Method for monitoring the temperature of an electrical device and monitoring device therefor

Abstract

The application belongs to the technical field of online monitoring of electrical equipment, and provides an electrical equipment temperature monitoring method and a monitoring device thereof. The method comprises the following steps: a remote monitoring end obtains a temperature anomaly probability of a target temperature monitoring device in a next time period based on a second temperature data sequence uploaded by the target temperature monitoring device in a current time period and an associated temperature trend feature; the remote monitoring end dynamically determines a wake-up period based on the temperature anomaly probability, and sends the wake-up period to the target temperature monitoring device; and the target temperature monitoring device receives and adopts the wake-up period, performs periodic temperature collection, and uploads collected temperature data to the remote monitoring end through a base station. The application can solve the balance problem between energy consumption and safety in a fixed cycle monitoring mode, and is helpful to improve the intelligent level of power system operation and maintenance.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of online monitoring of electrical equipment, in particular to an electrical equipment temperature monitoring method and a monitoring device thereof. BACKGROUND

[0002] In a power system, a high-voltage switch cabinet is a key device to ensure power supply reliability. The internal contacts and connecting points of the high-voltage switch cabinet will cause an increase in contact resistance due to aging, loosening or surface oxidation, and abnormal temperature rise will occur when passing through a load current. If this is not discovered in time, it will easily cause serious accidents such as fire and explosion. Therefore, it is crucial to monitor the temperature of the switch cabinet contacts in real time.

[0003] Currently, there are temperature online monitoring schemes based on wireless sensor networks. The monitoring device is usually installed on the internal contacts of the switch cabinet and is composed of a temperature sensor, a micro control unit (MCU) and a LoRa low-power wireless communication module. This type of monitoring device generally adopts a periodic wake-up working mode: the MCU is in a standby state most of the time, and automatically wakes up every fixed time period (for example, 5 minutes) by default, collects temperature data and uploads it wirelessly, and then enters the standby state again. This mode balances energy consumption and data update frequency to some extent.

[0004] However, a fixed wake-up period cannot adapt to the dynamic changes in the running state of the device. In a long period of time when the load is stable and the temperature is normal, too frequent wake-up and reporting will cause waste of energy and shorten the service life of the device; while in periods when the load increases suddenly and the device state is abnormal, a fixed long period may cause the omission of critical temperature change data, and the device cannot capture the fault precursor in time, missing the early warning opportunity.

[0005] To overcome the above-mentioned defects, how to determine a more appropriate wake-up period for the monitoring device to minimize monitoring power consumption while ensuring timely warning is a technical problem that needs to be solved at present. SUMMARY

[0006] To solve the above technical problems, the present application provides an electrical equipment temperature monitoring method, a monitoring device, an electronic device, a non-transitory computer readable storage medium and a computer program product.

[0007] The present application provides an electrical equipment temperature monitoring method, comprising the following steps:

[0008] The remote monitoring end extracts a temperature trend feature set from the first temperature data sequence uploaded by each other temperature monitoring device in the set area, and fuses each temperature trend feature in the temperature trend feature set into an equivalent temperature trend feature using a dynamic sampling method;

[0009] predicting a target temperature monitoring device based on the equivalent temperature trend characteristics and a preset electrical topology relationship; wherein the electrical topology relationship refers to the correlation relationship between the target temperature monitoring device and each other temperature monitoring device based on the power grid structure;

[0010] predicting a temperature anomaly probability of the target temperature monitoring device in the next period based on the second temperature data sequence uploaded by the target temperature monitoring device in the current period and the correlation temperature trend characteristics;

[0011] dynamically determining a wake-up period based on the temperature anomaly probability, and sending the wake-up period to the target temperature monitoring device; the target temperature monitoring device receives and adopts the wake-up period, performs periodic temperature collection, and uploads the collected temperature data to the remote monitoring end through the base station.

[0012] The embodiment of the present application also discloses an electrical equipment temperature monitoring device, which comprises a temperature sensor, a low-power MCU, a wireless LORA module and a power management circuit.

[0013] The wireless LORA module is used for receiving the wake-up period downloaded by the remote monitoring end through the base station, and uploading the temperature monitoring data collected by the temperature sensor to the remote monitoring end through the base station.

[0014] The low-power MCU is used for periodically automatically waking up based on the wake-up period, starting the temperature sensor after waking up and reading the temperature monitoring data of the high-voltage electrical equipment collected by the temperature sensor, and closing the temperature sensor and entering a sleep state after reading is completed.

[0015] The power management circuit is used for supplying power for the temperature sensor, the low-power MCU and the wireless LORA module.

[0016] The present application also provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the method according to any one of the preceding embodiments is implemented.

[0017] The present application also provides a non-transitory computer readable storage medium, wherein the non-transitory computer readable storage medium has instructions embodied thereon, and the instructions can be executed by one or more processors to perform the method according to any one of the preceding embodiments.

[0018] The present application also provides a computer program product, wherein the computer program product comprises an executable computer program, and the computer program is executed to implement the method according to any one of the preceding embodiments.

[0019] The application realizes the cooperative prediction of the temperature state of the target device by introducing the temperature trend feature fusion and the electrical topology relationship analysis, the method can effectively perceive the systematic thermal risk propagation, and converts the traditional passive monitoring based on single device historical data into active early warning based on group trend. Through the probabilistic evaluation of the temperature abnormal risk and the dynamic adjustment of the monitoring strategy, the terminal power consumption can be significantly reduced while ensuring the timeliness of the early warning, the balance between energy consumption and safety under the fixed cycle monitoring mode is solved, and the intelligent level of power system operation and maintenance is improved. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A flowchart of an electrical equipment temperature monitoring method disclosed by the embodiment of the application is shown.

[0021] Figure 2 A structural diagram of an electrical equipment temperature monitoring device disclosed by the embodiment of the application is shown. DETAILED DESCRIPTION

[0022] The technical solutions of the application will be further described in detail below with the aid of the drawings and embodiments.

[0023] To make the purpose, technical solutions and advantages of the embodiments of the application clearer, the technical solutions in the embodiments of the application will be described clearly and completely below with the aid of the drawings of the embodiments of the application. Obviously, the described embodiments are some but not all of the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the application.

[0024] To make the purpose, technical solutions and advantages of the embodiments of the application clearer, the technical solutions in the embodiments of the application will be described clearly and completely below with the aid of the drawings of the embodiments of the application. Obviously, the described embodiments are some but not all of the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the application.

[0025] As shown in the drawings, Figure 1 The embodiment of the application discloses an electrical equipment temperature monitoring method 100, including the following steps:

[0026] 10, the remote monitoring end obtains the temperature trend feature set from the first temperature data sequence uploaded by each other temperature monitoring device in the set region, and fuses each temperature trend feature in the temperature trend feature set into an equivalent temperature trend feature by using a dynamic sampling method;

[0027] In this step, the remote monitoring end obtains the first temperature data sequence (i.e. time series data) periodically uploaded by each other temperature monitoring device (deployed on other monitoring objects) in the set region. It can be understood that the set region refers to a local environment with a compact physical space, for example, the inside of the same power distribution room, switch station, compact box-type substation, etc.

[0028] Subsequently, trend features are extracted from the first temperature data sequence of each other temperature monitoring device, and a plurality of temperature trend features are obtained. The extracted temperature trend features are quantifications of temperature change dynamics, including but not limited to: short-term temperature rise slope, temperature fluctuation variance, deviation from historical same-period mean, etc. Among them, the short-term temperature rise slope is used to reflect the recent temperature change rate; the temperature fluctuation variance is used to characterize the stability of the temperature; and the deviation from the historical same-period mean is used to eliminate the periodic influence of the environment to highlight the anomaly. The temperature trend features of all other temperature monitoring devices are summarized to form a temperature trend feature set.

[0029] Next, all temperature trend features in the temperature trend feature set are fused by using a dynamic sampling method, and an equivalent temperature trend feature is obtained. It can be understood that the fusion process is not a simple average, but a dynamic weighting according to the significance of the feature value of each device or the historical change rate. For example, a device feature with a significantly abnormal temperature rise trend is given a higher weight, so that the equivalent temperature trend feature after fusion can better represent the dynamic level of the overall thermal state in the region. In this way, the data dimension can be simplified, and the key change information can be highlighted.

[0030] 20, based on the equivalent temperature trend feature and the preset electrical topology relationship, the associated temperature trend feature of the target temperature monitoring device is predicted; wherein the electrical topology relationship refers to the correlation relationship between the target temperature monitoring device and each other temperature monitoring device based on the power grid structure;

[0031] In this step, the remote monitoring end combines the equivalent temperature trend feature reflecting the overall thermal dynamics of the region obtained in step 10 with the electrical topology relationship to predict the associated temperature trend feature of the target temperature monitoring device. Specifically, based on the above equivalent temperature trend feature and the electrical topology relationship, the remote monitoring end can predict and generate the associated temperature trend feature of the target temperature monitoring device. The prediction process can be completed based on a first prediction model, for example, Transformer, CNN, etc., and specific details are not repeated.

[0032] The electrical topology relationship is a model established in advance according to the primary wiring diagram of the power grid, which defines the correlation between the target temperature monitoring device and other temperature monitoring devices in the region based on the physical structure of the power grid, such as direct upstream and downstream power supply relationship or parallel relationship in the same bus section. It can be understood that the electrical topology relationship (such as upstream and downstream power supply, same bus parallel) essentially defines a temperature influence network, which can be used to analyze the influence relationship and the strength of the influence between electrical equipment. For example, the operating state (including temperature) of the upstream power switch will directly affect the downstream load switch: the closer the electrical connection (the smaller the impedance, the closer the distance) between the devices, the greater the influence between them. In the above prediction process, the role of the electrical topology relationship is to analyze the most important part of the influence on the target temperature monitoring device from the mixed signal, i.e., the equivalent temperature trend feature, according to the rules of the temperature influence network.

[0033] For example: Assuming that a 10kV bus is connected to an incoming line switch (upstream), a target switch (downstream), and an adjacent switch. When the equivalent temperature trend feature shows that the temperature of the incoming line switch rises significantly, the topology relationship will determine that this signal has the greatest influence on the downstream target switch (because of the direct power supply relationship), and has less influence on the adjacent switch that is far away in electrical distance. The first prediction model accordingly extracts the influence trend of the incoming line switch from the mixed signal as the dominant component of the target switch associated temperature trend feature, thereby accurately warning the thermal risk from the upstream.

[0034] 30, based on the second temperature data sequence uploaded by the target temperature monitoring device in the current period, the associated temperature trend feature is predicted to obtain the temperature abnormality probability of the target temperature monitoring device in the next period;

[0035] In this step, the remote monitoring end synchronously acquires the second temperature data sequence uploaded by the target temperature monitoring device itself in the current period, which directly reflects the individual operating state of the target temperature monitoring device. The associated temperature trend feature obtained in the foregoing step 20, which characterizes the system environmental influence, and the second temperature data sequence, which characterizes the real-time state of the target temperature monitoring device, are taken as joint inputs and sent to a preset second prediction model. The second prediction model analyzes the internal correlation between the above-mentioned input features and outputs a temperature abnormality probability, which represents the possibility of temperature overrun and other abnormal conditions of the target device in the next period.

[0036] 40, based on the temperature abnormality probability, the wake-up period is dynamically determined and sent to the target temperature monitoring device; the target temperature monitoring device receives and adopts the wake-up period, performs periodic temperature collection, and reports the collected temperature data to the remote monitoring end through the base station.

[0037] In this step, the remote monitoring end is internally preset with a mapping rule for mapping continuous temperature anomaly probability into a specific wake-up period instruction. For example, when the temperature anomaly probability is lower than 0.1, it is determined that the risk level is low, and a longer wake-up period (such as 60 minutes) is set accordingly; when the probability is between 0.1 and 0.5, it is determined that the risk level is medium, and a medium wake-up period (such as 10 minutes) is set accordingly; and when the probability is higher than 0.5, it is determined that the risk level is high, and a short wake-up period (such as 1 minute) is set or a continuous monitoring mode is started.

[0038] After the decision is made, the remote monitoring end sends the determined wake-up period instruction to the target temperature monitoring device through a wireless communication network (such as a LoRa base station). After each timed wake-up, the target temperature monitoring device will actively detect whether the control instruction sent by the remote end is received, in addition to completing the basic tasks of temperature collection and reporting. Once a new wake-up period instruction is received, the internal micro control unit (MCU) immediately updates the timer parameters to replace the original period with the newly received wake-up period. Thereafter, the target temperature monitoring device works according to the optimized period: timed wake-up, temperature collection, wireless data transmission, and low-power sleep state. Through this mechanism, the system realizes a complete adaptive cycle from data collection, intelligent analysis, decision optimization to terminal execution.

[0039] The present application realizes the cooperative prediction of the target equipment temperature state by introducing regional temperature trend feature fusion and electrical topology relationship analysis. This method can effectively perceive the systematic heat risk propagation, and changes the traditional passive monitoring based on single device historical data into active early warning based on group trend. By probabilistic evaluation of temperature anomaly risk and dynamic adjustment of monitoring strategy, the terminal power consumption can be significantly reduced while ensuring the timeliness of early warning, solving the balance problem between energy consumption and safety in the fixed cycle monitoring mode, and helping to improve the intelligent level of power system operation and maintenance.

[0040] Optionally, the dynamic sampling method is used to fuse each temperature trend feature in the temperature trend feature set into an equivalent temperature trend feature, comprising:

[0041] 101, calculating the change rate of each temperature trend feature, automatically matching the sampling frequency based on the change rate, and extracting trend sampling data from the corresponding temperature trend feature according to the sampling frequency;

[0042] In this step, first, the historical data sequence of each temperature trend feature is analyzed, and the change rate of its characteristic value is calculated. For example, the change trend slope value of the temperature rise slope feature sequence in the past period of time is obtained by linear fitting, and the change rate is calculated. Based on the calculation result, according to the preset mapping relationship, appropriate sampling frequency is allocated to each feature: the features with large change rate (such as rapid increase of temperature rise slope) are allocated higher sampling frequency in order to densely capture its dynamic change; and the features with slow change are allocated lower sampling frequency to improve sampling efficiency.

[0043] According to the sampling frequency determined above, the corresponding data points are extracted from the historical data sequence of each temperature trend feature to form the trend sampling data corresponding to each temperature trend feature.

[0044] 102, dynamically calculate the significance evaluation value of the corresponding temperature trend feature according to the trend sampling data, and calculate the decay weight of the corresponding temperature trend feature based on the time decay function;

[0045] In this step, the statistical characteristics of each temperature trend feature are calculated based on the trend sampling data, such as the coefficient of variation of its characteristic value or the deviation from the reference value. The greater the deviation, the higher the significance evaluation value.

[0046] At the same time, the time decay function is introduced to calculate the decay weight. The time decay function determines the weight value of the data point according to its timestamp. The weight of recent data is higher, and the weight of historical data gradually decays over time. It can be understood that by introducing the time decay factor into the weight calculation, the evaluation of the importance of the feature can consider both its abnormality and its timeliness.

[0047] 103, dynamically allocate the fusion weight of each temperature trend feature according to the significance evaluation value and the decay weight, and output the equivalent temperature trend feature by sliding weighted average of each temperature trend feature in the sampling time window according to the fusion weight.

[0048] In this step, the normalized significance evaluation value and the decay weight are converted into the fusion weight of each temperature trend feature by weighted synthesis method, and the weight proportion of the significance evaluation value is higher to ensure that the abnormal features can be given enough attention.

[0049] In the set sampling time window, all temperature trend features are calculated by sliding weighted average. Specifically: when new data points are added, the fusion weight is automatically updated, and the weighted average value is recalculated. At the same time, the old data beyond the time window is removed, and then the equivalent temperature trend feature is obtained. This dynamic weighted fusion mechanism can ensure that the output equivalent temperature trend feature can highlight important abnormal information and reflect the latest state change.

[0050] The embodiment can dynamically adjust the sampling frequency according to the change of each temperature trend feature, significantly improve the sampling efficiency under the premise of ensuring the integrity of feature capture; at the same time, by introducing the double weight adjustment mechanism of saliency and time decay, the feature fusion process considers both the abnormality degree and the timeliness, ensuring that the equivalent temperature trend feature can not only capture abnormal signals sharply, but also accurately reflect the latest state of the system.

[0051] Optionally, the dynamic calculation of the saliency evaluation value of the corresponding temperature trend feature according to the trend sampling data comprises:

[0052] 1021, respectively calculating a first statistical deviation and a second statistical deviation of the trend sampling data relative to a normal working condition reference; wherein the first statistical deviation is a normalized Euclidean distance between the mean value of the trend sampling data and the median value of the normal temperature interval, and the second statistical deviation is a ratio of the variance of the trend sampling data to the variance of the normal temperature interval;

[0053] In this step, the abnormality degree of the trend sampling data is quantified from two dimensions. The first statistical deviation represents the change of temperature level, and by calculating the normalized Euclidean distance between the mean value of the trend sampling data and the median value of the normal temperature interval, the dimension influence is eliminated, and the overall temperature deviation degree is accurately reflected. The greater the distance value, the farther the temperature deviates from the normal level. The second statistical deviation represents the temperature volatility, and by calculating the ratio of the variance of the trend sampling data to the variance of the normal temperature interval, the temperature stability change is reflected. A ratio greater than 1 indicates that the fluctuation is intensified, and a ratio less than 1 tends to be stable. These two deviations depict the temperature abnormality feature from different angles.

[0054] 1022, based on the system load level at the current time, querying the first dynamic weight and the second dynamic weight corresponding to the first statistical deviation and the second statistical deviation from a preset weight configuration table; wherein the system load level refers to the running level of the total current or total power of the set region;

[0055] In this step, the system running context information is introduced to make the evaluation more consistent with the actual working condition of the electrical equipment, and the system load level therein reflects the running level of the total current or total power of the set region. According to different system load levels, the optimal weight combination is dynamically selected from the preset weight configuration table. Specifically, in the high load working condition, the weight of the first statistical deviation (i.e. the first dynamic weight) is increased, because the risk of temperature level deviation is greater at this time; in the load fluctuation condition, the weight of the second statistical deviation (i.e. the second dynamic weight) is increased, and the change of temperature stability is focused on.

[0056] 1023, calculate a preliminary saliency according to the first statistical deviation and the first dynamic weight, and the second statistical deviation and the second dynamic weight, input the preliminary saliency into a preset S-shaped function for normalization processing, and map an output result to a [0, 1] interval to obtain a final saliency evaluation value.

[0057] In this step, the deviation information of two dimensions is integrated into a preliminary saliency by using a weighted sum, and the weight configuration reflects the evaluation strategy focus under different operating conditions. Then, the preliminary saliency is mapped to the [0, 1] standard interval through an S-shaped function (such as a sigmoid function), which not only ensures the uniformity of the output value range, but also amplifies the discrimination of the significant abnormal signal, and finally obtains a standardized saliency evaluation value that can be compared across features.

[0058] The embodiment comprehensively considers the temperature level deviation and volatility change, and dynamically adjusts the evaluation focus from the actual operating condition, so that the saliency evaluation value can more comprehensively and accurately quantify the abnormal risk. Finally, through nonlinear normalization processing, a standardized evaluation result is generated, providing a stable and reliable input for subsequent feature fusion. The scheme of the embodiment can significantly improve the discrimination ability and evaluation accuracy of the temperature abnormal state under different load conditions, and enhance the adaptability and reliability of temperature monitoring.

[0059] Optionally, the associated temperature trend feature of the target temperature monitoring device is predicted based on the equivalent temperature trend feature and a preset electrical topology relationship, and the method comprises the following steps of:

[0060] 201, injecting the equivalent temperature trend feature as an initial node feature into a topology graph corresponding to the electrical topology relationship, wherein a node in the topology graph represents a temperature monitoring device, and an edge represents an electrical connection relationship, and a weight of the edge is determined based on a connected electrical distance or impedance value;

[0061] In this step, the equivalent temperature trend feature obtained in step 10 is taken as an initial feature and is assigned to the corresponding node in the topology graph. The topology graph is a mathematical model constructed according to the actual wiring structure of the power grid, i.e., the electrical topology relationship described above. The nodes in the topology graph represent various temperature monitoring devices, and the edges represent the electrical connection relationship (such as power supply and parallel relationship) between the devices. The weight of the edge is determined according to the connected electrical distance or impedance value. The smaller the impedance or the closer the distance, the greater the weight, indicating a stronger electrical association.

[0062] 202, in a graph neural network constructed based on the electrical topology relationship, the feature information of the neighborhood nodes of each node is aggregated through a multi-round message passing mechanism, wherein the aggregation weight is determined by the weight of the edge and the similarity of the neighborhood node features;

[0063] In this step, the nodes in the topology graph are mapped to the neuron units of the graph neural network, and the edges and their weights are converted into the message passing paths and initial connection strengths in the neural network.

[0064] In the graph neural network, each node interacts with its electrical neighborhood nodes through a multi-round message passing mechanism. On this basis, feature propagation is achieved through a multi-round message passing mechanism. Specifically, each node aggregates the feature information of its neighborhood nodes, and the aggregation process adopts a double weight mechanism. The first weight is based on the edge weight of the topology graph (electrical distance or impedance value), which ensures that the information transmission conforms to the physical structure of the power grid; the second weight is based on the real-time similarity calculation of node features, which is used to dynamically adjust the information transmission strength.

[0065] The neural network construction method based on electrical topology adopted in this embodiment can make the graph neural network inherit the connection characteristics of the power grid and ensure that the learning process conforms to the basic laws of the power system.

[0066] 203, after a preset number of rounds of feature propagation, the output features of the node corresponding to the target temperature monitoring device are extracted and used as the associated temperature trend features.

[0067] In this step, after a preset number of rounds of feature propagation, the features of each node contain its own initial information and the state information of the related nodes in the multi-hop neighborhood. At this time, the output features of the node corresponding to the target temperature monitoring device are extracted. The output features have been deeply processed by the topology network and have integrated the temperature trend influence of the electrically associated equipment, and can more comprehensively reflect the real thermal dynamics of the target device in the power grid system. Finally, they are used as the predicted associated temperature trend features.

[0068] This embodiment maps and converts the topology graph and the neural network, so that the graph neural network can inherit the connection characteristics of the power grid and ensure that the learning process conforms to the physical laws of the power system. At the same time, a double weight aggregation mechanism based on electrical impedance and feature similarity is adopted, so that the feature propagation can both maintain structural constraints and dynamically adapt to state changes. Therefore, the finally output associated temperature trend features deeply integrate the influence of the neighborhood equipment, which can improve the accuracy and physical interpretability of the prediction results, and provide a more reliable basis for temperature anomaly early warning.

[0069] Optionally, based on the second temperature data sequence uploaded by the target temperature monitoring device in the current period and the associated temperature trend features, the temperature anomaly probability of the target temperature monitoring device in the next period is predicted, including:

[0070] 301, time sequence feature extraction is performed on the second temperature data sequence to obtain a local time sequence feature vector of the target temperature monitoring device; the local time sequence feature vector and the associated temperature trend features are spliced to form a comprehensive feature vector;

[0071] In this step, the second temperature data sequence of the target temperature monitoring device itself is subjected to time sequence feature extraction to obtain a local time sequence feature vector. The local time sequence feature vector is used to represent the individual real-time running state of the target temperature monitoring device, including but not limited to: recent temperature rise rate, temperature fluctuation amplitude, deviation from the historical same period mean, and other key indicators. It can be understood that these features are used to reflect the independent temperature change dynamics of the target temperature monitoring device.

[0072] The above local time sequence feature vector is spliced with the associated temperature trend feature obtained in step 20. The associated temperature trend feature reflects the systematic thermal influence based on the electrical topology relationship, and the splicing of the two makes the finally formed comprehensive feature vector contain both the individual state information of the target temperature monitoring device and the systematic influence it suffers in the power grid environment.

[0073] 302, input the comprehensive feature vector into a pre-trained probability prediction model, and the probability prediction model outputs a temperature anomaly probability value between 0 and 1; wherein the probability prediction model is trained based on historical data, and the training samples include historical temperature data sequences, historical associated temperature trend features, and corresponding actual temperature anomaly labels.

[0074] In this step, the above comprehensive feature vector is input into the pre-trained probability prediction model. The probability prediction model can be constructed based on classification or regression algorithms (such as logistic regression, gradient boosting tree, or simple neural network), and the output layer adopts Sigmoid or other activation functions to ensure that the final output is a continuous probability value between 0 and 1.

[0075] It can be understood that the above temperature anomaly probability value represents the possibility of the target temperature monitoring device occurring temperature anomaly (such as exceeding the safety threshold) in the next period (such as the next 15 minutes). The closer the temperature anomaly probability value is to 1, the higher the abnormal risk is; the closer it is to 0, the safer the state is.

[0076] The embodiment combines the local time sequence feature reflecting the state of the device itself with the associated trend feature representing the systematic influence to construct a comprehensive feature input that describes the temperature state. At the same time, the probability prediction model trained based on historical data can effectively learn the non-linear relationship between complex features and abnormal states, and output continuous probability values. The embodiment realizes more accurate prediction of the temperature anomaly probability, so that the rationality of the subsequent determined wake-up period is higher.

[0077] As Figure 2As shown, the embodiment of the application further discloses an electrical equipment temperature monitoring device 200, which comprises a temperature sensor 2001, a low-power MCU 2002, a wireless LORA module 2003, and a power management circuit 2004.

[0078] The wireless LORA module is configured to receive a wake-up period transmitted by a remote monitoring terminal from a base station, and upload temperature monitoring data collected by the temperature sensor to the remote monitoring terminal through the base station.

[0079] The low-power MCU is configured to automatically wake up periodically based on the wake-up period, start the temperature sensor after waking up, read temperature monitoring data of the high-voltage electrical equipment collected by the temperature sensor, and turn off the temperature sensor and enter a sleep state after reading is completed.

[0080] The power management circuit is configured to supply power to the temperature sensor, the low-power MCU, and the wireless LORA module.

[0081] The temperature monitoring device can be installed on a contact in a power distribution cabinet, for example, and the temperature sensor is in close contact with the contact conductor. Initially, the low-power MCU automatically wakes up after default standby for 5 minutes, turns on the temperature sensor power and reads the temperature sensor data, packs the temperature monitoring data according to the communication protocol, turns off the temperature sensor, turns on the wireless LORA module, initializes the communication parameters, sends the packed data to the base station, and enters the standby state immediately after the sending is completed. The above process is repeated to realize remote collection of the contact temperature in the switch cabinet. When the low-power MCU receives an updated wake-up period fed back by the remote monitoring terminal, the low-power MCU automatically wakes up according to the wake-up period, and then performs the above monitoring and transmission process.

[0082] Optionally, the device further comprises a CT coil 2005, which supplies power to the power management circuit through magnetic induction.

[0083] The embodiment of the application further discloses an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the method according to any one of the preceding embodiments is implemented.

[0084] The embodiment of the application further discloses a non-transitory computer readable storage medium having instructions embodied thereon, which can be executed by one or more processors to perform the method according to any one of the preceding embodiments.

[0085] While the application has been particularly shown and described with reference to preferred embodiments, it will be understood to those skilled in the art that various changes in form and details can be made therein without departing from the spirit and scope of the application. Accordingly, the disclosed application is to be considered as illustrative and not restrictive, and the application is defined by the scope of the appended claims.

Claims

1. A method of electrical equipment temperature monitoring, characterized by, The method comprises the following steps: The remote monitoring terminal extracts temperature trend features from the first temperature data sequence uploaded by each other temperature monitoring device in the set area, and fuses each temperature trend feature in the temperature trend feature set into an equivalent temperature trend feature using a dynamic sampling method; Based on the equivalent temperature trend feature and the preset electrical topology relationship, the associated temperature trend feature of the target temperature monitoring device is predicted; wherein, the electrical topology relationship refers to the correlation between the target temperature monitoring device and each other temperature monitoring device based on the power grid structure; Based on the second temperature data sequence uploaded by the target temperature monitoring device in the current period and the associated temperature trend feature, the temperature anomaly probability of the target temperature monitoring device in the next period is predicted; Based on the temperature anomaly probability, the wake-up period is dynamically determined and sent to the target temperature monitoring device; the target temperature monitoring device receives and uses the wake-up period to periodically collect temperature data, and reports the collected temperature data to the remote monitoring terminal through the base station; The method for fusing each temperature trend feature in the temperature trend feature set into an equivalent temperature trend feature using a dynamic sampling method comprises the following steps: The change rate of each temperature trend feature is calculated, and the sampling frequency is automatically matched based on the change rate; the trend sampling data is extracted from the corresponding temperature trend feature according to the sampling frequency; The significance evaluation value of the corresponding temperature trend feature is dynamically calculated according to the trend sampling data, and the decay weight of the corresponding temperature trend feature is calculated based on a time decay function; The fusion weight of each temperature trend feature is dynamically allocated according to the significance evaluation value and the decay weight, and the sliding weighted average of each temperature trend feature in the sampling time window is performed according to the fusion weight, and the equivalent temperature trend feature is output; The method for dynamically calculating the significance evaluation value of the corresponding temperature trend feature according to the trend sampling data comprises the following steps: The first statistical deviation and the second statistical deviation of the trend sampling data relative to the normal working condition reference are calculated respectively; wherein, the first statistical deviation is the normalized Euclidean distance between the mean value of the trend sampling data and the median value of the normal temperature interval, and the second statistical deviation is the ratio of the variance of the trend sampling data to the variance of the normal temperature interval; Based on the system load level at the current time, the first dynamic weight and the second dynamic weight corresponding to the first statistical deviation and the second statistical deviation are queried from the preset weight configuration table; wherein, the system load level refers to the operation level of the total current or total power of the set area; The preliminary significance is calculated according to the first statistical deviation and the first dynamic weight, and the second statistical deviation and the second dynamic weight; the preliminary significance is input into the preset S-type function for normalization processing, and the output result is mapped to the [0, 1] interval to obtain the final significance evaluation value.

2. The method of claim 1, wherein: The method for predicting the associated temperature trend feature of the target temperature monitoring device based on the equivalent temperature trend feature and the preset electrical topology relationship comprises the following steps: Inject the equivalent temperature trend feature as an initial node feature into a topology graph corresponding to the electrical topology relationship, where nodes in the topology graph represent temperature monitoring devices, edges represent electrical connection relationships, and weights of edges are determined based on connection electrical distance or impedance values; In a graph neural network constructed based on the electrical topology relationship, a multi-round message passing mechanism is used to make each node aggregate feature information of its neighbor nodes, where aggregation weights are determined by weights of edges and similarity of neighbor node features; After a preset number of rounds of feature propagation, output features of a node corresponding to the target temperature monitoring device are extracted and used as the associated temperature trend feature.

3. A method of monitoring the temperature of an electrical device as defined in claim 2, wherein: Based on a second temperature data sequence uploaded by the target temperature monitoring device in a current period and the associated temperature trend feature, a temperature anomaly probability of the target temperature monitoring device in a next period is predicted, including: Temporal feature extraction is performed on the second temperature data sequence to obtain a local temporal feature vector of the target temperature monitoring device; the local temporal feature vector and the associated temperature trend feature are spliced to form a comprehensive feature vector; The comprehensive feature vector is input into a pre-trained probability prediction model, and the probability prediction model outputs a temperature anomaly probability value between 0 and 1; the probability prediction model is trained based on historical data, and training samples thereof include historical temperature data sequences, historical associated temperature trend features, and corresponding actual temperature anomaly labels.

4. An electrical equipment temperature monitoring apparatus based on the method of any of claims 1-3, characterized by: The monitoring device includes a temperature sensor, a low-power MCU, a wireless LORA module, and a power management circuit. The wireless LORA module is configured to receive a wake-up period transmitted by a remote monitoring end from a base station and upload temperature monitoring data collected by the temperature sensor to the remote monitoring end through the base station. The low-power MCU is configured to automatically wake up periodically based on the wake-up period, start the temperature sensor after waking up, read temperature monitoring data of the high-voltage electrical equipment collected by the temperature sensor, and turn off the temperature sensor and enter a sleep state after reading is completed. The power management circuit is configured to supply power to the temperature sensor, the low-power MCU, and the wireless LORA module.

5. An electrical device temperature monitoring apparatus according to claim 4, wherein: The device further includes a CT coil that supplies power to the power management circuit through magnetic induction.

6. An electronic device, comprising: The device includes a memory and a processor, and the memory stores a computer program that, when executed by the processor, implements the method of any one of claims 1-3.

7. A non-transitory computer-readable storage medium having instructions embodied thereon, wherein, The instructions can be executed by one or more processors to perform the method for implementing any one of claims 1-3.

8. A computer program product comprising a computer program that can be executed in a computer, characterized in that: The computer program, when executed, implements the method of any one of claims 1-3.

Images