A method and device for predicting the temperature of a ring main unit, electronic equipment and storage medium

Abstract

The application provides a kind of ring main unit temperature prediction method, device, electronic equipment and storage medium, the method comprises: obtaining the operating temperature value of multiple target positions of ring main unit, load current value, environmental temperature value and operating condition identifier, and carries out feature construction, obtains time series lag feature, contact deterioration indirect representation feature, thermal dynamic characteristic and representation heat dissipation capacity and heat dissipation condition and environmental feature;Based on the time series lag feature, contact deterioration indirect representation feature, thermal dynamic characteristic and the multi-dimensional feature vector of heat dissipation condition and environmental feature constructed by feature and preset time series prediction model, obtain the temperature variation prediction value of future preset time step, and generate early warning signal based on the temperature variation prediction value.The application realizes high-precision prediction and reliable early warning of ring main unit temperature by fusing multi-dimensional feature construction and time series prediction model of ring main unit thermodynamics physical mechanism.

Description

Technical Field

[0001] This application relates to the field of ring main unit technology, and more specifically, to a method, apparatus, electronic device, and storage medium for predicting the temperature of a ring main unit. Background Technology

[0002] Ring main units (RMMs) are critical equipment in power distribution networks. During long-term operation, their internal electrical connection points are prone to abnormal temperature increases due to factors such as load current, contact resistance, and heat dissipation. Failure to detect and address these issues promptly can lead to insulation aging, equipment damage, and even accidents. Therefore, accurate prediction and timely warning of RMM temperatures are essential for ensuring the safe operation of power distribution networks.

[0003] Currently, the common method for monitoring temperature in ring main units is to collect temperature data in real time using sensors and set fixed thresholds for alarms. When the monitored temperature value exceeds the preset threshold, the system issues an alarm signal. However, this method is a reactive mechanism, lacking predictive capabilities and unable to provide early warnings in the early stages of a temperature upward trend.

[0004] To overcome the shortcomings of fixed threshold alarms, existing technologies have proposed machine learning-based methods for predicting ring main unit (RMU) temperatures. For example, patent application number 202511823740.X discloses a training method for a RMU temperature prediction model. This method acquires a time-series operational dataset of the RMU, performs differential preprocessing on the data, constructs a feature set to characterize the dynamic temperature change patterns of the RMU, trains a random forest regression model based on the training set as the temperature prediction model, fine-tunes the parameters using a validation set, and finally confirms the training is complete after verifying that the prediction accuracy meets the requirements using a test set. This method, to a certain extent, achieves a shift from passive alarm to proactive early warning.

[0005] However, the aforementioned existing technologies still have the following shortcomings: First, they use random forest regression models for temperature prediction. Random forests are essentially static models and cannot effectively capture the time dependence and thermal hysteresis effect of temperature changes, thus limiting further improvement in prediction accuracy. Second, their feature engineering is mainly based on general statistical features and simple derived features, failing to fully integrate the thermodynamic physical mechanism of ring main units, especially lacking the ability to characterize the trend of contact resistance degradation, resulting in insufficient early warning accuracy under common fault modes such as poor contact. Summary of the Invention

[0006] In view of this, the purpose of this application is to provide a method, apparatus, electronic device and storage medium for predicting the temperature of a ring main unit, so as to overcome the problems in the prior art.

[0007] In a first aspect, embodiments of this application provide a method for predicting the temperature of a ring main unit, the method comprising: Acquire multi-source operating data of the ring main unit; wherein, the multi-source operating data includes operating temperature values, load current values, ambient temperature values, and operating condition identifiers of multiple target locations; The operating temperature value, load current value, ambient temperature value and operating condition identifier are used to construct features to obtain the time lag feature that characterizes the evolution law of temperature and current, the indirect characterization feature of contact degradation that characterizes the relative change trend of contact resistance, the thermal dynamic feature that characterizes the thermodynamic dynamic response characteristics, and the heat dissipation condition and environmental features that characterize the heat dissipation capacity and environmental impact. Based on the multi-dimensional feature vector constructed from the time lag characteristics, indirect characterization characteristics of contact degradation, thermal dynamic characteristics, and heat dissipation conditions and environmental characteristics, and the preset time prediction model, the predicted value of temperature change over a preset time step is obtained, and an early warning signal is generated based on the predicted value of temperature change.

[0008] In some technical solutions of this application, the aforementioned timing lag characteristics include temperature lag characteristics, current lag characteristics, and temperature rise lag characteristics, which are constructed in the following manner: Using the current moment as a reference, the system traces back along a preset sampling interval, using the operating temperature values ​​of multiple historical sampling points as the temperature hysteresis feature; using the load current values ​​of multiple historical sampling points as the current hysteresis feature; and using the difference between the operating temperature values ​​of multiple historical sampling points and the operating temperature values ​​of their respective previous sampling points as the temperature rise hysteresis feature.

[0009] In some technical solutions of this application, the aforementioned indirect characterization features of contact degradation include the current-temperature rise ratio anomaly index, temperature rise residual, and temperature difference ratio; the indirect characterization features of contact degradation are constructed in the following manner: The ratio of the temperature rise change to the current change at the current moment is calculated, and the statistical mean of the ratios of temperature rise change to current change at multiple historical moments within the backtracking window is subtracted to obtain the current-temperature rise ratio anomaly index; wherein, the temperature rise change is the difference between the operating temperature value at the current moment and the operating temperature value at the previous sampling moment, and the current change is the difference between the load current value at the current moment and the load current value at the previous sampling moment. The expected temperature rise is calculated based on a simplified thermal model that relies solely on the square term of the load current. The temperature rise residual is obtained by dividing the difference between the actual temperature rise and the expected temperature rise by the rated allowable temperature of the equipment. The actual temperature rise is the difference between the operating temperature value at the current moment and the operating temperature value at the previous sampling moment. The temperature difference ratio is obtained by calculating the ratio of the operating temperature of the target location to the average operating temperature of other similar locations within the same ring main unit.

[0010] In some technical solutions of this application, the aforementioned thermal dynamic characteristics include short-time temperature rise rate, cumulative thermal effect index, and cooling rate characteristics; the thermal dynamic characteristics are constructed in the following manner: Based on the operating temperature value, the difference between the current operating temperature value and the operating temperature value at the previous sampling time is divided by the sampling interval to obtain the short-time temperature rise rate. Based on the load current value, the squared values ​​of the load current at each sampling time within the preset backtracking window are summed by exponential decay to obtain the cumulative heat effect index. When the load current is detected to be in a decreasing phase, the linear fitting slope of the operating temperature values ​​of a preset number of sampling points is calculated based on the operating temperature value to obtain the cooling rate characteristic.

[0011] In some technical solutions of this application, the aforementioned heat dissipation conditions and environmental characteristics include the temperature difference characteristics inside and outside the cabinet, the coding characteristics of the installation environment, and the coding characteristics of the cabinet's location. These heat dissipation conditions and environmental characteristics are constructed in the following manner: Based on the ambient temperature value, the difference between the reference temperature inside the cabinet and the ambient temperature value is calculated to obtain the temperature difference characteristics inside and outside the cabinet; The installation environment type in the operating condition identifier is encoded to obtain the installation environment encoding feature; The cabinet location type in the operating condition identifier is encoded to obtain the cabinet location encoding feature.

[0012] In some technical solutions of this application, the aforementioned preset temporal prediction model is a hybrid model of temporal convolutional network and long short-term memory network; the output layer of the hybrid model includes a first output branch and a second output branch, the first output branch is used to output the predicted value of temperature change in the future preset time step, and the second output branch is used to output the probability value of the temperature exceeding the preset safety threshold in the future preset time step.

[0013] In some technical solutions of this application, the generation of an early warning signal from the predicted temperature change includes: Based on the current operating temperature value and the predicted temperature change value, calculate the predicted temperature value for a future preset time step. Based on the predicted temperature value and the probability value, a joint mapping is performed in a preset two-dimensional judgment matrix to generate a graded early warning signal; When the graded early warning signal meets the preset triggering conditions, the contribution value of each input feature currently predicted is analyzed based on the model interpretation tool to identify the dominant cause features, and a diagnostic report containing fault cause classification is generated according to the type of the dominant cause features.

[0014] Secondly, embodiments of this application provide a device for predicting the temperature of a ring main unit, the device comprising: The data acquisition module is used to acquire multi-source operating data of the ring main unit; wherein, the multi-source operating data includes operating temperature values, load current values, ambient temperature values ​​and operating condition identifiers of multiple target locations; The feature construction module is used to construct features from the operating temperature value, load current value, ambient temperature value and operating condition identifier to obtain time lag features that characterize the evolution of temperature and current, indirect contact degradation features that characterize the relative change trend of contact resistance, thermodynamic features that characterize the thermodynamic dynamic response characteristics, and heat dissipation conditions and environmental features that characterize heat dissipation capacity and environmental impact. The prediction module is used to obtain the predicted value of temperature change over a preset time step based on the multi-dimensional feature vector constructed from the time lag characteristics, indirect characterization characteristics of contact degradation, thermal dynamic characteristics, and heat dissipation conditions and environmental characteristics, and a preset time prediction model, and to generate an early warning signal based on the predicted value of temperature change.

[0015] Thirdly, embodiments of this application provide an electronic device, a processor, a memory, and a bus. The memory stores machine instructions executed by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine instructions are executed by the processor, the steps of the above-described method for predicting the temperature of a ring main unit are performed.

[0016] Fourthly, embodiments of this application provide a computer storage medium storing a computer program, which, when run by a processor, executes the steps of the above-described method for predicting the temperature of a ring main unit.

[0017] The technical solutions provided by the embodiments of this application may include the following beneficial effects: This application provides a method for predicting the temperature of a ring main unit (RNB). The method includes: acquiring multi-source operating data of the RNB; wherein the multi-source operating data includes operating temperature values, load current values, ambient temperature values, and operating condition identifiers of multiple target locations; constructing features from the operating temperature values, load current values, ambient temperature values, and operating condition identifiers to obtain time-series lag features characterizing the evolution of temperature and current, indirect contact degradation features characterizing the relative change trend of contact resistance, thermal dynamic features characterizing the thermodynamic dynamic response characteristics, and heat dissipation conditions and environmental features characterizing heat dissipation capacity and environmental impact; based on the multi-dimensional feature vector constructed from the time-series lag features, indirect contact degradation features, thermal dynamic features, and heat dissipation conditions and environmental features, and a preset time-series prediction model, obtaining a predicted value of temperature change over a preset time step, and generating an early warning signal based on the predicted value of temperature change.

[0018] This application acquires multi-source operating data of the ring main unit and constructs features from operating temperature, load current, ambient temperature, and operating condition indicators. This results in time-lag features characterizing the evolution of temperature and current, indirect contact degradation features characterizing the relative change trend of contact resistance, thermal dynamic features characterizing the thermodynamic dynamic response, and heat dissipation conditions and environmental features characterizing heat dissipation capacity and environmental impact. The constructed features comprehensively describe the temperature change pattern of the ring main unit from multiple physical dimensions, including time-series evolution, contact degradation, thermal dynamic response, and heat dissipation conditions, overcoming the deficiency in existing technologies where feature engineering is disconnected from the thermodynamic physical mechanism of the ring main unit. Based on this, the application obtains predicted values ​​of temperature changes at a preset time step and generates early warning signals based on the multi-dimensional feature vectors and a preset time-series prediction model. By utilizing the time-series prediction model's ability to capture time dependence and the multi-dimensional features' ability to characterize physical mechanisms, high-precision prediction and reliable early warning of the ring main unit's temperature are achieved.

[0019] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart illustrating a method for predicting the temperature of a ring main unit provided in an embodiment of this application is shown. Figure 2 A schematic diagram of a device for predicting the temperature of a ring main unit provided in an embodiment of this application is shown; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0023] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0024] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0025] Ring main units (RMMs) are critical equipment in power distribution networks. During long-term operation, their internal electrical connection points are prone to abnormal temperature increases due to factors such as load current, contact resistance, and heat dissipation. Failure to detect and address these issues promptly can lead to insulation aging, equipment damage, and even accidents. Therefore, accurate prediction and timely warning of RMM temperatures are essential for ensuring the safe operation of power distribution networks.

[0026] Currently, the common method for monitoring temperature in ring main units is to collect temperature data in real time using sensors and set fixed thresholds for alarms. When the monitored temperature value exceeds the preset threshold, the system issues an alarm signal. However, this method is a reactive mechanism, lacking predictive capabilities and unable to provide early warnings in the early stages of a temperature upward trend.

[0027] To overcome the shortcomings of fixed threshold alarms, existing technologies have proposed machine learning-based methods for predicting ring main unit (RMU) temperatures. For example, patent application number 202511823740.X discloses a training method for a RMU temperature prediction model. This method acquires a time-series operational dataset of the RMU, performs differential preprocessing on the data, constructs a feature set to characterize the dynamic temperature change patterns of the RMU, trains a random forest regression model based on the training set as the temperature prediction model, fine-tunes the parameters using a validation set, and finally confirms the training is complete after verifying that the prediction accuracy meets the requirements using a test set. This method, to a certain extent, achieves a shift from passive alarm to proactive early warning.

[0028] However, the aforementioned existing technologies still have the following shortcomings: First, they use random forest regression models for temperature prediction. Random forests are essentially static models and cannot effectively capture the time dependence and thermal hysteresis effect of temperature changes, thus limiting further improvement in prediction accuracy. Second, their feature engineering is mainly based on general statistical features and simple derived features, failing to fully integrate the thermodynamic physical mechanism of ring main units, especially lacking the ability to characterize the trend of contact resistance degradation, resulting in insufficient early warning accuracy under common fault modes such as poor contact.

[0029] Based on this, the present application provides a method, apparatus, electronic device and storage medium for predicting the temperature of a ring main unit, which will be described below through embodiments.

[0030] Figure 1 The diagram illustrates a flowchart of a method for predicting the temperature of a ring main unit according to an embodiment of this application, wherein the method includes steps S101-S104; specifically: S101. Obtain multi-source operating data of the ring main unit; wherein, the multi-source operating data includes operating temperature values, load current values, ambient temperature values, and operating condition identifiers of multiple target locations; S102. The operating temperature value, load current value, ambient temperature value and operating condition identifier are used to construct features to obtain the time lag feature that characterizes the evolution law of temperature and current, the indirect characterization feature of contact degradation that characterizes the relative change trend of contact resistance, the thermal dynamic feature that characterizes the thermodynamic dynamic response characteristics, and the heat dissipation condition and environmental features that characterize the heat dissipation capacity and environmental impact. S103. Based on the time lag characteristics, indirect characterization characteristics of contact degradation, thermal dynamic characteristics, and heat dissipation conditions and environmental characteristics, a multi-dimensional feature vector is constructed and a preset time prediction model is used to obtain the predicted value of temperature change over a preset time step in the future, and an early warning signal is generated based on the predicted value of temperature change.

[0031] This application acquires multi-source operating data of the ring main unit and constructs features from operating temperature, load current, ambient temperature, and operating condition indicators. This results in time-lag features characterizing the evolution of temperature and current, indirect contact degradation features characterizing the relative change trend of contact resistance, thermal dynamic features characterizing the thermodynamic dynamic response, and heat dissipation conditions and environmental features characterizing heat dissipation capacity and environmental impact. The constructed features comprehensively describe the temperature change pattern of the ring main unit from multiple physical dimensions, including time-series evolution, contact degradation, thermal dynamic response, and heat dissipation conditions, overcoming the deficiency in existing technologies where feature engineering is disconnected from the thermodynamic physical mechanism of the ring main unit. Based on this, the application obtains predicted values ​​of temperature changes at a preset time step and generates early warning signals based on the multi-dimensional feature vectors and a preset time-series prediction model. By utilizing the time-series prediction model's ability to capture time dependence and the multi-dimensional features' ability to characterize physical mechanisms, high-precision prediction and reliable early warning of the ring main unit's temperature are achieved.

[0032] The following describes some embodiments of this application in detail. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0033] The method provided in this application first requires acquiring multi-source operational data of the ring main unit. Multi-source operational data refers to multiple monitoring data sequences with different data sources and physical meanings, which together form the information basis for subsequent feature construction and temperature prediction. Specifically, this multi-source operational data includes at least the operating temperature values ​​of multiple target locations within the ring main unit, the load current values ​​of the corresponding circuits, the ambient temperature value of the environment in which the unit is located, and operating condition indicators reflecting the installation and deployment conditions of the ring main unit.

[0034] The operating temperature values ​​of the aforementioned target locations refer to real-time temperature measurements collected by temperature sensors deployed at key electrical connection points inside the ring main unit. These key electrical connection points may include, but are not limited to, busbar connections, cable termination joints, and disconnector contacts. These locations, due to their long-term bearing of load current, are concentrated areas of heat generation and thermal failure. In practical applications, the temperature sampling frequency should be able to capture the dynamic trend of temperature changes; for example, a sampling frequency of no less than once per minute can be used.

[0035] It should be noted that this application specifically obtains the operating temperature values ​​of multiple target parts, rather than the temperature data of a single part or a general ring main unit. The purpose is to provide a data foundation for the subsequent construction of indirect characterization features of contact degradation, so as to discover abnormal temperature rises in a specific part by comparing the temperatures of similar parts, thereby more effectively identifying local faults such as contact resistance degradation.

[0036] Load current refers to the real-time current data flowing through the corresponding circuit of the monitored ring main unit, which can be synchronously acquired through devices such as current transformers or smart meters. According to Joule's law, the heat generated when current flows through a conductor is proportional to the square of the current. Therefore, the magnitude and changes of load current are the main driving factors causing temperature changes in the ring main unit. During data acquisition, the load current value should be synchronized with the operating temperature value to ensure that the physical relationship between the two can be accurately reflected in subsequent feature construction.

[0037] Ambient temperature refers to the measured ambient temperature outside or inside the ring main unit, away from the heat source. It can be obtained through an independent ambient temperature sensor. Ambient temperature directly affects the heat dissipation conditions of the ring main unit. When the ambient temperature is high, the temperature difference for heat dissipation decreases, reducing heat dissipation efficiency, and resulting in higher operating temperatures under the same load conditions. Therefore, ambient temperature is an indispensable input data for understanding and predicting temperature change patterns.

[0038] Operating condition identifiers refer to categorical information describing the installation and operation scenarios of ring main units (RNBs). Unlike the continuously sampled numerical data mentioned above, these identifiers are typically pre-entered into the system in the form of tags or codes. Operating condition identifiers include at least the installation environment type and the racking location type. The installation environment type distinguishes the installation scenario of the RNB, such as direct sunlight outdoors, outdoor shade, indoor or basement environments, where ventilation and heat dissipation conditions differ significantly. The racking location type identifies the relative position of the RNB when multiple RNBs are installed side-by-side, such as end racks or middle racks. RNBs in the middle of a rack have significantly weaker heat dissipation capacity than those at the ends because their heat dissipation surfaces are blocked by adjacent racks, making them more prone to temperature increases under the same operating conditions. By using these operating condition identifiers as input data, this application overcomes the shortcomings of existing technologies that rely solely on continuously changing physical quantities such as ambient temperature and lack a structured description of actual heat dissipation conditions. This allows subsequent temperature prediction models to learn differentiated temperature change patterns for different deployment scenarios, avoiding prediction bias caused by ignoring differences in installation conditions.

[0039] After acquiring the aforementioned multi-source operating data, this application further constructs features for the operating temperature value, load current value, ambient temperature value, and operating condition identifier to obtain four types of features that characterize the temperature change law of the ring main unit from different physical dimensions: the time lag feature characterizing the evolution law of temperature and current, the indirect characterization feature characterizing the relative change trend of contact resistance, the thermal dynamic feature characterizing the thermodynamic dynamic response characteristics, and the heat dissipation conditions and environmental characteristics characterizing the heat dissipation capacity and environmental impact.

[0040] Time lag characteristics are used to characterize the evolution of the operating state of a ring main unit (RNB) over time. The temperature change of a RNB exhibits significant thermal inertia; the current temperature depends not only on the current operating conditions but also on the heat accumulation and transfer processes over a past period. Therefore, relying solely on the instantaneous value at the current moment is insufficient to reflect the dynamic trend of temperature change. Time lag characteristics specifically include temperature lag characteristics, current lag characteristics, and temperature rise lag characteristics. These three types of characteristics characterize the evolution of temperature change in the RNB from different physical dimensions, and together they constitute a complete time-series information description. Among these features, temperature lag directly provides snapshots of temperature over multiple past moments, enabling the model to perceive the inherent continuity and trend of temperature. Current lag provides historical information on driving factors directly related to temperature changes. According to Joule's law, load current is the fundamental cause of heat generation, and the magnitude of current in the past determines the historical accumulation of heat. Historical temperature values ​​alone cannot distinguish between two different operating conditions: "high temperature caused by continuous high current" and "current has decreased but heat has not yet dissipated." Temperature rise lag, through the temperature difference between adjacent sampling moments, directly quantifies the magnitude and direction of temperature changes over a short period, explicitly providing the model with information on the acceleration of temperature changes, eliminating the need for the model to indirectly deduce acceleration information from absolute temperature values. If only temperature lag is used without current lag, the model cannot determine the cause of temperature changes and will struggle to distinguish between normal load fluctuations and equipment malfunctions. If only temperature and current lag features are used without temperature rise lag, the model must learn the differential patterns from temperature values ​​during training, increasing the learning difficulty and convergence time. Therefore, the above three types of characteristics together constitute the time lag characteristics, which fully describe the evolution of the ring main unit temperature over time from the three perspectives of state quantity, driving quantity, and change quantity.

[0041] It should be noted that this application uses time-lag features, rather than directly using the raw temperature value sequence of continuous sampling as model input. The reason for this is that directly inputting all historical temperature data into the model would, on the one hand, cause the input dimension to grow uncontrollably with the sampling time, making the model unable to operate stably; on the other hand, temperature values ​​from a long time ago have almost no substantial impact on predicting short-term future temperatures, instead serving as noise that interferes with the model's learning of key trends. More importantly, temperature itself is a slow variable with significant inertia. If the model receives an excessively long historical temperature sequence, it is highly likely to learn an identity mapping that directly copies the current temperature to the predicted value for the next moment. This seemingly high-precision prediction is actually a lagging, spurious prediction, completely losing the ability to anticipate temperature change trends. This application constructs time-series lag features by selecting a few representative historical sampling points. This not only compresses historical sequences of arbitrary length into fixed-dimensional feature vectors to ensure the consistency of model input, but also actively injects the physical knowledge of the thermal inertia of the ring main unit into feature engineering. This enables the model to learn key change pattern information such as "the current temperature is not high, but it has been rising rapidly over a period of time in the past", thereby effectively capturing the time dependence and thermal lag effect of temperature changes.

[0042] As a preferred embodiment, when selecting representative historical sampling points, the principle of "dense recent data and sparse distant data" can be followed to balance the model's sensitivity to short-term dynamic changes and its stable perception of long-term trends. Specifically, for time periods closer to the current moment, smaller sampling intervals can be selected, such as selecting data from the 1st, 3rd, and 6th sampling points respectively. These dense recent sampling points can accurately depict the rate and acceleration of temperature change in the short term. For time periods farther from the current moment, larger sampling intervals can be selected, such as selecting data from the 12th and 24th sampling points respectively. These sparse distant sampling points are used to provide the background of temperature change trends and a benchmark for comparison. The physical basis of this non-equidistant historical sampling point selection method lies in the law of decay of the influence of the ring main unit's thermal inertia over time. That is, the state of historical sampling points closer to the current moment has a greater influence on future temperature, while the state of historical sampling points farther from the current moment has a smaller influence on future temperature. Therefore, by using dense sampling in the near term and sparse sampling in the long term, the model can obtain sufficient temporal resolution to identify rapid temperature changes, while avoiding the introduction of too much redundant long-term data and increasing noise interference, thus achieving a balance between feature dimensions and predictive information.

[0043] Taking temperature hysteresis as an example, the current moment can be used as a benchmark, and the model can be traced back at a preset sampling interval to select the operating temperature values ​​of the 1st, 3rd, 6th, 12th, and 24th sampling points as the specific values ​​of this feature. The construction method for current hysteresis and temperature rise hysteresis features is similar. By selecting multiple representative historical sampling points, the historical sequence of arbitrary length is compressed into a feature vector of fixed dimension to ensure the consistency of the model input. At the same time, the physical knowledge of the thermal inertia of the ring main unit is injected into the feature engineering, enabling the model to effectively capture the time dependence and thermal hysteresis effect of temperature changes.

[0044] Indirect characterization of contact degradation is used to reflect the relative changing trend of contact resistance at electrical connection points inside a ring main unit. Increased contact resistance is one of the most common fault modes leading to abnormal heating in ring main units. However, since contact resistance itself is difficult to measure directly online, it needs to be indirectly characterized by other observable physical quantities.

[0045] The indirect characterization features of contact degradation constructed in this application include the current-temperature rise ratio anomaly index, temperature rise residual, and temperature difference ratio. The current-temperature rise ratio anomaly index determines whether the same current increment causes a larger temperature rise than normal by comparing the ratio of the current change to the current change at the current moment with its historical statistical mean. If so, it suggests that the contact resistance may be deteriorating. The temperature rise residual compares the actual temperature rise with the expected temperature rise calculated by a simplified thermal model that relies only on the square of the load current, and normalizes the difference relative to the equipment's rated allowable temperature. This extracts the abnormal temperature rise component that cannot be explained by current changes, which is also an important indicator of contact resistance degradation. The temperature difference ratio identifies abnormal temperature deviations of a specific location relative to similar locations by calculating the ratio of the operating temperature of the target location to the average operating temperature of other locations with the same electrical function and connection type within the same ring main unit. This ratio should be close to 1.0 under normal operating conditions. If the ratio is consistently significantly greater than 1.0, it indicates that the temperature of that location is abnormally high compared to other similar locations.

[0046] This application selects three types of indicators—current-temperature rise ratio anomaly index, temperature rise residual, and temperature difference ratio—to constitute the indirect characterization features of contact deterioration, rather than using only one type or selecting other indicators. This is based on the analysis of the physical mechanism of contact resistance deterioration faults and full consideration of the complementarity of various indicators.

[0047] Contact resistance degradation is a gradual process, and its manifestation in operational data is multi-dimensional. A single indicator often only captures some abnormal signals under specific operating conditions, making it difficult to comprehensively cover various degradation modes and operating scenarios. The current-temperature rise ratio anomaly index, by longitudinally comparing the temperature rise response changes of the same part at different times, can effectively capture the changes in temperature rise efficiency caused by the gradual increase in contact resistance. Its advantage lies in its sensitivity to degradation signals during current fluctuations. However, under operating conditions where the load current is stable for a long period, since the current change is close to zero, the calculation of this index may be affected by division by zero, resulting in a decrease in reliability.

[0048] The temperature rise residual extracts the abnormal temperature rise component that cannot be explained by current by comparing the actual temperature with the expected temperature calculated solely from the current. Its advantage lies in its independence from current changes; even with a constant load current, as long as the contact resistance is slowly deteriorating, causing an abnormal temperature rise, the residual will continue to increase and signal the problem. This characteristic effectively compensates for the shortcomings of the current-temperature rise ratio anomaly index under steady-state conditions.

[0049] The temperature difference ratio, by comparing the temperature differences of similar parts within the same ring main unit, can eliminate interference from systemic factors such as increased load and rising ambient temperature. When the entire ring main unit heats up due to increased load, the current-temperature rise ratio anomaly index and temperature rise residual may fluctuate abnormally simultaneously. However, at this time, the temperature of each part rises synchronously, and the temperature difference ratio remains at a normal level. This allows differentiation between systemic load changes and local defects in a specific part. Conversely, when a joint becomes loose or oxidized, the temperature of that part will rise independently relative to other similar parts, and the temperature difference ratio will deviate from the normal range, thus enabling precise location of local faults.

[0050] The three types of indicators mentioned above indirectly characterize the degradation trend of contact resistance from three different dimensions: longitudinal time comparison, self-consistency verification, and horizontal spatial comparison. They complement each other functionally, jointly covering different scenarios such as current fluctuation conditions, current stability conditions, and the distinction between systemic and local faults. If only one or two types of indicators are used, monitoring blind spots will appear under certain operating conditions, making it impossible to comprehensively and reliably perceive the degradation trend of contact resistance.

[0051] Thermal dynamic characteristics are used to describe the dynamic response characteristics of ring main units (RNBs) at the thermodynamic level. The temperature rise process of a RNB is not only related to the current magnitude but also closely related to the rate of heat accumulation and the rate of heat dissipation. Specifically, thermal dynamic characteristics include short-time temperature rise rate, cumulative thermal effect index, and cooling rate characteristics. The short-time temperature rise rate is obtained by dividing the difference between the current operating temperature and the previous sampling time by the sampling interval, reflecting the instantaneous speed of temperature change. The cumulative thermal effect index is obtained by exponentially decaying weighted summation of the squared values ​​of the load current at each sampling time within a preset backtracking window. The decay coefficient is set according to the thermal time constant of the RNB, giving a greater weight to the contribution of recent current to heat, while the contribution of long-term current decays exponentially, thus accurately simulating the heat accumulation process in conductors and insulation materials. The cooling rate characteristic is obtained by linearly fitting the operating temperature values ​​of several recent sampling points when the load current is detected to be in a decreasing phase. The slope of the fitted line represents the heat dissipation capacity at the current moment. When the contact resistance is normal, the heat dissipation rate mainly depends on the environmental conditions; however, when the heat dissipation rate slows down abnormally, it may indicate problems such as obstructed ventilation or equipment aging.

[0052] The reason for choosing short-term temperature rise rate, cumulative thermal effect index, and cooling rate characteristics for constructing thermal dynamic characteristics is that they correspond to three independent yet interconnected core aspects of the ring main unit's thermodynamic dynamic process: the rate of temperature rise, heat accumulation, and heat dissipation capacity. The short-term temperature rise rate, calculated by dividing the temperature change between adjacent sampling times by the sampling interval, directly quantifies the rate of temperature change per unit time, reflecting the equipment's instantaneous thermal response to current changes. It is the most direct indicator for determining whether the temperature is in a rapid rise phase. The cumulative thermal effect index, by exponentially decaying weighted summation of the square values ​​of the load current over a past period, simulates the physical process of heat accumulation in conductors and insulation materials over time. The decay coefficient is set so that the contribution weight of recent current to the current thermal state is greater than that of long-term current, consistent with the law of heat dissipation in actual thermodynamics. This index reflects how much heat the equipment has currently stored, thus providing a basis for judging future temperature rise potential. The cooling rate characteristic is used to characterize the current heat dissipation capacity of the equipment by calculating the linear fitting slope of the temperature drop when the load current is decreasing. Its physical meaning is that when the heat generation decreases, the rate of temperature drop directly depends on the quality of heat dissipation conditions. If the heat dissipation channel is blocked or the environmental heat dissipation conditions deteriorate, the cooling rate will slow down significantly.

[0053] The three types of indicators described above comprehensively characterize the thermodynamic dynamic response of the ring main unit from three dimensions: temperature rise, heat storage, and temperature drop. These three indicators complement each other and are indispensable. If only the short-term temperature rise rate is used without the cumulative heat effect index, the model can only perceive the rate of change at the current moment, but cannot determine how much historical heat accumulation this change is based on, easily leading to misjudgments of the severity of the temperature rise. If the cooling rate characteristic is missing, the model cannot distinguish whether "slow temperature drop is due to poor heat dissipation conditions or because the current has not yet truly decreased," thus losing the ability to identify problems related to poor heat dissipation during fault diagnosis. Therefore, these three types of indicators together constitute the thermal dynamic characteristics, providing the prediction model with complete thermodynamic dynamic response information.

[0054] Heat dissipation conditions and environmental characteristics are used to comprehensively describe various external and structural factors affecting the heat dissipation of ring main units (RNBs). The actual operating temperature of a RNB depends not only on its internal heat generation but also, to a large extent, on its heat dissipation conditions. These characteristics include the temperature difference between the inside and outside of the RNB, the installation environment coding characteristics, and the parallel location coding characteristics. The temperature difference between the inside and outside of the RNB is obtained by calculating the difference between the reference temperature inside the RNB and the ambient temperature, reflecting the driving force of the RNB dissipating heat to the surrounding environment at the current moment; the greater the temperature difference, the stronger the heat dissipation. The installation environment coding characteristics are obtained by numerically coding the installation environment type. Different installation environments, such as direct sunlight outdoors, shaded outdoor environments, indoor environments, or basements, have significantly different ventilation conditions and solar radiation effects. These differences are converted into discrete characteristics that the model can recognize through coding. The parallel location coding characteristics are obtained by coding the parallel location type, distinguishing between different locations such as end cabinets and intermediate cabinets. As mentioned earlier, intermediate cabinets have significantly weaker heat dissipation capacity than end cabinets because their heat dissipation surfaces are blocked on both sides, making them more prone to temperature increases under the same operating conditions. By incorporating the aforementioned heat dissipation conditions and environmental characteristics, the prediction model can learn differentiated temperature rise patterns for different installation and deployment scenarios, avoiding prediction bias caused by ignoring differences in heat dissipation conditions, and improving the model's adaptability and prediction accuracy in different field environments.

[0055] The reason for choosing three types of indicators—internal and external temperature difference characteristics, installation environment coding characteristics, and parallel cabinet location coding characteristics—to construct the heat dissipation conditions and environmental features is that they comprehensively describe the heat dissipation state of the ring main unit from three aspects: heat dissipation driving force, external environmental conditions, and its own structural constraints. The internal and external temperature difference characteristic directly quantifies the driving force of the ring main unit dissipating heat to the surrounding environment by calculating the difference between the reference temperature inside the cabinet and the ambient temperature. The larger the temperature difference, the more heat is dissipated per unit time. This characteristic can reflect the activity level of the heat dissipation process in real time. However, it only describes the heat dissipation driving force and cannot explain the differences in heat dissipation effects between different ring main units or the same ring main unit under different installation conditions at the same temperature difference. The installation environment coding feature was introduced to compensate for this deficiency. Installation environment types, such as direct outdoor sunlight, outdoor shade, indoor, or basement, determine the ventilation conditions, airflow speed, and degree of solar radiation around the ring main unit. These factors directly affect heat dissipation efficiency. Under the same temperature difference between inside and outside the unit, the actual heat dissipation rate in a well-ventilated, shaded outdoor environment can differ by several times from that in a poorly ventilated basement environment. Therefore, encoding the installation environment type as a discrete feature recognizable by the model allows the prediction model to adjust its temperature rise expectations based on different environmental conditions. The parallel unit location coding feature is designed for the unique installation structure of ring main units. When multiple units are installed side-by-side, the heat dissipation area of ​​the ring main unit at the ends differs significantly from that in the middle. The heat dissipation surfaces on the left and right sides of the middle unit are blocked by adjacent units, resulting in a much smaller effective heat dissipation area than the end units. Under the same load and environmental conditions, the operating temperature of the middle unit is usually higher. This factor cannot be derived from the temperature difference between inside and outside the unit or the installation environment type and must be described separately through the parallel unit location coding feature. The combined effect of these three types of characteristics enables the predictive model to distinguish whether the high temperature is caused by deteriorating heat dissipation conditions or internal faults. For example, when the temperature difference between the inside and outside of the cabinet is small and the installation environment is a basement, a slightly higher temperature may be normal and does not require triggering an alarm. However, when the temperature difference between the inside and outside of the cabinet is sufficient, the installation environment is good, and it is an end cabinet, a slightly higher temperature is very likely a signal of internal contact degradation or other faults, and a timely warning should be issued. Therefore, the three types of indicators systematically describe the heat dissipation driving force, the external heat dissipation environment, and the internal heat dissipation structure, respectively. They complement each other and are indispensable, providing complete information on heat dissipation conditions for the predictive model to accurately assess the thermal state of the ring main unit.

[0056] After constructing the four types of features mentioned above, the time lag features, indirect contact degradation characteristics, thermal dynamic features, and heat dissipation conditions are concatenated with environmental features in a preset order to form a multi-dimensional feature vector. Each dimension in this multi-dimensional feature vector corresponds to a specific feature index, and the values ​​of all dimensions together describe the complete state information of the ring main unit at the current moment across the four physical dimensions of time evolution, contact state, thermal dynamic response, and heat dissipation conditions. By combining the four types of features into a unified feature vector, the subsequent prediction model can simultaneously consider the coupling relationships between different physical dimensions, rather than analyzing the impact of a single factor on temperature in isolation.

[0057] The multidimensional feature vector is input into a pre-trained offline time-series prediction model, which outputs a predicted value of the temperature change at a preset time step. It should be noted that the model in this application predicts the amount of temperature change rather than the absolute temperature value. This is because temperature is a relatively slow-changing, inertial variable. If the absolute temperature value is used directly as the prediction target, the model may learn an identity mapping during training, directly copying the current temperature to the predicted value for the next moment. While this prediction may show a low numerical error, it is essentially a lagging, spurious prediction that fails to anticipate the temperature change trend. By using the amount of temperature change as the prediction target, the model must learn the mapping relationship between the input features and the direction and magnitude of temperature change, forcing the model to extract deeper patterns related to the temperature rise trend from the multidimensional features, thereby achieving true forward prediction.

[0058] After obtaining the predicted temperature change value, an early warning signal is generated based on this prediction. Specifically, the current operating temperature value is added to the predicted temperature change value to obtain the predicted temperature value for a preset time step in the future. This predicted temperature value is compared with a pre-set safe temperature threshold. If the predicted temperature value exceeds the safe temperature threshold, an early warning signal is generated to alert maintenance personnel that there is an overheating risk in that part within the preset time step in the future, so that inspection and handling can be arranged in advance. Compared with traditional fixed threshold alarms, this application, by predicting the temperature change trend in advance and generating an early warning based on this, advances the time point of fault detection from the moment the temperature actually exceeds the standard to the stage where the temperature shows an upward trend but has not yet exceeded the standard, thus gaining valuable handling time for maintenance personnel and realizing a transformation from passive response to proactive early warning.

[0059] In an optional implementation, when generating an early warning signal, this application does not rely solely on a single predicted temperature value, but instead employs a graded early warning mechanism that jointly determines the predicted temperature value and the probability value of over-temperature. Furthermore, it introduces fault cause attribution analysis based on model interpretation tools to achieve a complete closed loop from early warning to diagnosis.

[0060] First, the current operating temperature is added to the model's predicted temperature change to obtain the predicted temperature for a predetermined time step in the future. The physical meaning of this calculation is that the current operating temperature reflects the current thermal state of the equipment, while the predicted temperature change reflects the model's judgment on future temperature rise or fall trends. Adding the two together gives the expected temperature the equipment will reach in the future. Compared to directly predicting absolute temperature, this "current baseline plus predicted increment" calculation mode effectively avoids the risk of model learning lag mapping.

[0061] Then, the calculated predicted temperature value and the over-temperature probability value output by the second output branch of the model are jointly mapped in a preset two-dimensional judgment matrix to generate a graded early warning signal. The reason for using two-dimensional joint judgment is that a single predicted temperature value or a single over-temperature probability value cannot fully reflect the overall situation of thermal failure risk. If only the predicted temperature value is relied upon for judgment, when the predicted temperature value is slightly lower than the safety threshold but the model has a low confidence in the prediction result, a single threshold judgment may miss potential risks; if only the over-temperature probability value is relied upon for judgment, false alarms may occur in some low-risk situations where the probability value fluctuates occasionally. This application constructs a two-dimensional judgment matrix with the predicted temperature value and the over-temperature probability value as the horizontal and vertical axes, respectively, and divides the early warning signal into multiple levels. For example, when the predicted temperature value is lower than 80% of the safety threshold and the over-temperature probability value is lower than 0.3, it is judged as a normal state; when the predicted temperature value exceeds the safety threshold and the over-temperature probability value exceeds 0.5, it is judged as a warning state; and when the predicted temperature value exceeds 110% of the safety threshold and the over-temperature probability value exceeds 0.7, it is judged as an alarm state. This two-dimensional judgment method only triggers the corresponding level of warning when both the temperature prediction value and the probability confidence level support the risk assessment, thereby effectively reducing the false alarm rate while ensuring the warning coverage rate.

[0062] When the generated graded early warning signal meets the preset triggering conditions, such as reaching the orange or red alert level, this application further utilizes a model interpretation tool to perform attribution analysis on the contribution values ​​of each input feature that led to the prediction, in order to identify one or more dominant inducing features that have the greatest impact on the prediction result. Specifically, the model interpretation tool can employ the SHAP method, which is based on the Shapley value principle in game theory and can fairly allocate a prediction result to each input feature, quantifying the marginal contribution of each feature to the prediction. By extracting one or more features with the largest absolute contribution value, and based on the type of these features, the fault causes are initially classified. For example, when the dominant cause is a characteristic directly related to the load current, such as a load current lag term or a cumulative heat effect index, it can be preliminarily determined to be overload-type overheating. Maintenance personnel are advised to check for unreasonable operating methods or sudden load increases. When the dominant cause is a characteristic related to the contact state, such as an abnormal current-temperature rise ratio index or a temperature rise residual, it can be preliminarily determined to be contact-deterioration-type overheating. The focus should be on checking the tightening torque of joints and the oxidation of contact surfaces in the corresponding areas. When the dominant cause is a characteristic related to heat dissipation conditions, such as the temperature difference between inside and outside the cabinet or the cooling rate, it can be preliminarily determined to be poor heat dissipation-type overheating. The focus should be on checking whether the ventilation channels are unobstructed or whether the installation environment has changed. Finally, the above information, including the warning location, warning level, predicted temperature value, probability of overheating, dominant cause type, and recommended handling measures, is integrated to generate a structured early warning diagnostic report. Through this method, this application not only issues early warning signals but also provides maintenance personnel with preliminary fault cause analysis and targeted handling suggestions, significantly improving fault diagnosis efficiency and the pertinence of emergency response.

[0063] In an optional implementation, this application uses a time-series prediction model for prediction. A time-series prediction model is a type of prediction model specifically designed for processing time-series data and capable of capturing the time dependencies within the data. Unlike static models such as random forests, time-series prediction models, when processing input data, not only consider the feature values ​​at the current moment but also model the evolution of features over time through their internal memory or attention mechanisms. For the specific task of predicting the temperature of a ring main unit, temperature change is essentially a dynamic process with significant thermal inertia and time lag effects. Current changes at the current moment often require a period of energy accumulation before being reflected in the temperature. Therefore, the model must be able to capture such causal relationships across time steps, and the time-series prediction model is a suitable choice to meet this requirement.

[0064] In its implementation, this time-series prediction model can employ a hybrid structure of temporal convolutional networks and long short-term memory networks. The temporal convolutional network slides along the time dimension through one-dimensional convolution operations, extracting local temporal patterns at different time scales, such as rapid temperature rise responses caused by sudden current changes or slow temperature drifts caused by changes in ambient temperature. The long short-term memory network, through its gating mechanism, models long-range dependencies, remembering state information from longer periods, such as the impact of morning load peaks on afternoon temperatures or seasonal temperature baseline drift. The combination of these two approaches can achieve both precise capture of short-term dynamic changes and stable perception of long-term trends. Furthermore, the output layer of this hybrid model can be configured as a multi-task structure, including a first output branch and a second output branch. The first output branch outputs the predicted temperature change over a future preset time step, while the second output branch outputs the probability that the temperature will exceed a preset safety threshold within that preset time step. This allows the model to simultaneously provide point estimates and uncertainty estimates in a single forward inference, providing richer decision-making basis for subsequent graded early warning determinations.

[0065] In one specific embodiment, the preset temporal prediction model adopts a hybrid structure of temporal convolutional network and long short-term memory network, and its specific architecture is as follows.

[0066] The input layer receives a three-dimensional tensor of shape (batch size, time step, feature dimension). The time step is set to 24, which means that the input is a sequence of feature vectors from the current time step and the previous 23 time steps. The feature vector at each time step contains several dimensions, including temporal lag features, indirect contact degradation features, thermal dynamic features, and heat dissipation conditions and environmental features.

[0067] The input sequence first enters the temporal convolutional network module. This module contains two stacked one-dimensional convolutional layers. The first layer uses 64 convolutional kernels with a kernel size of 3 and a stride of 1, employing a rectified linear unit (RCU) activation function. The second layer uses 128 convolutional kernels with a kernel size of 5 and a stride of 1, also employing an RCU activation function. Each convolutional layer is followed by a batch normalization layer and a dropout layer with a dropout rate of 0.2 to prevent overfitting. The output of the temporal convolutional network module is a feature sequence that extracts multi-scale local temporal patterns.

[0068] The output of the temporal convolutional network module then enters the long short-term memory (LSM) network module. This module consists of two stacked LSM layers, each with 128 hidden units, and employs a bidirectional structure, processing the sequence simultaneously from both forward and backward directions in the temporal dimension, ensuring that the output at each time step incorporates contextual information. A dropout layer with a dropout rate of 0.2 is also set between the two LSM layers. The output of the last LSM layer at the last time step is taken as the representation vector of the entire sequence, and this vector has a dimension of 256.

[0069] The representation vector is fed into two parallel fully connected output branches. The first output branch consists of two fully connected layers: the first layer contains 64 neurons with a modified linear unit (MRU) activation function, and the second layer contains one neuron without an activation function. This branch outputs a predicted temperature change over a future preset time step, in degrees Celsius. The second output branch also consists of two fully connected layers: the first layer contains 64 neurons with a MRU activation function, and the second layer contains one neuron using an sigmoid growth curve function as its activation function. This branch outputs a probability value between 0 and 1, representing the probability that the temperature will exceed a preset safety threshold within the future preset time step.

[0070] During the training phase, the model's total loss function consists of a weighted sum of two parts. The first part is the mean squared error loss of the first output branch, which measures the deviation between the predicted and actual temperature change values. The second part is the binary cross-entropy loss of the second output branch, which measures the deviation between the predicted overtemperature probability and the actual overtemperature label. The weight ratio of the two losses can be set to 1:0.5, ensuring that the model prioritizes the accuracy of temperature change prediction while also considering the accuracy of probability estimation during training.

[0071] The above embodiments are only one specific implementation of this application. In practical applications, the number of channels, number of hidden units, dropout rate and other hyperparameters of each layer can be adjusted according to the data scale and computing resources, or other types of time series prediction model structures can be adopted, such as the Transformer model based on self-attention mechanism or the network structure based on gated recurrent units, as long as it has the ability to model and predict time series data.

[0072] Figure 2 This illustration shows a structural diagram of a device for predicting the temperature of a ring main unit according to an embodiment of this application. The device includes: The data acquisition module is used to acquire multi-source operating data of the ring main unit; wherein, the multi-source operating data includes operating temperature values, load current values, ambient temperature values ​​and operating condition identifiers of multiple target locations; The feature construction module is used to construct features from the operating temperature value, load current value, ambient temperature value and operating condition identifier to obtain time lag features that characterize the evolution of temperature and current, indirect contact degradation features that characterize the relative change trend of contact resistance, thermodynamic features that characterize the thermodynamic dynamic response characteristics, and heat dissipation conditions and environmental features that characterize heat dissipation capacity and environmental impact. The prediction module is used to obtain the predicted value of temperature change over a preset time step based on the multi-dimensional feature vector constructed from the time lag characteristics, indirect characterization characteristics of contact degradation, thermal dynamic characteristics, and heat dissipation conditions and environmental characteristics, and a preset time prediction model, and to generate an early warning signal based on the predicted value of temperature change.

[0073] The time lag characteristics include temperature lag characteristics, current lag characteristics, and temperature rise lag characteristics, which are constructed in the following manner: Using the current moment as a reference, the system traces back along a preset sampling interval, using the operating temperature values ​​of multiple historical sampling points as the temperature hysteresis feature; using the load current values ​​of multiple historical sampling points as the current hysteresis feature; and using the difference between the operating temperature values ​​of multiple historical sampling points and the operating temperature values ​​of their respective previous sampling points as the temperature rise hysteresis feature.

[0074] The indirect characterization features of contact degradation include the current-to-temperature rise ratio anomaly index, temperature rise residual, and temperature difference ratio; these indirect characterization features are constructed in the following manner: The ratio of the temperature rise change to the current change at the current moment is calculated, and the statistical mean of the ratios of temperature rise change to current change at multiple historical moments within the backtracking window is subtracted to obtain the current-temperature rise ratio anomaly index; wherein, the temperature rise change is the difference between the operating temperature value at the current moment and the operating temperature value at the previous sampling moment, and the current change is the difference between the load current value at the current moment and the load current value at the previous sampling moment. The expected temperature rise is calculated based on a simplified thermal model that relies solely on the square term of the load current. The temperature rise residual is obtained by dividing the difference between the actual temperature rise and the expected temperature rise by the rated allowable temperature of the equipment. The actual temperature rise is the difference between the operating temperature value at the current moment and the operating temperature value at the previous sampling moment. The temperature difference ratio is obtained by calculating the ratio of the operating temperature of the target location to the average operating temperature of other similar locations within the same ring main unit.

[0075] The thermal dynamic characteristics include short-time temperature rise rate, cumulative heat effect index, and cooling rate characteristics; these thermal dynamic characteristics are constructed in the following manner: Based on the operating temperature value, the difference between the current operating temperature value and the operating temperature value at the previous sampling time is divided by the sampling interval to obtain the short-time temperature rise rate. Based on the load current value, the squared values ​​of the load current at each sampling time within the preset backtracking window are summed by exponential decay to obtain the cumulative heat effect index. When the load current is detected to be in a decreasing phase, the linear fitting slope of the operating temperature values ​​of a preset number of sampling points is calculated based on the operating temperature value to obtain the cooling rate characteristic.

[0076] The heat dissipation conditions and environmental characteristics include the temperature difference characteristics inside and outside the cabinet, the coding characteristics of the installation environment, and the coding characteristics of the cabinet's location. These heat dissipation conditions and environmental characteristics are constructed in the following way: Based on the ambient temperature value, the difference between the reference temperature inside the cabinet and the ambient temperature value is calculated to obtain the temperature difference characteristics inside and outside the cabinet; The installation environment type in the operating condition identifier is encoded to obtain the installation environment encoding feature; The cabinet location type in the operating condition identifier is encoded to obtain the cabinet location encoding feature.

[0077] The generation of the early warning signal based on the predicted temperature change includes: Based on the current operating temperature value and the predicted temperature change value, calculate the predicted temperature value for a future preset time step. Based on the predicted temperature value and the probability value, a joint mapping is performed in a preset two-dimensional judgment matrix to generate a graded early warning signal; When the graded early warning signal meets the preset triggering conditions, the contribution value of each input feature currently predicted is analyzed based on the model interpretation tool to identify the dominant cause features, and a diagnostic report containing fault cause classification is generated according to the type of the dominant cause features.

[0078] like Figure 3 As shown, this application provides an electronic device for executing the method for predicting the temperature of a ring main unit in this application. The device includes a memory, a processor, a bus, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the method for predicting the temperature of a ring main unit.

[0079] Specifically, the aforementioned memory and processor can be general-purpose memory and processor, without any specific limitations. When the processor runs the computer program stored in the memory, it can execute the aforementioned method for predicting the temperature of the ring main unit.

[0080] Corresponding to the method for predicting the temperature of a ring main unit in this application, this application embodiment also provides a computer storage medium storing a computer program, which is executed by a processor to perform the steps of the above-described method for predicting the temperature of a ring main unit.

[0081] Specifically, the storage medium can be a general-purpose storage medium, such as a portable disk or hard disk. When the computer program on the storage medium is run, it can execute the above-mentioned method for predicting the temperature of the ring main unit.

[0082] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. The system embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interface; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.

[0083] 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; 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, depending on actual needs.

[0084] In addition, the functional units in the embodiments provided in this application 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.

[0085] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion 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 described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0086] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0087] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application. All should be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method for predicting the temperature of a ring main unit, characterized in that, The method includes: Acquire multi-source operating data of the ring main unit; wherein, the multi-source operating data includes operating temperature values, load current values, ambient temperature values, and operating condition identifiers of multiple target locations; The operating temperature value, load current value, ambient temperature value and operating condition identifier are used to construct features to obtain the time lag feature that characterizes the evolution law of temperature and current, the indirect characterization feature of contact degradation that characterizes the relative change trend of contact resistance, the thermal dynamic feature that characterizes the thermodynamic dynamic response characteristics, and the heat dissipation condition and environmental features that characterize the heat dissipation capacity and environmental impact. Based on the time lag characteristics, indirect characterization characteristics of contact degradation, thermal dynamic characteristics, and heat dissipation conditions and environmental characteristics, a multi-dimensional feature vector is constructed, and a preset time prediction model is used to obtain the predicted value of temperature change over a preset time step in the future, and an early warning signal is generated based on the predicted value of temperature change. The time lag characteristics include temperature lag characteristics, current lag characteristics, and temperature rise lag characteristics, which are constructed in the following manner: Using the current moment as a reference, the system traces back along a preset sampling interval, using the operating temperature values ​​of multiple historical sampling points as the temperature hysteresis feature; using the load current values ​​of multiple historical sampling points as the current hysteresis feature; and using the difference between the operating temperature values ​​of multiple historical sampling points and the operating temperature values ​​of their respective previous sampling points as the temperature rise hysteresis feature. The indirect characterization features of contact degradation include the current-to-temperature rise ratio anomaly index, temperature rise residual, and temperature difference ratio; these indirect characterization features are constructed in the following manner: The ratio of the temperature rise change to the current change at the current moment is calculated, and the statistical mean of the ratios of temperature rise change to current change at multiple historical moments within the backtracking window is subtracted to obtain the current-temperature rise ratio anomaly index; wherein, the temperature rise change is the difference between the operating temperature value at the current moment and the operating temperature value at the previous sampling moment, and the current change is the difference between the load current value at the current moment and the load current value at the previous sampling moment. The expected temperature rise is calculated based on a simplified thermal model that relies solely on the square term of the load current. The temperature rise residual is obtained by dividing the difference between the actual temperature rise and the expected temperature rise by the rated allowable temperature of the equipment. The actual temperature rise is the difference between the operating temperature value at the current moment and the operating temperature value at the previous sampling moment. The temperature difference ratio is obtained by calculating the ratio of the operating temperature of the target location to the average operating temperature of other similar locations within the same ring main unit. The thermal dynamic characteristics include short-time temperature rise rate, cumulative heat effect index, and cooling rate characteristics; these thermal dynamic characteristics are constructed in the following manner: Based on the operating temperature value, the difference between the current operating temperature value and the operating temperature value at the previous sampling time is divided by the sampling interval to obtain the short-time temperature rise rate. Based on the load current value, the squared values ​​of the load current at each sampling time within the preset backtracking window are summed by exponential decay to obtain the cumulative heat effect index. When the load current is detected to be in a decreasing phase, the linear fitting slope of the operating temperature values ​​of a preset number of sampling points is calculated based on the operating temperature value to obtain the cooling rate characteristic; The heat dissipation conditions and environmental characteristics include the temperature difference characteristics inside and outside the cabinet, the coding characteristics of the installation environment, and the coding characteristics of the cabinet's location. These heat dissipation conditions and environmental characteristics are constructed in the following way: Based on the ambient temperature value, the difference between the reference temperature inside the cabinet and the ambient temperature value is calculated to obtain the temperature difference characteristics inside and outside the cabinet; The installation environment type in the operating condition identifier is encoded to obtain the installation environment encoding feature; The cabinet location type in the operating condition identifier is encoded to obtain the cabinet location encoding feature.

2. The method according to claim 1, characterized in that, The preset temporal prediction model is a hybrid model of temporal convolutional network and long short-term memory network; the output layer of the hybrid model includes a first output branch and a second output branch. The first output branch is used to output the predicted value of temperature change in the future preset time step, and the second output branch is used to output the probability value of the temperature exceeding the preset safety threshold in the future preset time step.

3. The method according to claim 2, characterized in that, The generation of the early warning signal based on the predicted temperature change includes: Based on the current operating temperature value and the predicted temperature change value, calculate the predicted temperature value for a future preset time step. Based on the predicted temperature value and the probability value, a joint mapping is performed in a preset two-dimensional judgment matrix to generate a graded early warning signal; When the graded early warning signal meets the preset triggering conditions, the contribution value of each input feature currently predicted is analyzed based on the model interpretation tool to identify the dominant cause features, and a diagnostic report containing fault cause classification is generated according to the type of the dominant cause features.

4. A device for predicting the temperature of a ring main unit, characterized in that, The device includes: The data acquisition module is used to acquire multi-source operating data of the ring main unit; wherein, the multi-source operating data includes operating temperature values, load current values, ambient temperature values ​​and operating condition identifiers of multiple target locations; The feature construction module is used to construct features from the operating temperature value, load current value, ambient temperature value and operating condition identifier to obtain time lag features that characterize the evolution of temperature and current, indirect contact degradation features that characterize the relative change trend of contact resistance, thermodynamic features that characterize the thermodynamic dynamic response characteristics, and heat dissipation conditions and environmental features that characterize heat dissipation capacity and environmental impact. The prediction module is used to obtain the predicted value of temperature change over a preset time step based on the multi-dimensional feature vector constructed from the time lag characteristics, indirect characterization characteristics of contact degradation, thermal dynamic characteristics, and heat dissipation conditions and environmental characteristics, and a preset time prediction model, and to generate an early warning signal based on the predicted value of temperature change. The time lag characteristics include temperature lag characteristics, current lag characteristics, and temperature rise lag characteristics, which are constructed in the following manner: Using the current moment as a reference, the system traces back along a preset sampling interval, using the operating temperature values ​​of multiple historical sampling points as the temperature hysteresis feature; using the load current values ​​of multiple historical sampling points as the current hysteresis feature; and using the difference between the operating temperature values ​​of multiple historical sampling points and the operating temperature values ​​of their respective previous sampling points as the temperature rise hysteresis feature. The indirect characterization features of contact degradation include the current-to-temperature rise ratio anomaly index, temperature rise residual, and temperature difference ratio; these indirect characterization features are constructed in the following manner: The ratio of the temperature rise change to the current change at the current moment is calculated, and the statistical mean of the ratios of temperature rise change to current change at multiple historical moments within the backtracking window is subtracted to obtain the current-temperature rise ratio anomaly index; wherein, the temperature rise change is the difference between the operating temperature value at the current moment and the operating temperature value at the previous sampling moment, and the current change is the difference between the load current value at the current moment and the load current value at the previous sampling moment. The expected temperature rise is calculated based on a simplified thermal model that relies solely on the square term of the load current. The temperature rise residual is obtained by dividing the difference between the actual temperature rise and the expected temperature rise by the rated allowable temperature of the equipment. The actual temperature rise is the difference between the operating temperature value at the current moment and the operating temperature value at the previous sampling moment. The temperature difference ratio is obtained by calculating the ratio of the operating temperature of the target location to the average operating temperature of other similar locations within the same ring main unit. The thermal dynamic characteristics include short-time temperature rise rate, cumulative heat effect index, and cooling rate characteristics; these thermal dynamic characteristics are constructed in the following manner: Based on the operating temperature value, the difference between the current operating temperature value and the operating temperature value at the previous sampling time is divided by the sampling interval to obtain the short-time temperature rise rate. Based on the load current value, the squared values ​​of the load current at each sampling time within the preset backtracking window are summed by exponential decay to obtain the cumulative heat effect index. When the load current is detected to be in a decreasing phase, the linear fitting slope of the operating temperature values ​​of a preset number of sampling points is calculated based on the operating temperature value to obtain the cooling rate characteristic; The heat dissipation conditions and environmental characteristics include the temperature difference characteristics inside and outside the cabinet, the coding characteristics of the installation environment, and the coding characteristics of the cabinet's location. These heat dissipation conditions and environmental characteristics are constructed in the following way: Based on the ambient temperature value, the difference between the reference temperature inside the cabinet and the ambient temperature value is calculated to obtain the temperature difference characteristics inside and outside the cabinet; The installation environment type in the operating condition identifier is encoded to obtain the installation environment encoding feature; The cabinet location type in the operating condition identifier is encoded to obtain the cabinet location encoding feature.

5. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine instructions that the processor executes. When the electronic device is running, the processor communicates with the memory via the bus. When the machine instructions are executed by the processor, they perform the steps of the method for predicting the temperature of a ring main unit as described in any one of claims 1 to 3.

6. A computer storage medium, characterized in that, The computer storage medium stores a computer program that, when executed by a processor, performs the steps of the method for predicting the temperature of a ring main unit as described in any one of claims 1 to 3.

Images