Method, device and medium for predicting residual overload running time of transformer

Abstract

The embodiment of the application provides a kind of transformer's residual overload running time prediction method, device equipment and medium.The method comprises: obtaining the measured basic parameter signal of transformer;Based on measured basic parameter signal, using pre-trained transformer hot spot temperature prediction model for prediction processing, obtain the predicted transformer hot spot temperature of transformer current period;Based on predicted transformer hot spot temperature, obtain the comprehensive aging factor of transformer current period;Based on the comprehensive aging factor of current period and the comprehensive aging factor of multiple historical periods, calculate the residual overload running time of transformer current.The method is used to improve the prediction accuracy of the overload running time of transformer, to improve the safety and reliability of transformer operation.

Description

Technical Field

[0001] This application relates to the field of power grids, and more particularly to a method, apparatus, and medium for predicting the remaining overload operating time of a transformer. Background Technology

[0002] In the context of the new power system, with the large-scale grid connection of renewable energy sources such as wind power and photovoltaics, and the rapid popularization of new loads such as electric vehicles and data centers, the operating environment of power transformers faces unprecedented complexity. Against this backdrop, transformer overload operation has become the norm, posing unprecedented challenges to the safe and reliable operation of power transformers.

[0003] To ensure the safe operation of transformers, it is usually necessary to effectively predict the remaining overload operating time. Traditional prediction methods mainly rely on static models and empirical formulas, and their accuracy still needs improvement. For example, the IEC 60076-7 standard describes transient thermal processes through differential equations, but does not fully consider the dynamic effects of oil flow hysteresis; the IEEE C57.91 standard optimizes the hot spot temperature calculation method and introduces a dynamic load correction factor, but its model is still based on a single time scale characteristic and is difficult to cope with multi-source load fluctuations.

[0004] In view of this, there is an urgent need for a more accurate prediction technology for the remaining overload operating time of transformers, in order to further ensure the safe and reliable operation of transformers in power systems. Summary of the Invention

[0005] This application provides a method, device, and medium for predicting the remaining overload operating time of a transformer, in order to improve the accuracy of predicting the remaining overload operating time of the transformer, thereby improving the safety and reliability of transformer operation.

[0006] In a first aspect, embodiments of this application provide a method for predicting the remaining overload operating time of a transformer, comprising:

[0007] Acquire the measured basic parameter signals of the transformer, which include the load current signal, top oil temperature signal, ambient temperature signal, oil flow rate signal, and the number of coolers in operation for the current period of the transformer.

[0008] Based on the measured basic parameter signals, a pre-trained transformer hotspot temperature prediction model is used for prediction processing to obtain the predicted transformer hotspot temperature for the current time period; the transformer hotspot temperature prediction model is obtained by training an initial model based on a tensor spatiotemporal residual network.

[0009] Based on the predicted transformer hot spot temperature, the comprehensive aging factor of the transformer for the current period is obtained; the comprehensive aging factor is used to characterize the degree of aging and loss of the transformer per unit time.

[0010] Based on the comprehensive aging factor of the current period and the comprehensive aging factor of multiple historical periods, the remaining overload operating time of the transformer is calculated.

[0011] In one possible implementation, obtaining the comprehensive aging factor of the transformer for the current period based on the predicted transformer hotspot temperature includes:

[0012] Based on the predicted transformer hot spot temperature and the duration of the current period, calculate the temperature aging factor of the transformer for the current period;

[0013] Calculate the time aging factor of the transformer for the current time period based on the duration of the current time period;

[0014] Based on preset factor weights, the temperature aging factor and the time aging factor are weighted and summed to obtain the comprehensive aging factor of the transformer for the current period.

[0015] In one possible implementation, calculating the transformer's current remaining overload operating time based on the comprehensive aging factor of the current time period and the comprehensive aging factor of multiple historical time periods includes:

[0016] The product of the comprehensive aging factor for different time periods and the duration of the corresponding time period is summed, and the summation result is compared with the service time of the transformer to obtain the cumulative life loss of the transformer; the different time periods include the current time period and the multiple historical time periods;

[0017] Based on the cumulative life loss, the current remaining overload operating time of the transformer is obtained.

[0018] In one possible implementation, the step of using a pre-trained transformer hotspot temperature prediction model to perform prediction processing based on the measured basic parameter signal to obtain the predicted transformer hotspot temperature for the current time period includes:

[0019] For each of the load current signal, the top oil temperature signal, the ambient temperature signal, and the oil flow velocity signal, multi-scale feature extraction is performed on the signal to obtain the reconstructed signal and trend term of the signal;

[0020] The reconstructed signal and trend term of each of the load current signal, the top oil temperature signal, the ambient temperature signal, and the oil flow velocity signal, as well as the number of coolers put into operation, are input into the transformer hot spot temperature prediction model for prediction processing. The transformer hot spot temperature output by the model is obtained as the predicted transformer hot spot temperature for the current period.

[0021] In one possible implementation, the method further includes:

[0022] The initial model is constructed based on the tensor-space-time residual network;

[0023] The initial model is trained based on the pre-built sample set to obtain the transformer hot spot temperature prediction model;

[0024] The sample set is constructed based on historically collected load current signals, top-layer oil temperature signals, ambient temperature signals, oil flow velocity signals, the number of coolers in operation, and actual transformer hot spot temperatures.

[0025] In one possible implementation, training the initial model based on a pre-built sample set to obtain the transformer hotspot temperature prediction model includes:

[0026] Based on the pre-built sample set, the initial model is trained using the AdamW optimizer to obtain the transformer hotspot temperature prediction model.

[0027] In one possible implementation, obtaining the current remaining overload operating time of the transformer based on the cumulative life loss includes:

[0028] Based on the cumulative life loss, the current remaining overload operating time of the transformer is obtained using a preset formula for calculating the remaining overload operating time.

[0029] The formula for calculating the remaining overload operating time is as follows:

[0030]

[0031] In the formula, This indicates the remaining overload operating time; This indicates the cumulative lifespan loss; This represents the thermal capacity coefficient of the transformer; This indicates the predicted hot spot temperature of the transformer; This indicates the upper limit of the hot spot temperature of the transformer; This indicates the service life of the transformer; This indicates the preset safety threshold for transformer life loss.

[0032] Secondly, embodiments of this application provide a device for predicting the remaining overload operating time of a transformer, comprising:

[0033] The acquisition module is used to acquire the measured basic parameter signals of the transformer. The measured basic parameter signals include the load current signal, top oil temperature signal, ambient temperature signal, oil flow rate signal, and the number of coolers in operation for the current period of the transformer.

[0034] The prediction module is used to perform prediction processing based on the measured basic parameter signals and a pre-trained transformer hotspot temperature prediction model to obtain the predicted transformer hotspot temperature for the current period. The transformer hotspot temperature prediction model is obtained by training an initial model based on a tensor spatiotemporal residual network.

[0035] The processing module is used to obtain the comprehensive aging factor of the transformer for the current period based on the predicted transformer hot spot temperature; the comprehensive aging factor is used to characterize the degree of aging loss of the transformer per unit time.

[0036] The calculation module is used to calculate the remaining overload operating time of the transformer based on the comprehensive aging factor of the current time period and the comprehensive aging factor of multiple historical time periods.

[0037] In one possible implementation, the processing module is specifically used for:

[0038] Based on the predicted transformer hot spot temperature and the duration of the current period, calculate the temperature aging factor of the transformer for the current period;

[0039] Calculate the time aging factor of the transformer for the current time period based on the duration of the current time period;

[0040] Based on preset factor weights, the temperature aging factor and the time aging factor are weighted and summed to obtain the comprehensive aging factor of the transformer for the current period.

[0041] In one possible implementation, the computing module includes:

[0042] The first calculation unit is used to sum the product between the comprehensive aging factor of different time periods and the duration of the corresponding time period, and compare the summation result with the service time of the transformer to obtain the cumulative life loss of the transformer; the different time periods include the current time period and the multiple historical time periods;

[0043] The second calculation unit is used to obtain the current remaining overload operating time of the transformer based on the cumulative life loss.

[0044] In one possible implementation, the prediction module is specifically used for:

[0045] For each of the load current signal, the top oil temperature signal, the ambient temperature signal, and the oil flow velocity signal, multi-scale feature extraction is performed on the signal to obtain the reconstructed signal and trend term of the signal;

[0046] The reconstructed signal and trend term of each of the load current signal, the top oil temperature signal, the ambient temperature signal, and the oil flow velocity signal, as well as the number of coolers put into operation, are input into the transformer hot spot temperature prediction model for prediction processing. The transformer hot spot temperature output by the model is obtained as the predicted transformer hot spot temperature for the current period.

[0047] In one possible implementation, the device further includes:

[0048] The training module includes:

[0049] A construction unit is used to construct the initial model based on a tensor-space-time residual network;

[0050] The training unit is used to train the initial model based on a pre-built sample set to obtain the transformer hotspot temperature prediction model.

[0051] The sample set is constructed based on historically collected load current signals, top-layer oil temperature signals, ambient temperature signals, oil flow velocity signals, the number of coolers in operation, and actual transformer hot spot temperatures.

[0052] In one possible implementation, the training unit is specifically used for:

[0053] Based on the pre-built sample set, the initial model is trained using the AdamW optimizer to obtain the transformer hotspot temperature prediction model.

[0054] In one possible implementation, the second computing unit is specifically used for:

[0055] Based on the cumulative life loss, the current remaining overload operating time of the transformer is obtained using a preset formula for calculating the remaining overload operating time.

[0056] The formula for calculating the remaining overload operating time is as follows:

[0057]

[0058] In the formula, This indicates the remaining overload operating time; This indicates the cumulative lifespan loss; This represents the thermal capacity coefficient of the transformer; This indicates the predicted hot spot temperature of the transformer; This indicates the upper limit of the hot spot temperature of the transformer; This indicates the service life of the transformer; This indicates the preset safety threshold for transformer life loss.

[0059] Thirdly, embodiments of this application provide a computer device, including: a memory and a processor;

[0060] The memory stores computer-executed instructions;

[0061] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0062] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0063] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0064] The method, apparatus, and medium for predicting the remaining overload operating time of a transformer provided in this application first acquire multi-dimensional measured basic parameter signals of the transformer. Based on these signals, a transformer hotspot temperature prediction model, pre-trained from an initial model constructed using a tensor spatiotemporal residual network, is used to dynamically capture the spatiotemporal dependencies of parameters such as load current, oil temperature, and oil flow velocity. This achieves high-accuracy prediction of the transformer hotspot temperature for the current period. Then, based on the predicted hotspot temperature, a comprehensive aging factor for the current period is obtained. Using this comprehensive aging factor, along with comprehensive aging factors from multiple historical periods, the remaining overload operating time of the transformer is calculated. This method effectively predicts the overload operating time of the transformer, providing a basis for transformer life assessment, operation and maintenance strategies, and economic operation of the power grid. Attached Figure Description

[0065] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0066] Figure 1 A flowchart illustrating a method for predicting the remaining overload operating time of a transformer, provided in Embodiment 1 of this application;

[0067] Figure 2 This is a flowchart illustrating a method for predicting the remaining overload operating time of a transformer, as provided in Embodiment 2 of this application.

[0068] Figure 3 A schematic diagram of the architecture of a transformer hotspot temperature prediction model;

[0069] Figure 4 A schematic diagram of the structure of a device for predicting the remaining overload operating time of a transformer, provided in Embodiment 3 of this application;

[0070] Figure 5 A schematic diagram of the structure of a device for predicting the remaining overload operating time of a transformer, provided in Embodiment 4 of this application;

[0071] Figure 6 A schematic diagram of the structure of the computer device provided in this application.

[0072] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0073] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0074] To facilitate understanding of the technical content of this solution, the background technology is described in detail below:

[0075] Traditional methods for predicting the remaining overload operating time of transformers mainly rely on static models and empirical formulas, and the accuracy of these predictions still needs to be improved. For example, the new IEC 60076-7 standard has added a description of transient thermal processes and introduced a differential equation model that considers oil flow hysteresis; the IEEE C57.91 standard has improved the hot spot temperature calculation method and added a correction coefficient under dynamic load conditions; the new national standard GB / T 1094.7 (2023) has for the first time included relevant provisions for artificial intelligence-assisted evaluation, such as recommending the integration of three types of data sources: real-time monitoring data (oil temperature, load, etc.), historical operation and maintenance records, and equipment parameters; prioritizing the use of interpretable algorithms such as random forests and GBDT; and requiring model retraining at least every quarter. Some researchers have also proposed distributed improved thermal circuit models or introduced real-time interactive calculation of oil flow velocity field and temperature field to improve the accuracy of transformer dynamic thermal evaluation and thus effectively predict the remaining overload operating time of transformers. In addition, technologies such as fluorescent fiber optic temperature measurement and wireless sensor networks have also been used for distributed measurement of transformer internal temperature, significantly improving the evaluation accuracy of transformer overload capacity.

[0076] However, even with the efforts made by existing technologies, the current prediction results still deviate significantly from the actual state of the transformer. Therefore, there is an urgent need for a more accurate prediction technology for the remaining overload operating time of transformers based on dynamic thermal assessment, so as to promote the development of transformer overload operation technology towards a more intelligent and reliable direction, thereby further ensuring the safe and reliable operation of transformers in the power system.

[0077] Based on the above background technology, the inventors discovered during their research that by combining real-time collected parameters such as transformer load current, top layer temperature, and ambient temperature with a prediction model based on tensor spatiotemporal modeling, the spatiotemporal dependence of parameters such as load current, oil temperature, and oil flow velocity can be dynamically captured, thereby achieving accurate prediction of transformer winding hot spot temperature. Then, based on the high-accuracy transformer winding hot spot temperature, the degree of aging loss of the transformer in the current stage per unit time is assessed, and combined with the degree of aging loss of the transformer in each unit time in historical stages, the remaining overload operating time of the transformer can be quantitatively calculated, thus achieving accurate prediction of the remaining overload operating time of the transformer.

[0078] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0079] Figure 1This is a flowchart illustrating a method for predicting the remaining overload operating time of a transformer, as provided in Embodiment 1 of this application. Figure 1 As shown, the method includes:

[0080] S101. Obtain the measured basic parameter signals of the transformer.

[0081] The measured basic parameter signals include the transformer's load current signal, top oil temperature signal, ambient temperature signal, oil flow velocity signal, and the number of coolers in operation during the current period.

[0082] In this step, sensors designed for different signals to be acquired can be used to collect various measured basic parameters of the transformer in real time according to a preset acquisition frequency, thereby obtaining the measured basic parameter signals of the transformer. The acquisition frequency, for example, is taken from 1Hz to 10kHz. In practical applications, data can be collected at the transformer side using sensors and transmitted via a high-speed bus to the data processing unit of an electronic device (such as a computer) to execute the method for predicting the remaining overload operating time of the transformer provided in this application at the device end.

[0083] It should be understood that for any measured basic parameter, the corresponding signal contains the instantaneous values ​​of that physical parameter at multiple points in time, either continuous or discrete. In practical applications, the measured basic parameters collected for each type of transformer can be saved according to a preset time interval, for example, at a storage interval of 15 minutes per group. Accordingly, each measured basic parameter signal includes multiple instantaneous values ​​of the measured basic parameter stored in sequence at 15-minute intervals during the current stage (such as within 1 day or 1 week).

[0084] Specifically, load current refers to the operating current of the high-voltage or low-voltage winding of the transformer, used to reflect the current load of the transformer; top oil temperature refers to the real-time temperature of the insulating oil at the top of the transformer tank, used to indirectly reflect the heating trend of the winding; ambient temperature refers to the ambient temperature of the transformer installation location, used to assess the transformer's heat dissipation capacity; oil flow velocity refers to the flow velocity of the insulating oil in the transformer cooling circuit (such as the oil pipes of a forced oil circulation air-cooled system), used to assess the transformer's heat dissipation capacity; and the number of coolers in operation refers to the actual number of coolers currently in operation in the transformer, used to assess the transformer's heat dissipation capacity.

[0085] As a specific example, for a 110kV transformer, a current transformer can be used to collect the transformer's load current; a temperature sensor embedded in the top layer of the oil tank can be used to collect the transformer's top oil temperature; a temperature sensor installed near the transformer can be used to collect the transformer's ambient temperature; an ultrasonic flow meter can be used to collect the transformer's oil flow velocity; and a current sensor can be used to collect the current in the cooler circuit. If the current value is greater than zero, it indicates that the cooler is in use, thus obtaining the number of coolers in operation. The sampling frequency of each sensor is, for example, 1kHz for the current transformer, and for example, 1Hz for the temperature sensor and the ultrasonic flow meter.

[0086] S102. Based on the measured basic parameter signals, a pre-trained transformer hot spot temperature prediction model is used for prediction processing to obtain the predicted transformer hot spot temperature for the current period.

[0087] The transformer hotspot temperature prediction model was obtained by training an initial model based on a tensor spatiotemporal residual network.

[0088] In this step, the acquired various measured basic parameter signals are input into the transformer hot spot temperature prediction model so that the model can jointly predict the hot spot temperature of the transformer in the current stage from the perspectives of time and space. The output of the model is used as the predicted hot spot temperature of the transformer in the current period.

[0089] S103. Based on the predicted transformer hot spot temperature, obtain the comprehensive aging factor of the transformer for the current period. The comprehensive aging factor is used to characterize the degree of aging loss of the transformer per unit time.

[0090] In this step, considering that the degree of aging loss of the transformer is mainly driven by thermal aging, the degree of aging loss of the transformer in the current stage (such as within 1 day or 1 week) can be evaluated based on the predicted transformer hot spot temperature, so as to obtain the comprehensive aging factor of the transformer in the current period.

[0091] In one possible implementation, to further improve the accuracy of predicting the remaining overload runtime, this step can be implemented using steps 3.1 to 3.3 as follows:

[0092] Step 3.1: Calculate the temperature aging factor of the transformer for the current period based on the predicted transformer hot spot temperature and the duration of the current period.

[0093] In this step, it is necessary to evaluate the degree of aging loss of the transformer per unit time from the perspective of the transformer hot spot temperature, and obtain the temperature aging factor of the transformer in the current period.

[0094] As a specific example, the temperature aging factor in the i-th time period The following formula can be used to calculate it:

[0095]

[0096] in, This represents the duration of the i-th time period; This indicates the predicted hot spot temperature of the transformer.

[0097] Step 3.2: Calculate the time aging factor of the transformer for the current time period based on the duration of the current time period.

[0098] In this step, the degree of aging and loss of the transformer per unit time will be evaluated from the perspective of the transformer's service life to obtain the time aging factor of the transformer in the current period.

[0099] As a specific example, the time aging factor of the i-th time period The following formula can be used to calculate it:

[0100]

[0101] in, The aging shape factor represents the transformer and is determined by the inherent characteristics of the transformer's insulation material composition, manufacturing process, impregnation treatment, etc. For example, it is suitable for materials whose aging process is slow in the early stage and accelerates in the later stage (such as oil paper insulation). For example, 1.5.

[0102] Step 3.3: Based on the preset factor weights, the temperature aging factor and the time aging factor are weighted and summed to obtain the comprehensive aging factor of the transformer in the current period.

[0103] As a specific example, the comprehensive aging factor for the i-th time period can be calculated using the following formula. :

[0104]

[0105] in, This indicates the preset weight of the time aging factor. This indicates the preset temperature aging factor weight; in practical applications, and It can be determined based on the actual application situation, for example It is 0.3. It is 0.7.

[0106] The method provided in this implementation constructs a time-temperature dual-factor coupled model. By weighted fusion of the temperature aging factor and the time aging factor, it quantifies the cumulative damage of overload operation to the transformer life from the perspective of thermal stress and time accumulation. This makes the comprehensive aging factor more accurately reflect the degree of aging loss of the transformer per unit time, and thus makes the remaining overload operation time calculated based on the comprehensive aging factor more accurate.

[0107] S104. Based on the comprehensive aging factor of the current period and the comprehensive aging factor of multiple historical periods, calculate the remaining overload operating time of the transformer.

[0108] In this step, the contribution of the current period to the aging loss of transformer oil will be calculated based on the comprehensive aging factor of the current period. The total cumulative aging loss (i.e., cumulative life loss) of the transformer from the time the transformer was put into use to the current period will be calculated based on the comprehensive aging factor of multiple historical periods. Then, the remaining overload operating time of the transformer can be calculated based on the cumulative life loss and the allowable aging loss of the transformer.

[0109] In one possible implementation, this step can be achieved using steps 4.1 to 4.2 as follows:

[0110] Step 4.1: Sum the products of the comprehensive aging factor for different time periods and the corresponding duration of the time periods, and compare the summation result with the service time of the transformer to obtain the cumulative life loss of the transformer.

[0111] The different time periods include the current time period and multiple historical time periods.

[0112] Specifically, this step can be represented by the following formula:

[0113]

[0114] in, This indicates the service life of the transformer (i.e., the operating time of the transformer from when it was put into use until the current moment, which is equivalent to the combination of the time sequence of multiple historical periods and the current period). This indicates the number of segments into which the transformer's service life is divided; This indicates the cumulative lifespan loss of the transformer.

[0115] Step 4.2: Based on the cumulative life loss, obtain the current remaining overload operating time of the transformer.

[0116] In this step, the remaining overload operating time of the transformer will be calculated based on the cumulative life loss and the transformer's allowable aging loss.

[0117] As a specific example, the current remaining overload operating time of the transformer can be obtained based on the cumulative lifespan loss using a preset formula for calculating the remaining overload operating time. The formula for calculating the remaining overload operating time is as follows:

[0118]

[0119] In the formula, Indicates the remaining overload operating time of the transformer; This represents the thermal capacity coefficient of a transformer; This indicates the predicted hot spot temperature of the transformer; Indicates the upper limit of the transformer's hot spot temperature; This represents the preset safe threshold for transformer lifespan loss. Among them, The specific requirements can be determined based on the actual transformer operation needs in the application of this solution, and this application does not impose any specific restrictions on this.

[0120] It should be understood that in the above formula, if This indicates that the transformer's cumulative lifespan loss has exceeded its limit; therefore, its remaining overload operating time is determined to be 0. Then it will be based on the remaining life capacity of the transformer insulation. Quantification of thermal safety redundancy Cumulative life loss and overload duration correction item Calculate the remaining overload operating time that satisfies the dual constraints of not exceeding the lifespan limit and not exceeding the temperature limit.

[0121] The formula provided in this step calculates the remaining overload operating time by comparing the relationship between cumulative lifetime loss and the cumulative lifetime loss safety threshold. When the cumulative lifetime loss has not reached the cumulative lifetime loss safety threshold, it integrates core constraints such as remaining lifetime capacity, thermal safety redundancy, and overload duration. This ensures that the calculation of the remaining overload operating time takes into account both the remaining redundancy of insulation lifetime and the safety margin of hot spot temperature. This avoids the danger of continuing overload after the lifetime is exhausted, and when the lifetime redundancy is sufficient, it can combine the transformer's thermal buffering capacity and the overload duration to dynamically adapt the operating time. Ultimately, it maximizes the transformer's overload operating flexibility under safety constraints.

[0122] The method provided in this implementation transforms the current predicted transformer hot spot temperature into a comprehensive aging factor that characterizes the degree of aging loss per unit time in the current period, and integrates the comprehensive aging factors of the current period with those of multiple historical periods to calculate the remaining overload operating time. This effectively quantifies the cumulative life loss of the transformer and ensures the accuracy of the calculation of the remaining overload operating time.

[0123] The method for predicting the remaining overload operating time of a transformer provided in this application first acquires multi-dimensional measured basic parameter signals of the transformer. Based on these measured basic parameter signals, a transformer hotspot temperature prediction model, pre-trained from an initial model constructed based on a tensor spatiotemporal residual network, is used for prediction processing. This dynamically captures the spatiotemporal dependencies of parameters such as load current, oil temperature, and oil flow velocity, achieving high-accuracy prediction of the transformer hotspot temperature for the current period. Then, based on the predicted transformer hotspot temperature, a comprehensive aging factor for the transformer for the current period is obtained. Based on the comprehensive aging factor for the current period and the comprehensive aging factors for multiple historical periods, the remaining overload operating time of the transformer is calculated. This method effectively predicts the overload operating time of the transformer, thereby providing a basis for transformer life assessment, operation and maintenance strategies, and economic operation of the power grid.

[0124] Figure 2 A flowchart illustrating a method for predicting the remaining overload operating time of a transformer, as provided in this application embodiment, is shown below. Figure 2 As shown, based on the above embodiments, this application also provides a method for obtaining a transformer hotspot temperature prediction model, including:

[0125] S201. Construct an initial model based on tensor spatiotemporal residual network.

[0126] Specifically, a tensor-space-time residual network should include at least a tensor-space-time feature learning module and a residual network. More specifically, the tensor-space-time residual tensor structure is used to effectively capture the spatial correlation (i.e., the mutual influence between different parameters) and temporal correlation (i.e., the changing trend of the same parameter at different time points) of the measured basic parameter signals; the jump connection design of the residual network is used to solve the gradient vanishing problem during deep network training, ensuring the stability of model training.

[0127] It should be understood that the initial model built on this architecture can lay the core network structure foundation for subsequent model training and accurate prediction of hotspot temperatures.

[0128] S202. Based on the pre-built sample set, train the initial model to obtain the transformer hotspot temperature prediction model.

[0129] The sample set is constructed based on historically collected load current signals, top-level oil temperature signals, ambient temperature signals, oil flow velocity signals, the number of coolers in operation, and actual transformer hot spot temperatures.

[0130] In this step, the model is optimized and trained based on the predicted transformer hotspot temperature output by the model for each sample and the actual transformer hotspot temperature corresponding to the sample, so that the model output approaches the actual value.

[0131] In one possible implementation, each sample in the sample set includes a reconstructed signal and trend term corresponding to each of the historically acquired load current signal, top-level oil temperature signal, ambient temperature signal, oil flow velocity signal, and actual transformer hot spot temperature, as well as the number of coolers put into operation.

[0132] Correspondingly, Figure 3 A schematic diagram of the transformer hotspot temperature prediction model provided for this implementation is shown below. Figure 3 As shown, the architecture includes an input layer, a spatiotemporal feature extraction module, a feature dimensionality reduction module, a bidirectional temporal coding layer, a residual network layer, and an output layer.

[0133] In detail, the input layer consists of two parts. The first part is used to input a third-order tensor with dimensions (B1×W1×F1), where B1 is the batch size, which can be set according to the number of training samples, W1 is the time window, and F2 is the feature dimension. The second part is used to input a third-order tensor with dimensions (B2×W2×F2), where B2 is the batch size, W2 is the time window, and F2 is the feature dimension. It should be understood that in practical applications, B1 and B2, as well as W1 and W2, are generally the same.

[0134] It should be noted that the first part is used to input the first training sample, and the second part is used to input the second training sample. The first training sample includes the reconstructed signal corresponding to each of the historically collected load current signal, top oil temperature signal, ambient temperature signal, oil flow velocity signal, and actual transformer hot spot temperature signal, as well as the number of coolers put into operation. The second training sample includes the trend term corresponding to each of the historically collected load current signal, top oil temperature signal, ambient temperature signal, oil flow velocity signal, and actual transformer hot spot temperature signal.

[0135] The spatiotemporal feature extraction module receives training sample data in tensor form from the input layer, performs convolutions in both the temporal and spatial dimensions, and enhances feature extraction capabilities based on temporal attention mechanisms (relative position encoding) and channel attention mechanisms (feature cross-layer). It then outputs a cross-modal feature tensor to the feature dimensionality reduction module. Specifically, the number of cooler inputs is encoded using one-hot encoding, which uses N bits to represent N categories, with each category corresponding to a unique position.

[0136] The feature dimensionality reduction module is used to perform Tucker decomposition on cross-modal feature tensors. Tucker decomposition refers to the preprocessing of cross-modal feature tensors by centering (subtracting the mean), expanding the cross-modal feature tensors into matrix form and performing truncated singular value decomposition on them, and using alternating least squares method to iteratively optimize and obtain factor matrices. Based on this, the core tensor is calculated using the projected factor matrix, and finally the core tensor is output to the bidirectional temporal coding layer.

[0137] The formula for calculating the truncated singular value decomposition is as follows:

[0138]

[0139] In the above formula, Represents the cross-modal feature tensor; n represents the number of modes; Represents the factor matrix; represents the right singular value matrix; T represents the transpose.

[0140] The formula for calculating the core tensor is:

[0141]

[0142] In the above formula, Represents the product modulo n; Represents the core tensor; This represents the original cross-modal high-dimensional feature tensor.

[0143] The bidirectional temporal coding layer includes a forward gated recurrent unit (GRU) and a backward GRU, used for bidirectional temporal coding to obtain bidirectional features, which are then input into the residual network layer. Specifically, the forward GRU is used to process the forward sequence. The backward GRU performs time-step reverse sorting and encoding on the input sequence, denoted as... Then, bidirectional features are obtained by fusing them through a fully connected layer and a sigmoid function. The bidirectional feature is represented as:

[0144]

[0145] The residual network layer is used to fuse the bidirectional features corresponding to the two inputs to obtain the fused features, and then input the fused features into the output layer.

[0146] The output layer performs a non-linear transformation on the fused features using the Tanh activation function to map the features to a reasonable numerical range. Then, the numerical range is scaled and mapped to the actual range of transformer hotspot temperatures to obtain the predicted transformer hotspot temperature. And output it.

[0147] It should be understood that during the training phase, after constructing the initial model described above, the training parameters (such as network weights, temporal features, and regularization parameters) will be initialized first. Then, the training samples will be input into the input layer to obtain the predicted transformer hotspot temperature output by the output layer. Next, the model parameters will be iteratively optimized based on the predicted transformer hotspot temperature and the actual transformer hotspot temperature in the samples. After training is completed, the transformer hotspot temperature prediction model will be obtained.

[0148] Optionally, the initial model can be trained using the AdamW optimizer based on a pre-built sample set to obtain a transformer hotspot temperature prediction model.

[0149] The loss function of the AdamW optimizer is expressed as:

[0150]

[0151] in, The value represents the actual hot spot temperature of the transformer; λ represents the regularization coefficient. This represents the set of trainable parameters of the model, including network weights, etc.

[0152] It should be understood that using the AdamW optimizer for model training has two advantages. First, it can dynamically adjust the learning rate according to the update requirements of different modules in the dual-path spatiotemporal residual network, adapting to the learning logic of the model's complex multi-layer structure and multimodal features. This allows deep parameters to iterate efficiently, significantly accelerating the convergence speed in the early stages of training and avoiding the convergence oscillation problem that is prone to occur with fixed learning rate optimizers. Second, its improved weight decay mechanism can directly apply precise regularization constraints to the parameters, avoiding the model from overfitting local fluctuations in the training data. Ultimately, this ensures both stable and efficient convergence during the training phase and improves the model's generalization ability to new data from actual transformer operation.

[0153] In practical applications, an optional incremental training strategy can be used for model training. Specifically, the network branch with the reconstructed signal as input is trained first, then the network branch with the trend term as input is added, and then the complete model training is performed. During the training process, the grid search method is used to fine-tune the three hyperparameters: learning rate, batch size, and number of residual blocks.

[0154] In one possible implementation, when using the above-mentioned transformer hotspot temperature prediction model for prediction processing, Example 1 S102 can be implemented using the following steps 2.1 to 2.2:

[0155] Step 2.1: For each of the load current signal, top oil temperature signal, ambient temperature signal, and oil flow velocity signal, perform multi-scale feature extraction to obtain the reconstructed signal and trend term.

[0156] The following uses the transformer top oil temperature signal For example, this section illustrates the multi-scale feature extraction of the top-layer oil temperature signal and the resulting reconstructed signal. and trend items

[0157] The method includes the following steps a to f:

[0158] Step a: Take the derivative of the top oil temperature signal to obtain the time series of the first derivative of the top oil temperature. ;

[0159] Step b: Based on the top oil temperature signal Time series of the first derivative of top oil temperature By determining the sign of the product of two adjacent points, we sequentially search for all local maxima and minima of the top oil temperature, and then apply cubic spline interpolation to calculate the upper envelope of the transformer's top oil temperature. and lower envelope .

[0160] Step c: Calculate the frequency modulation signal of the top oil temperature signal. The calculation formula is:

[0161]

[0162]

[0163]

[0164] in, This represents the envelope estimation function. If... It is not a pure frequency modulation signal, that is, it does not satisfy... Then Repeat steps a to c as the initial signal until... It is a pure frequency modulation signal (denoted as ) ).

[0165] Step d: Multiply the product of all envelope estimation functions generated in the above iteration process with the pure modulation signal to obtain the first top oil temperature component of the top oil temperature signal. The calculation formula is:

[0166]

[0167] in, Let represent the envelope estimation function obtained in the i-th iteration; This indicates the number of repetitions in step c.

[0168] Step e: From the top oil temperature signal Separate the oil in the middle layer to obtain the remaining top layer oil temperature signal. and will Steps a to e are repeated using the initial top-layer oil temperature signal until the energy of the remaining transformer top-layer oil temperature is less than a preset energy threshold, thereby decomposing all top-layer oil temperature components. The preset energy threshold is, for example, the original top-layer oil temperature signal. 5% of the energy.

[0169] After steps a through e above, the initial transformer top oil temperature... It has been decomposed into the sum of several top-layer oil temperature components and the residual signal, and its expression is:

[0170]

[0171] in, This represents the remaining top-layer oil temperature signal obtained during the k-th decomposition iteration; Let i represent the i-th top-layer oil temperature component, i∈[1,k-1], and k represent the total number of decomposition iterations.

[0172] Step f: Calculate the reconstructed signal and trend term of the top oil temperature using the following formulas:

[0173]

[0174] Therefore, the reconstructed signal and trend term of the top oil temperature signal are calculated. For the transformer load current signal, ambient temperature signal, and oil flow velocity signal, multi-scale feature extraction can be performed using the methods described in steps a to f above to obtain the corresponding reconstructed signal and trend term. The specific multi-scale extraction methods for each signal will not be elaborated here.

[0175] Step 2.2: Input the reconstructed signal and trend term of each of the load current signal, top oil temperature signal, ambient temperature signal and oil flow velocity signal, as well as the number of coolers put into operation, into the transformer hot spot temperature prediction model for prediction processing, and obtain the transformer hot spot temperature output by the model as the predicted transformer hot spot temperature for the current period.

[0176] In this step, the reconstructed signals and trend terms of each of the load current signal, top oil temperature signal, ambient temperature signal, and oil flow velocity signal need to be input into the input layer of the model, and the predicted transformer hot spot temperature output by the model in the output layer is used as the predicted transformer hot spot temperature for the current period.

[0177] The method provided in this implementation effectively separates the multi-scale features in the original signal by decomposing it into multiple scales, accurately quantifying the long-term trend and short-term fluctuations in the time series. Then, the reconstructed signal and trend term obtained from the decomposition are used as input to the prediction model to obtain the predicted transformer hot spot temperature. This allows the model to independently capture the spatiotemporal features of the long-term trend term and short-term fluctuation components in the original signal, avoiding the problem of insufficient feature learning caused by mutual interference between features of different scales. At the same time, it allows the model to form a more accurate feature representation of the temperature influencing factors under different operating conditions of the transformer, thereby significantly improving the accuracy and real-time performance of hot spot temperature prediction, and enhancing the model's robustness in prediction under complex operating conditions such as overload and sudden environmental changes.

[0178] The method for predicting the remaining overload operating time of a transformer provided in this application constructs an initial model based on a tensor spatiotemporal residual network and trains the initial model according to a pre-constructed sample set to obtain a transformer hotspot temperature prediction model. This allows the model to dynamically capture the spatiotemporal dependencies of multi-dimensional parameters and accelerate model convergence by relying on the residual network layer, thereby mitigating the gradient minimization problem and effectively ensuring the prediction accuracy and training cost of the model.

[0179] Figure 4 This is a schematic diagram of the structure of a device for predicting the remaining overload operating time of a transformer, as provided in Embodiment 3 of this application. Figure 4 As shown, the transformer remaining overload operating time prediction device 30 provided in this embodiment includes:

[0180] The acquisition module 301 is used to acquire the measured basic parameter signals of the transformer. The measured basic parameter signals include the load current signal, top oil temperature signal, ambient temperature signal, oil flow rate signal, and the number of coolers in operation during the current period of the transformer.

[0181] The prediction module 302 is used to perform prediction processing based on the measured basic parameter signals and a pre-trained transformer hot spot temperature prediction model to obtain the predicted transformer hot spot temperature for the current period. The transformer hot spot temperature prediction model is obtained by training an initial model based on a tensor spatiotemporal residual network.

[0182] The processing module 303 is used to obtain the comprehensive aging factor of the transformer for the current period based on the predicted transformer hot spot temperature; the comprehensive aging factor is used to characterize the degree of aging loss of the transformer per unit time.

[0183] The calculation module 304 is used to calculate the remaining overload operating time of the transformer based on the comprehensive aging factor of the current time period and the comprehensive aging factor of multiple historical time periods.

[0184] The transformer remaining overload operating time prediction device 30 provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0185] Figure 5 This is a schematic diagram of the structure of a device for predicting the remaining overload operating time of a transformer, as provided in Embodiment 4 of this application. Figure 5 As shown, based on the above embodiments, the transformer remaining overload operation time prediction device 30 provided in this embodiment further includes:

[0186] Training module 305 includes:

[0187] The building unit is used to construct the initial model based on the tensor-space-time residual network;

[0188] The training unit is used to train the initial model based on a pre-built sample set to obtain a transformer hotspot temperature prediction model.

[0189] The sample set is constructed based on historically collected load current signals, top-level oil temperature signals, ambient temperature signals, oil flow velocity signals, the number of coolers in operation, and actual transformer hot spot temperatures.

[0190] In one possible implementation, the processing module 303 is specifically used for:

[0191] Calculate the temperature aging factor of the transformer for the current period based on the predicted transformer hot spot temperature and the duration of the current period.

[0192] Calculate the time aging factor of the transformer for the current time period based on the duration of the current time period;

[0193] Based on preset factor weights, the temperature aging factor and the time aging factor are weighted and summed to obtain the comprehensive aging factor of the transformer in the current period.

[0194] In one possible implementation, the computing module 304 includes:

[0195] The first calculation unit is used to sum the product between the comprehensive aging factor of different time periods and the duration of the corresponding time period, and compare the summation result with the service time of the transformer to obtain the cumulative life loss of the transformer; different time periods include the current time period and multiple historical time periods;

[0196] The second calculation unit is used to obtain the current remaining overload operating time of the transformer based on the cumulative life loss.

[0197] In one possible implementation, the prediction module 302 is specifically used for:

[0198] For each of the load current signal, top oil temperature signal, ambient temperature signal, and oil flow velocity signal, multi-scale feature extraction is performed to obtain the reconstructed signal and trend term.

[0199] The reconstructed signals and trend terms of each of the load current signal, top oil temperature signal, ambient temperature signal, and oil flow velocity signal, as well as the number of coolers in operation, are input into the transformer hot spot temperature prediction model for prediction processing. The transformer hot spot temperature output by the model is then obtained as the predicted transformer hot spot temperature for the current period.

[0200] In one possible implementation, the training unit is specifically used for:

[0201] Based on the pre-built sample set, the initial model is trained using the AdamW optimizer to obtain the transformer hot spot temperature prediction model.

[0202] In one possible implementation, the second computing unit is specifically used for:

[0203] Based on the cumulative life loss, the current remaining overload operating time of the transformer is obtained using a preset formula for calculating the remaining overload operating time.

[0204] The formula for calculating the remaining overload running time is as follows:

[0205]

[0206] In the formula, Indicates the remaining overload running time; Indicates cumulative lifespan loss; This represents the thermal capacity coefficient of a transformer; This indicates the predicted hot spot temperature of the transformer; Indicates the upper limit of the transformer's hot spot temperature; Indicates the service life of the transformer; This indicates the preset safety threshold for transformer life loss.

[0207] The transformer remaining overload operating time prediction device 30 provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0208] Figure 6 A schematic diagram of the structure of the computer device provided in this application. Figure 6 As shown, the electronic device 40 provided in this embodiment includes at least one processor 401 and a memory 402. Optionally, the device 40 further includes a communication component 403. The processor 401, memory 402, and communication component 403 are connected via a bus 404.

[0209] In a specific implementation, at least one processor 401 executes computer execution instructions stored in memory 402, causing at least one processor 401 to perform the above-described method.

[0210] The specific implementation process of processor 401 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0211] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0212] The memory may include read-only memory and random access memory. The memory may be volatile or non-volatile, or may include both. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which serves as an external cache. Many forms of RAM are available by way of example, but not limitation. Examples include Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Sync Link DRAM (SLDRAM), and Direct Rambus RAM (DR RAM).

[0213] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0214] This application also provides a computer program product, including a computer program that, when executed, implements the above-described method.

[0215] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0216] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as SRAM, EEPROM, EPROM, PROM, ROM, magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0217] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside within an ASIC. Alternatively, the processor and the readable storage medium can exist as discrete components in a device.

[0218] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0219] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0220] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0221] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0222] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0223] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for predicting the remaining overload operating time of a transformer, characterized in that, include: Acquire the measured basic parameter signals of the transformer, which include the load current signal, top oil temperature signal, ambient temperature signal, oil flow rate signal, and the number of coolers in operation for the current period of the transformer. Based on the measured basic parameter signals, a pre-trained transformer hotspot temperature prediction model is used for prediction processing to obtain the predicted transformer hotspot temperature for the current time period; the transformer hotspot temperature prediction model is obtained by training an initial model based on a tensor spatiotemporal residual network. Based on the predicted transformer hot spot temperature, the comprehensive aging factor of the transformer for the current period is obtained; The comprehensive aging factor is used to characterize the degree of aging loss of the transformer per unit time. Based on the comprehensive aging factor of the current period and the comprehensive aging factor of multiple historical periods, the remaining overload operating time of the transformer is calculated.

2. The method according to claim 1, characterized in that, The step of obtaining the comprehensive aging factor of the transformer for the current period based on the predicted transformer hotspot temperature includes: Based on the predicted transformer hot spot temperature and the duration of the current period, calculate the temperature aging factor of the transformer for the current period; Calculate the time aging factor of the transformer for the current time period based on the duration of the current time period; Based on preset factor weights, the temperature aging factor and the time aging factor are weighted and summed to obtain the comprehensive aging factor of the transformer for the current period.

3. The method according to claim 1 or 2, characterized in that, The calculation of the transformer's current remaining overload operating time based on the comprehensive aging factor of the current time period and the comprehensive aging factor of multiple historical time periods includes: The product of the comprehensive aging factor for different time periods and the duration of the corresponding time period is summed, and the summation result is compared with the service time of the transformer to obtain the cumulative life loss of the transformer; the different time periods include the current time period and the multiple historical time periods; Based on the cumulative life loss, the current remaining overload operating time of the transformer is obtained.

4. The method according to claim 1 or 2, characterized in that, The process of using a pre-trained transformer hotspot temperature prediction model to predict the transformer hotspot temperature for the current time period, based on the measured basic parameter signals, includes: For each of the load current signal, the top oil temperature signal, the ambient temperature signal, and the oil flow velocity signal, multi-scale feature extraction is performed on the signal to obtain the reconstructed signal and trend term of the signal; The reconstructed signal and trend term of each of the load current signal, the top oil temperature signal, the ambient temperature signal, and the oil flow velocity signal, as well as the number of coolers put into operation, are input into the transformer hot spot temperature prediction model for prediction processing. The transformer hot spot temperature output by the model is obtained as the predicted transformer hot spot temperature for the current period.

5. The method according to claim 1 or 2, characterized in that, The method further includes: The initial model is constructed based on the tensor-space-time residual network; The initial model is trained based on the pre-built sample set to obtain the transformer hot spot temperature prediction model; The sample set is constructed based on historically collected load current signals, top-layer oil temperature signals, ambient temperature signals, oil flow velocity signals, the number of coolers in operation, and actual transformer hot spot temperatures.

6. The method according to claim 5, characterized in that, The step of training the initial model based on a pre-built sample set to obtain the transformer hotspot temperature prediction model includes: Based on the pre-built sample set, the initial model is trained using the AdamW optimizer to obtain the transformer hotspot temperature prediction model.

7. The method according to claim 3, characterized in that, The step of obtaining the current remaining overload operating time of the transformer based on the cumulative life loss includes: Based on the cumulative life loss, the current remaining overload operating time of the transformer is obtained using a preset formula for calculating the remaining overload operating time. The formula for calculating the remaining overload operating time is as follows: In the formula, This indicates the remaining overload operating time; This indicates the cumulative lifespan loss; This represents the thermal capacity coefficient of the transformer; This indicates the predicted hot spot temperature of the transformer; This indicates the upper limit of the hot spot temperature of the transformer; This indicates the service life of the transformer; This indicates the preset safety threshold for transformer life loss.

8. A device for predicting the remaining overload operating time of a transformer, characterized in that, The device includes: The acquisition module is used to acquire the measured basic parameter signals of the transformer. The measured basic parameter signals include the load current signal, top oil temperature signal, ambient temperature signal, oil flow rate signal, and the number of coolers in operation for the current period of the transformer. The prediction module is used to perform prediction processing based on the measured basic parameter signals and a pre-trained transformer hotspot temperature prediction model to obtain the predicted transformer hotspot temperature for the current period. The transformer hotspot temperature prediction model is obtained by training an initial model based on a tensor spatiotemporal residual network. The processing module is used to obtain the comprehensive aging factor of the transformer for the current period based on the predicted transformer hot spot temperature; the comprehensive aging factor is used to characterize the degree of aging loss of the transformer per unit time. The calculation module is used to calculate the remaining overload operating time of the transformer based on the comprehensive aging factor of the current time period and the comprehensive aging factor of multiple historical time periods.

9. A computer device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

Images