sample is greater than the time interval between the historical monitoring period and its corresponding first future time point in the first sample, so that after predicting partial discharge, the thermal runaway problem caused by insulation failure due to partial discharge can be predicted.

[0077] Specifically, the first sample monitoring data includes the operational data of the portable transformer at the sample time point, external environmental data, and parameter information of the portable transformer. The operational data includes the portable transformer's load current, hot spot temperature, magnetic flux density, and voltage frequency, while the external environmental data includes the ambient temperature. The parameter information of the portable transformer includes the volume of the core, the resistance of the windings at different temperatures, the effective heat dissipation area, the specific heat capacity, the total mass, and the temperature threshold. Here, the hot spot temperature is the temperature of the hottest point inside the portable transformer, which can be collected using a fiber optic grating hot spot temperature sensor. The second sample monitoring data includes the operational data of the portable transformer at a second future time point corresponding to the sample time point, and external environmental data. For example, the second sample monitoring data includes the sample hot spot temperature, magnetic flux density, voltage frequency, and winding resistance of the portable transformer at the second future time point corresponding to the sample time point.

[0078] Step 205: Input the second sample into the physical information neural network model, and construct the target loss function of the physical information neural network model based on the predicted hotspot temperature of the second sample output by the physical information neural network model; calculate the target loss function and update the model parameters of the physical information neural network model until the target loss function converges, and output the physical information neural network model as the temperature prediction model.

[0079] In this step, the second sample is input into the initial physical information neural network model to train the model. During training, the physical information neural network model can predict the hotspot temperature of the portable transformer at a second future time point corresponding to the sample time point, starting from the time point of the second sample and operating according to the state reflected in the monitoring data of the first sample after a second time interval. The model then outputs the predicted hotspot temperature of the second sample. Based on this predicted hotspot temperature, the target loss function of the physical information neural network model is constructed. The target loss function is calculated, and the model parameters of the physical information neural network model are updated until the target loss function converges. The physical information neural network model is then output as the temperature prediction model.

[0080] It should be noted that the Physical Information Neural Network (PINN) model is a machine learning model that combines deep learning with knowledge of physics. During the learning process, the PINN model utilizes physical laws to guide the model, thereby improving its generalization ability.

[0081] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to fully illustrate the specific implementation process of this embodiment, a target loss function of the physical information neural network model is constructed based on the predicted hotspot temperature of the second sample output by the physical information neural network model. Specifically, this includes: constructing a data loss function based on the predicted hotspot temperature of the second sample and the sample hotspot temperature in the second sample monitoring data; determining the heating power of the core based on the magnetic induction intensity and voltage frequency of the movable transformer in the second sample monitoring data and the volume of the core inside the movable transformer in the first sample monitoring data; and determining the heating power of the core based on the load current of the movable transformer and the winding of the movable transformer in the second sample monitoring data. The heating power of the winding is determined by the resistance at the hot spot temperature of the sample. The heat dissipation power of the mobile transformer is determined based on the effective heat dissipation area and ambient temperature of the mobile transformer in the first sample monitoring data and the hot spot temperature of the sample in the second sample monitoring data. The total loss power of the mobile transformer is determined based on the heating power of the core, the heating power of the winding, and the heat dissipation power of the mobile transformer. A physical loss function is constructed based on the predicted hot spot temperature of the second sample, the total loss power of the mobile transformer, and the specific heat capacity and total mass of the mobile transformer in the first sample monitoring data. A constraint loss function is constructed based on the predicted hot spot temperature of the second sample and the temperature threshold of the mobile transformer.

[0082] The target loss function includes the data loss function, the physical loss function, and the constraint loss function.

[0083] In this step, for example, the data loss function Represented as: , in, N The number of the second sample. For the second sample i The sample hotspot temperature within the second sample monitoring data. The second sample output by the physical information neural network model i Predicted hotspot temperatures.

[0084] Here, the data loss function calculates the average of the squared differences between the temperature predicted by the physical information neural network model and the actual measured temperature. The data loss function guides the physical information neural network model to learn from existing samples, ensuring that the predicted hotspot temperatures output by the model closely match the actual observations.

[0085] Physical loss function Represented as: , in, The second sample output by the physical information neural network model i The derivative of the predicted hotspot temperature with respect to time, i.e. the rate of change of the predicted hotspot temperature, reflects how much the predicted hotspot temperature increases or decreases per unit time. In the physical information neural network model, this term can be directly calculated from the predicted hotspot temperature output by the physical information neural network model through automatic differentiation. The magnetic loss coefficient of a portable transformer is the hysteresis loss coefficient, which measures the proportion of energy lost by the magnetic material in the portable transformer during magnetization and demagnetization in an alternating magnetic field. When the magnetic material in the portable transformer is in an alternating magnetic field, there will be energy loss during its magnetization process. This energy loss is converted into heat energy, causing the portable transformer to heat up. For the second sample i The magnetic induction intensity within the monitoring data of the second sample in the middle. For the second sample i Voltage frequency within the second sample monitoring data. For the second sample i The volume of the iron core in the first sample monitoring data. Indicates the movable transformer in the second sample i The actual heat loss of the iron core affected by hysteresis at the second future time point corresponding to the sample time point, i.e. the heating power of the iron core, reflects the influence of the iron core's hysteresis characteristics on the loss. The temperature coefficient of the winding resistance. For the second sample i Load current in the second sample monitoring data. For the second sample i The second sample monitoring data shows the resistance of the winding at the hot spot temperature of the sample. Here, the winding resistance changes with temperature. A function of the winding resistance changing with temperature can be preset, so that the resistance of the winding at different temperatures can be obtained directly through the mapping relationship of the function. Indicates a movable transformer in the second sample i The heat loss of the winding at the second future time point corresponding to the sample time point, that is, the heating power of the winding. For heat dissipation coefficient, For the second sample i The ambient temperature in the first sample monitoring data and the second sample i The actual temperature difference between hotspot temperatures within the second sample monitoring data. For the second sample i The effective heat dissipation area within the first sample monitoring data. This represents the heat dissipation power of a portable transformer under actual temperature differences. For the second sample iThe specific heat capacity within the first sample monitoring data. m For the second sample i The total mass within the first sample monitoring data. It is the overall heat capacity of a portable transformer, representing the amount of heat required to raise the temperature by one unit. Indicates the movable transformer in the second sample i The actual total power loss at the second future time point corresponding to the sample time point. According to the laws of thermodynamics, Indicates the movable transformer in the second sample i The rate of change of the sample hotspot temperature at the second future time point corresponding to the current sample time point, that is, .

[0086] It should be noted that the above-mentioned magnetic loss coefficient, temperature coefficient, and heat dissipation coefficient are specifically set according to the different materials inside the portable transformer in the actual application scenario.

[0087] Here, the physical loss function calculates the difference between the rate of temperature change predicted by the physical information neural network model and the actual rate of temperature change calculated by physical laws. By minimizing the physical loss function, the temperature change law learned by the physical information neural network model must strictly conform to the law of conservation of energy, thereby improving the reliability of the temperature prediction by the physical information neural network model.

[0088] The constraint loss function is determined based on the difference between the predicted hot spot temperature of the second sample of the temperature output of the physical information neural network model and the temperature threshold of the movable transformer, so that the physical information neural network model conforms to the constraints of the real physical world.

[0089] Furthermore, a target loss function is constructed based on the data loss function, physical loss function, and constraint loss function. L , is represented as: .in, To constrain the loss function, , , These are the weights for the data loss function, physical loss function, and constraint loss function, respectively, which can be set according to the specific application scenario.

[0090] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to fully illustrate the specific implementation process of this embodiment, the iron core adopts an orthogonal stacked structure of silicon steel sheets and nanocrystalline alloy strips. The method further includes: determining the hysteresis loop of the iron core based on the saturation magnetization of the iron core, the characteristic parameters of the nanocrystalline alloy strip, and the magnetic induction intensity of the movable transformer in the second sample monitoring data; performing loop integration on the hysteresis loop of the iron core to determine the hysteresis loss of the iron core; and updating the heating power of the iron core based on the hysteresis loss of the iron core.

[0091] In this step, the heating power of the iron core is updated based on the orthogonal laminated structure of silicon steel sheets and nanocrystalline alloy strips.

[0092] Specifically, when the iron core employs an orthogonal laminated structure of silicon steel sheets and nanocrystalline alloy strips, the actual magnetization process of the iron core exhibits both nonlinear hysteresis loss and linear eddy current loss. This step uses the Preisach model to describe the shape and area of ​​the hysteresis loop, where hysteresis loss is proportional to the area of ​​the hysteresis loop. Therefore, by calculating the heat loss of the iron core based on the dynamic loss of real-time magnetic induction intensity and material properties, the physical information neural network model can respond more sensitively to changes in the magnetic field, thereby improving the accuracy of temperature prediction. Ultimately, this allows the physical laws learned by the physical information neural network model to more closely approximate the real world.

[0093] It should be noted that the Preisach model is a mathematical model used to describe the nonlinear characteristics of hysteresis or piezoelectric materials.

[0094] For example, the Preisach model adopts The function is represented as follows: , in, Magnetization is , and magnetic flux density is . The function characterizes the magnetic flux density of the iron core. The degree to which something is magnetized under the influence of an action. The saturation magnetization of the iron core represents the maximum magnetization intensity when the iron core reaches magnetic saturation, reflecting the upper limit of the material's magnetization capability. After magnetic saturation, the magnetization intensity no longer increases significantly. The characteristic parameter for characterizing the initial magnetic permeability of nanocrystalline alloy strips determines the rate at which the magnetization intensity in the low magnetic field region increases with the magnetic induction intensity. To characterize the characteristic magnetic field parameters of nanocrystalline alloys with irreversible magnetic domain flipping, The weighting coefficient for the nonlinear term is the amplitude coefficient of the second term in the formula, which determines the contribution of the irreversible magnetization process to the total magnetization. For example, , , , .

[0095] Furthermore, through The nonlinear characteristics can be applied in different directions ( or On the same The value gives different MThe value is used to construct the entire hysteresis loop. The hysteresis loss within one cycle is equal to the area enclosed by the hysteresis loop. By performing loop integration on the hysteresis loop, the hysteresis loss of the iron core with an orthogonal stacked structure of silicon steel sheets and nanocrystalline alloy strips is obtained. The heating power of the iron core is then updated based on the calculated hysteresis loss.

[0096] Step 206: Acquire the ultrasonic signal generated inside the portable transformer during the monitoring period; perform a fast Fourier transform on the ultrasonic signal to extract the spectral features of the ultrasonic signal; input the spectral features into the partial discharge prediction model to obtain the predicted partial discharge result of the portable transformer at the first future time point corresponding to the monitoring period.

[0097] In this step, for real-time monitoring of portable transformers in practical application scenarios, ultrasonic signals generated inside the portable transformer are periodically acquired within different monitoring periods. Then, a Fast Fourier Transform (FFT) is performed on the ultrasonic signals to extract their spectral features. Next, these spectral features are arranged chronologically and input as a feature sequence into a pre-trained partial discharge prediction model. This model outputs the predicted partial discharge result for the portable transformer at the first future time point corresponding to the monitoring period. The predicted partial discharge result includes both normal and partial discharge. If the predicted partial discharge result is partial discharge, it indicates that continued operation of the portable transformer in its current state will lead to partial discharge in the future. Therefore, a first warning message is generated based on the first future time point corresponding to the monitoring period and the predicted partial discharge result to remind personnel to check promptly and prevent insulation failure caused by partial discharge, thus achieving early warning of insulation failure.

[0098] Here, the monitoring cycle refers to the time period for continuous monitoring of the portable transformer, which is a fixed time interval, such as every 5 minutes or every 10 minutes. This embodiment does not impose a specific limitation. The first future time point corresponding to the monitoring cycle refers to a future time point corresponding to a fixed first time interval counted backward from the end of the current monitoring cycle. It is used to predict whether partial discharge will occur at that future time point. For example, this time interval can be from a few minutes to tens of minutes. This embodiment does not impose a specific limitation.

[0099] Step 207: If the predicted partial discharge result is partial discharge, obtain the monitoring data of the portable transformer within the monitoring period; input the monitoring data into the temperature prediction model to obtain the predicted hot spot temperature of the portable transformer at the second future time point corresponding to the monitoring period.

[0100] The second future time point corresponding to the monitoring cycle refers to another future time point corresponding to another fixed second time interval counting backward from the end of the current monitoring cycle. It is used to predict whether thermal runaway will occur at that future time point. For example, this time interval can be from a few minutes to tens of minutes, and this embodiment does not impose specific limitations.

[0101] In this step, if partial discharge is predicted to occur in the portable transformer during future operation, monitoring data of the portable transformer within the monitoring period is acquired. This monitoring data is then input into a pre-trained temperature prediction model to obtain the predicted hotspot temperature of the portable transformer at the second future time point corresponding to the monitoring period, thus predicting whether thermal runaway will occur in the portable transformer. The monitoring data includes the operational data of the portable transformer and external environmental data.

[0102] It should be noted that the temperature prediction model will output the corresponding future predicted hot spot temperature for the monitoring data at different monitoring time points within the monitoring period. In this step, the highest predicted hot spot temperature corresponding to different monitoring time points within the monitoring period will be used as the predicted hot spot temperature of the portable transformer at the second future time point corresponding to the monitoring period.

[0103] Here, the time interval between the monitoring period and its corresponding first future time point is less than the time interval between the monitoring period and its corresponding second future time point. That is, the second future time point corresponding to the monitoring period is later than its corresponding first future time point, in order to conform to the actual situation where partial discharge occurs first and then thermal runaway occurs.

[0104] Step 208: If the predicted hot spot temperature is greater than the temperature threshold of the portable transformer, determine the heat generation of the portable transformer according to the preset overload rate, the rated power of the portable transformer, and the preset heat dissipation time; determine the heat dissipation of the mixed heat dissipation medium according to the flow rate of the mixed heat dissipation medium, monitoring data, and the predicted hot spot temperature; calculate the heat absorption of the phase change material in the heat dissipation module due to phase change; construct an objective function with the goal that the sum of the heat dissipation of the mixed heat dissipation medium and the heat absorption of the heat dissipation module is less than the heat generation of the portable transformer; solve the objective function to obtain the target flow rate of the mixed heat dissipation medium, and control the flow of the mixed heat dissipation medium according to the target flow rate for heat dissipation treatment.

[0105] In this step, if the predicted hotspot temperature exceeds the temperature threshold of the portable transformer, it indicates that continued operation of the portable transformer in its current state will not only lead to partial discharge and insulation failure in the future, but also further cause thermal runaway. Therefore, a second early warning message is generated based on the second future time point corresponding to the monitoring cycle and the predicted hotspot temperature to remind staff to reduce the load on the portable transformer in a timely manner, achieving early warning of thermal runaway. Simultaneously, efficient heat dissipation treatment is implemented for the portable transformer due to its unique heat dissipation structure.

[0106] Specifically, this step pre-sets a heat dissipation time, which is less than the time interval between the monitoring period and its second future time point. This ensures that the portable transformer can completely dissipate heat before thermal runaway occurs, preventing thermal runaway. Simultaneously, this step pre-sets a relatively high overload rate for the heat dissipation process, ensuring that the portable transformer can dissipate heat promptly under any overload condition.

[0107] For example, when it is predicted that thermal runaway will occur in a portable transformer during future operation, the heat generated by the portable transformer under a preset overload rate for a preset heat dissipation time is first calculated. Q , represented as ,in, To preset the overload rate, This is the rated power of the portable transformer. A preset heat dissipation time is set. For example, the preset overload rate can be set to 150%, and the preset heat dissipation time can be set to 45 minutes to control the temperature rise of the portable transformer to be less than or equal to its temperature threshold within 45 minutes of continuous 150% overload. Further, the amount of heat dissipated by the mixed heat dissipation medium when the temperature of the portable transformer rises to the predicted hot spot temperature within the preset heat dissipation time is calculated; that is, the heat dissipation capacity of the mixed heat dissipation medium. Q 1, represented as ,in, For the flow rate of the mixed heat dissipation medium, The specific heat capacity of the mixed heat dissipation medium, The density of the mixed heat dissipation medium, This represents the temperature difference between the hotspot temperature and the predicted hotspot temperature at the last monitoring time point in the monitoring cycle for the portable transformer. Next, the heat absorbed during the phase change of the phase change material is calculated. Q 2, ,in, For the quality of phase change materials, The latent heat of phase change materials. Therefore, an objective function is constructed with the goal that the sum of the heat dissipation of the mixed heat dissipation medium and the heat absorption of the heat dissipation module should be less than the heat generated by the portable transformer. Solving the objective function yields the target flow rate of the mixed heat dissipation medium, which is then controlled according to this target flow rate for heat dissipation.

[0108] In this step, different heat dissipation levels can also be set. If the predicted hot spot temperature is greater than the temperature threshold of the portable transformer but less than the temperature value corresponding to the first heat dissipation level, the cooling fan in the heat dissipation module can be turned on for forced air cooling. If the predicted hot spot temperature is greater than or equal to the temperature value corresponding to the first heat dissipation level but less than the temperature value corresponding to the second heat dissipation level, then while using forced air cooling, effective heat dissipation is achieved by controlling the flow rate of the mixed heat dissipation medium as described above. For example, the temperature value corresponding to the first heat dissipation level can be 80℃, the temperature value corresponding to the second heat dissipation level can be 130℃, and the temperature threshold of the portable transformer can be 60℃.

[0109] It should be noted that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0110] Furthermore, such as Figure 4 As shown, as a specific implementation of the above-mentioned mobile transformer operation monitoring method, this application embodiment provides a mobile transformer operation monitoring device 400, which includes: a first acquisition unit 401, a first prediction unit 402, a second acquisition unit 403, and a second prediction unit 404.

[0111] The first acquisition unit 401 is used to acquire the ultrasonic signal generated inside the movable transformer during the monitoring period. The first prediction unit 402 is used to perform a fast Fourier transform on the ultrasonic signal to extract the spectral features of the ultrasonic signal; and to input the spectral features into the partial discharge prediction model to obtain the predicted partial discharge result of the mobile transformer at the first future time point corresponding to the monitoring period. The partial discharge prediction model is trained based on historical ultrasonic signals and a long short-term memory network model. The second acquisition unit 403 is used to acquire monitoring data of the mobile transformer within the monitoring period based on the predicted partial discharge results. The monitoring data includes the operation data of the mobile transformer and the external environment data. The second prediction unit 404 is used to input the monitoring data into the temperature prediction model to obtain the predicted hot spot temperature of the portable transformer at the second future time point corresponding to the monitoring cycle. The temperature prediction model is trained based on historical monitoring data and physical information neural network model. Based on the predicted partial discharge results and the predicted hot spot temperature, the portable transformer is subjected to heat dissipation treatment and risk warning.

[0112] Optionally, the portable transformer operation monitoring device 400 also includes: The first training unit is used to acquire historical ultrasonic signals generated inside the portable transformer during the historical monitoring period. Based on the historical partial discharge results of the portable transformer at the first future time point corresponding to the historical monitoring period, the historical ultrasonic signals are labeled, and a first sample set is generated based on the labeled historical ultrasonic signals. The historical partial discharge results include normal and partial discharge. The first sample set is input into the Long Short-Term Memory (LSTM) network model, and the model parameters of the LTM network model are backpropagated and gradients are calculated based on a preset loss function. The model parameters of the LTM network model are iteratively updated using an accelerated gradient adaptive moment estimation algorithm and a variable parameter neurodynamics algorithm. When the change in the preset loss function is less than a preset threshold or the number of iterations is equal to the maximum number of iterations, the LTM network model is output as a partial discharge prediction model.

[0113] Optionally, the first training unit is specifically used to update the learning rate of the current iteration training time short-term memory network model based on a preset convergence rate constant and a preset mapping function, to obtain the learning rate of the next iteration training time short-term memory network model; calculate the first and second moment estimates of the gradient during the current iteration training time short-term memory network model based on the gradient during the current iteration training time short-term memory network model; correct the first and second moment estimates of the gradient based on the correction bias of the first and second moment estimates of the gradient during the current iteration training time short-term memory network model; and update the model parameters of the current iteration training time short-term memory network model based on the learning rate of the next iteration training time short-term memory network model and the corrected first and second moment estimates, to obtain the model parameters of the next iteration training time short-term memory network model.

[0114] Optionally, the portable transformer operation monitoring device 400 also includes: The second training unit is used to construct a second sample set of the portable transformer based on historical and simulated monitoring data. The second sample set includes the first sample monitoring data of the portable transformer at the sample time point and the second sample monitoring data at the second future time point corresponding to the sample time point. The second sample is input into the physical information neural network model, and the target loss function of the physical information neural network model is constructed based on the predicted hotspot temperature of the second sample output by the physical information neural network model. The target loss function is calculated and the model parameters of the physical information neural network model are updated until the target loss function converges. The physical information neural network model is then output as a temperature prediction model.

[0115] The second training unit is specifically used to: construct a data loss function based on the predicted hotspot temperature of the second sample and the sample hotspot temperature in the second sample monitoring data; determine the heating power of the core based on the magnetic induction intensity and voltage frequency of the portable transformer in the second sample monitoring data and the volume of the core of the portable transformer in the first sample monitoring data; determine the heating power of the winding based on the load current of the portable transformer in the second sample monitoring data and the resistance of the winding in the portable transformer at the sample hotspot temperature; determine the heat dissipation power of the portable transformer based on the effective heat dissipation area and ambient temperature of the portable transformer in the first sample monitoring data and the sample hotspot temperature in the second sample monitoring data; determine the total loss power of the portable transformer based on the heating power of the core, the heating power of the winding, and the heat dissipation power of the portable transformer; construct a physical loss function based on the predicted hotspot temperature of the second sample, the total loss power of the portable transformer, and the specific heat capacity and total mass of the portable transformer in the first sample monitoring data; and construct a constraint loss function based on the predicted hotspot temperature of the second sample and the temperature threshold of the portable transformer.

[0116] The second training unit is specifically used to determine the hysteresis loop of the iron core based on the saturation magnetization of the iron core, the characteristic parameters of the nanocrystalline alloy strip, and the magnetic induction intensity of the movable transformer in the second sample monitoring data; to perform loop integration on the hysteresis loop of the iron core to determine the hysteresis loss of the iron core; and to update the heating power of the iron core based on the hysteresis loss of the iron core.

[0117] The second prediction unit 404 is specifically used to determine the heat generation of the portable transformer if the predicted hot spot temperature is greater than the temperature threshold of the portable transformer, based on the preset overload rate, the rated power of the portable transformer, and the preset heat dissipation time; determine the heat dissipation of the mixed heat dissipation medium based on the flow rate of the mixed heat dissipation medium, monitoring data, and the predicted hot spot temperature; calculate the heat absorption of the phase change material in the heat dissipation module due to phase change; construct an objective function with the goal that the sum of the heat dissipation of the mixed heat dissipation medium and the heat absorption of the heat dissipation module is less than the heat generation of the portable transformer; solve the objective function to obtain the target flow rate of the mixed heat dissipation medium, and control the flow of the mixed heat dissipation medium according to the target flow rate for heat dissipation treatment.

[0118] Specific limitations regarding the mobile transformer operation monitoring device can be found in the limitations of the mobile transformer operation monitoring method described above, and will not be repeated here. Each module in the aforementioned mobile transformer operation monitoring device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0119] Based on the above, Figures 1 to 2 Accordingly, embodiments of this application also provide a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. Figures 1 to 2 The method for monitoring the operation of a portable transformer is shown.

[0120] Based on this understanding, the technical solution of this application can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, or portable hard drive), and includes several instructions to cause a computer device (such as a personal computer, server, or network device) to execute the methods described in the various implementation scenarios of this application.

[0121] Based on the above, Figures 1 to 2 The method shown, and Figure 4 To achieve the above objectives, the present application also provides a computer device, specifically a personal computer, server, network device, etc., as shown in the virtual device embodiment. This computer device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to achieve the above-described objectives. Figures 1 to 2 The method for monitoring the operation of a portable transformer is shown.

[0122] Optionally, the computer device may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB ports, card reader ports, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Bluetooth interfaces, Wi-Fi interfaces), etc.

[0123] Those skilled in the art will understand that the computer device structure provided in this embodiment does not constitute a limitation on the computer device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0124] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages and stores the hardware and software resources of a computer device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software within the physical device.

[0125] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platform, or the embodiments of this application can be implemented by hardware.

[0126] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application. Those skilled in the art will understand that the modules in the apparatus of the embodiment can be distributed within the apparatus of the embodiment as described, or can be modified to be located in one or more apparatuses different from this embodiment. The modules of the above-described embodiment can be combined into one module, or further divided into multiple sub-modules.

[0127] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of any particular implementation scenario. The above disclosures are merely a few specific implementation scenarios of this application; however, this application is not limited thereto, and any variations conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A method for monitoring the operation of a portable transformer, characterized in that, The method includes: Acquire ultrasonic signals generated inside the portable transformer during the monitoring period; Perform a Fast Fourier Transform on the ultrasonic signal to extract its spectral features; The spectral features are input into the partial discharge prediction model to obtain the predicted partial discharge result of the mobile transformer at the first future time point corresponding to the monitoring period. The partial discharge prediction model is trained based on historical ultrasonic signals and a long short-term memory network model. Based on the predicted partial discharge results, the monitoring data of the mobile transformer during the monitoring period is obtained. The monitoring data includes the operating data of the mobile transformer and the external environment data. The monitoring data is input into the temperature prediction model to obtain the predicted hot spot temperature of the portable transformer at the second future time point corresponding to the monitoring cycle. The temperature prediction model is trained based on historical monitoring data and a physical information neural network model. Based on the predicted partial discharge results and the predicted hot spot temperature, heat dissipation treatment and risk warning are carried out on the portable transformer.

2. The mobile transformer operation monitoring method according to claim 1, characterized in that, The method further includes: Acquire historical ultrasonic signals generated inside the movable transformer during the historical monitoring period; Based on the historical partial discharge results of the portable transformer at the first future time point corresponding to the historical monitoring cycle, the historical ultrasonic signals are labeled, and a first sample set is generated based on the labeled historical ultrasonic signals. The historical partial discharge results include normal and partial discharge. The first sample set is input into the long short-term memory network model, and the model parameters of the long short-term memory network model are backpropagated and gradients are calculated based on a preset loss function. The model parameters of the long short-term memory network model are iteratively updated using the accelerated gradient adaptive moment estimation algorithm and the variable parameter neurodynamics algorithm. When the change in the preset loss function is less than the preset threshold or the number of iterations is equal to the maximum number of iterations, the long short-term memory network model is output as a partial discharge prediction model.

3. The mobile transformer operation monitoring method according to claim 2, characterized in that, The iterative update of the model parameters of the Long Short-Term Memory network model using the accelerated gradient adaptive moment estimation algorithm and the variable parametric neurodynamics algorithm specifically includes: Based on a preset convergence rate constant and a preset mapping function, the learning rate of the long short-term memory network model during the current iteration of training is updated to obtain the learning rate of the long short-term memory network model during the next iteration of training. Based on the gradient of the Long Short-Term Memory network model during the current iteration of training, calculate the first and second moment estimates of the gradient during the current iteration of training; Based on the correction bias of the first and second moment estimates of the gradient during the current iteration of training, the first and second moment estimates of the gradient are corrected; Based on the learning rate of the Long Short-Term Memory (LSTM) network model and the corrected first and second moment estimates during the next iteration of training, the model parameters of the LSM network model during the current iteration of training are updated to obtain the model parameters of the LSM network model during the next iteration of training.

4. The mobile transformer operation monitoring method according to claim 1, characterized in that, The method further includes: A second sample set of the mobile transformer is constructed based on the historical monitoring data and simulated monitoring data of the mobile transformer. The second sample in the second sample set includes the first sample monitoring data of the mobile transformer at the sample time point, and the second sample monitoring data at the second future time point corresponding to the sample time point. The second sample is input into the physical information neural network model, and the target loss function of the physical information neural network model is constructed based on the predicted hotspot temperature of the second sample output by the physical information neural network model. The target loss function is calculated, and the model parameters of the physical information neural network model are updated until the target loss function converges. The physical information neural network model is then output as a temperature prediction model.

5. The mobile transformer operation monitoring method according to claim 4, characterized in that, The target loss function includes a data loss function, a physical loss function, and a constraint loss function. The construction of the target loss function for the physical information neural network model based on the predicted hotspot temperature of the second sample output by the physical information neural network model specifically includes: The data loss function is constructed based on the predicted hotspot temperature of the second sample and the sample hotspot temperature in the monitoring data of the second sample. The heating power of the core is determined based on the magnetic induction intensity and voltage frequency of the portable transformer in the second sample monitoring data and the volume of the core inside the portable transformer in the first sample monitoring data. Based on the load current of the portable transformer and the resistance of the winding in the portable transformer at the hot spot temperature of the sample in the second sample monitoring data, the heating power of the winding is determined; The heat dissipation power of the portable transformer is determined based on the effective heat dissipation area and ambient temperature of the portable transformer in the first sample monitoring data and the sample hot spot temperature in the second sample monitoring data. The total power loss of the portable transformer is determined based on the heating power of the iron core, the heating power of the winding, and the heat dissipation power of the portable transformer. The physical loss function is constructed based on the predicted hot spot temperature of the second sample and the total power loss of the portable transformer, as well as the specific heat capacity and total mass of the portable transformer in the monitoring data of the first sample. The constraint loss function is constructed based on the predicted hotspot temperature of the second sample and the temperature threshold of the movable transformer.

6. The mobile transformer operation monitoring method according to claim 5, characterized in that, The core adopts an orthogonal laminated structure of silicon steel sheets and nanocrystalline alloy strips, and the method further includes: The hysteresis loop of the iron core is determined based on the saturation magnetization of the iron core, the characteristic parameters of the nanocrystalline alloy strip, and the magnetic induction intensity of the movable transformer in the second sample monitoring data. The hysteresis loss of the iron core is determined by performing a loop integral on the hysteresis loop of the iron core. The heating power of the iron core is updated based on the hysteresis loss of the iron core.

7. The mobile transformer operation monitoring method according to claim 1, characterized in that, The portable transformer includes a transformer module and a heat dissipation module disposed outside the transformer module. The transformer module includes an oil tank and a transformer body immersed in the oil tank. The oil tank contains a mixed heat dissipation medium, the flow rate of which can be controlled. The heat dissipation treatment of the portable transformer based on the predicted hot spot temperature includes: If the predicted hot spot temperature is greater than the temperature threshold of the portable transformer, the heat generation of the portable transformer is determined according to the preset overload rate, the rated power of the portable transformer, and the preset heat dissipation time. The heat dissipation of the mixed heat dissipation medium is determined based on the flow rate of the mixed heat dissipation medium, the monitoring data, and the predicted hot spot temperature. Calculate the heat absorbed by the phase change material in the heat dissipation module during the phase change process; An objective function is constructed with the goal that the sum of the heat dissipation of the mixed heat dissipation medium and the heat absorption of the heat dissipation module is less than the heat generation of the portable transformer. Solve the objective function to obtain the target flow rate of the mixed heat dissipation medium, and control the flow of the mixed heat dissipation medium according to the target flow rate for heat dissipation treatment.

8. A portable transformer operation monitoring device, characterized in that, The device includes: The first acquisition unit is used to acquire ultrasonic signals generated inside the movable transformer during the monitoring period; The first prediction unit is used to perform a fast Fourier transform on the ultrasonic signal to extract the spectral features of the ultrasonic signal; and to input the spectral features into a partial discharge prediction model to obtain the predicted partial discharge result of the portable transformer at the first future time point corresponding to the monitoring period. The partial discharge prediction model is trained based on historical ultrasonic signals and a long short-term memory network model. The second acquisition unit is used to acquire monitoring data of the mobile transformer within the monitoring period based on the predicted partial discharge result. The monitoring data includes the operation data of the mobile transformer and the external environment data. The second prediction unit is used to input the monitoring data into a temperature prediction model to obtain the predicted hot spot temperature of the portable transformer at the second future time point corresponding to the monitoring cycle. The temperature prediction model is trained based on historical monitoring data and a physical information neural network model. Based on the predicted partial discharge result and the predicted hot spot temperature, the unit performs heat dissipation treatment and risk warning for the portable transformer.

9. A readable storage medium having a program or instructions stored thereon, characterized in that, When the program or instructions are executed by the processor, they implement the steps of the mobile transformer operation monitoring method as described in any one of claims 1 to 7.

10. A computer device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the program, it implements the mobile transformer operation monitoring method as described in any one of claims 1 to 7.

Images