Method, apparatus and device for monitoring electromechanical equipment based on agent, and medium

Abstract

Embodiments of the present application disclose an agent-based electromechanical equipment monitoring method, device, equipment and medium. The method comprises: obtaining current multi-source feature information of a target electrical circuit, the current multi-source feature information comprising basic parameters of each device in the target electrical circuit, real-time electrical parameters of each device, pipeline node temperature and environmental parameters; inputting the current multi-source feature information into a preset agent model, performing spatio-temporal fusion analysis and processing on the current multi-source feature information and historical multi-source feature information of the target electrical circuit within a preset first historical time length through a spatio-temporal attention mechanism in the agent model to obtain spatio-temporal fusion features of the target electrical circuit; inputting the spatio-temporal fusion features into an alarm threshold output layer in the agent model to output a target alarm threshold of the target electrical circuit; and monitoring the target electrical circuit according to the target alarm threshold. The accuracy of electromechanical equipment monitoring can be improved by implementing the method of the embodiments of the present application.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to an electromechanical equipment monitoring method and device based on an agent, an equipment and a medium. BACKGROUND

[0002] In the field of modern residential buildings, electromechanical systems (including electrical pipelines, distribution boxes, sockets, lamps, air conditioners, water heaters, and other electromechanical equipment) are the core infrastructure for ensuring the quality and safety of life. The pipelines and equipment in these systems generally show signs of aging after a certain period of use. When aging reaches a critical point, a series of safety hazards can easily occur, including but not limited to fires caused by overheated wires, building structure damage and resource waste caused by pipeline leaks, and fatal risks such as gas leaks. The traditional "after-maintenance" strategy, which takes action only after a failure or disaster occurs, has been unable to meet the needs of modern smart buildings for safe, reliable, and efficient operation and maintenance.

[0003] To overcome the drawbacks of the traditional mode, the concept of predictive maintenance has emerged, and a monitoring system based on sensor technology has been developed. This system relies on various sensors deployed on circuits or equipment to achieve early warning by collecting data in real time and comparing it with pre-set fixed thresholds.

[0004] However, the current monitoring system generally uses static and empirical alarm thresholds. Once these thresholds are set, they usually remain unchanged throughout the system's life cycle, and cannot be adjusted according to the running time of the equipment and the state of the running equipment in the circuit, affecting the accuracy of electromechanical equipment monitoring. SUMMARY

[0005] The embodiments of the present application provide an electromechanical equipment monitoring method, device, equipment and medium based on an agent, which can adaptively adjust the alarm threshold and improve the accuracy of electromechanical equipment monitoring.

[0006] In a first aspect, the embodiments of the present application provide an electromechanical equipment monitoring method based on an agent, which includes:

[0007] Obtaining current multi-source feature information of a target electrical circuit, the current multi-source feature information including basic parameters of each device in the target electrical circuit, real-time electrical parameters of each device, pipeline node temperature, and environmental parameters, the basic parameters including service life, standby power consumption, running power consumption, cumulative running time, and boot-up time;

[0008] Inputting the current multi-source feature information into a pre-set agent model, and performing spatio-temporal fusion analysis and processing on the current multi-source feature information and historical multi-source feature information of the target electrical circuit within a pre-set first historical time length through a spatio-temporal attention mechanism in the agent model to obtain spatio-temporal fusion features of the target electrical circuit.

[0009] inputting the spatio-temporal fusion feature into an alarm threshold output layer in the agent model to output a target alarm threshold of the target electrical circuit;

[0010] monitoring the target electrical circuit according to the target alarm threshold.

[0011] In a second aspect, the embodiments of the present application further provide an electromechanical equipment monitoring device based on an agent, which comprises:

[0012] a transceiving unit configured to acquire current multi-source feature information of a target electrical circuit, the current multi-source feature information comprising basic parameters of each device in the target electrical circuit, real-time electrical parameters of each device, pipeline node temperature and environmental parameters, the basic parameters comprising service life, standby power consumption, running power consumption, cumulative running time and boot-up time;

[0013] a processing unit configured to input the current multi-source feature information into a preset agent model, perform spatio-temporal fusion analysis and processing on the current multi-source feature information and historical multi-source feature information of the target electrical circuit within a preset first historical time length through a spatio-temporal attention mechanism in the agent model to obtain spatio-temporal fusion feature of the target electrical circuit, input the spatio-temporal fusion feature into an alarm threshold output layer in the agent model to output a target alarm threshold of the target electrical circuit, and monitor the target electrical circuit according to the target alarm threshold.

[0014] In a third aspect, the embodiments of the present application further provide a computer device, which comprises a memory and a processor, the memory having a computer program stored thereon, and the processor implements the above method when executing the computer program.

[0015] In a fourth aspect, the embodiments of the present application further provide a computer readable storage medium, the storage medium storing a computer program, the computer program comprising program instructions, and the program instructions can implement the above method when executed by a processor.

[0016] The embodiment of the present application provides a method, device, equipment and medium for monitoring electromechanical equipment based on an agent. The method comprises the following steps: acquiring current multi-source feature information of a target electrical circuit, wherein the current multi-source feature information comprises basic parameters of each device in the target electrical circuit, real-time electrical parameters of each device, pipeline node temperature and environmental parameters; inputting the current multi-source feature information into a preset agent model; performing spatio-temporal fusion analysis and processing on the current multi-source feature information and historical multi-source feature information of the target electrical circuit in a preset first historical time length through a spatio-temporal attention mechanism in the agent model, to obtain spatio-temporal fusion features of the target electrical circuit; inputting the spatio-temporal fusion features into an alarm threshold output layer in the agent model, and outputting a target alarm threshold of the target electrical circuit; and monitoring the target electrical circuit according to the target alarm threshold. The embodiment of the present application can adaptively adjust the alarm threshold of the electrical circuit through the preset agent model and the multi-source feature information of the electrical circuit, thereby improving the accuracy of the monitoring of the electromechanical equipment. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0018] Figure 1 The flowchart of the method for monitoring electromechanical equipment based on an agent provided by the embodiment of the present application is shown.

[0019] Figure 2 The connection diagram of the agent model and the feature library provided by the embodiment of the present application is shown.

[0020] Figure 3 The sub-flowchart of the method for monitoring electromechanical equipment based on an agent provided by the embodiment of the present application is shown.

[0021] Figure 4 The training step diagram of the agent model in the method for monitoring electromechanical equipment based on an agent provided by the embodiment of the present application is shown.

[0022] Figure 5 The schematic block diagram of the device for monitoring electromechanical equipment based on an agent provided by the embodiment of the present application is shown.

[0023] Figure 6 The schematic block diagram of the computer device provided by the embodiment of the present application is shown. DETAILED DESCRIPTION

[0024] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0025] It should be understood that the terms "comprising" and "including" as used in the specification and the appended claims indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0026] It should also be understood that the terms used in the present application specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, the singular forms "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0027] It should be further understood that the term "and / or" used in the present application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations thereof.

[0028] The embodiments of the present application provide an agent-based electromechanical equipment monitoring method, device, equipment and medium.

[0029] The execution subject of the agent-based electromechanical equipment monitoring method can be an agent-based electromechanical equipment monitoring device provided by the embodiments of the present application, or a computer device integrated with the agent-based electromechanical equipment monitoring device, wherein the agent-based electromechanical equipment monitoring device can be realized in the form of hardware or software, and the computer device can be a terminal or a server.

[0030] In the present embodiment, the agent-based electromechanical equipment monitoring method provided can be applied to the monitoring of electromechanical equipment of a target house, which can be a residential house or a commercial house, and a power circuit is arranged in the target house, and the agent-based electromechanical equipment monitoring method provided by the present application is executed for each power circuit.

[0031] In addition, when the target house includes multiple power circuits, the present application further judges the state of the main circuit in combination with the operation and maintenance results of the multiple power circuits.

[0032] Figure 1 FIG. 1 is a flowchart of an agent-based electromechanical equipment monitoring method provided by the embodiments of the present application. As shown in FIG. 1, the agent-based electromechanical equipment monitoring method provided by the embodiments of the present application includes the following steps. Figure 1As shown, the method comprises steps S110-S140.

[0033] S110, acquire current multi-source feature information of the target electrical circuit, the current multi-source feature information comprising basic parameters of each device in the target electrical circuit, real-time electrical parameters of each device, pipeline node temperature, and environmental parameters.

[0034] Specifically, each device in the target electrical circuit is an electromechanical device located in the electrical circuit, and the basic parameters of each device comprise service life, standby power consumption, running power consumption, cumulative running time, and boot-up time, etc.; the real-time electrical parameters comprise current, voltage, active power, apparent power, and residual current, etc.; the pipeline node temperature comprises node temperature of each preset node in the electrical pipeline of the electrical circuit, the preset node comprising interface nodes of each device, internal nodes of an electrical box, internal nodes of a socket, internal nodes of a switch, and cable specific position nodes, etc.; the environmental parameters comprise environmental temperature and environmental humidity.

[0035] In some embodiments, as Figure 2 As shown, different feature libraries are set for different sources of feature information in this embodiment, the current multi-source feature information is updated into the corresponding feature library, and all feature libraries are uniformly accessed to the agent model. Specifically, the feature library 1 comprises basic parameters of each device, the feature library 2 comprises real-time electrical parameters of each device, the feature library 3 comprises pipeline node temperature, and the feature library 4 comprises environmental parameters.

[0036] Among them, after the initial value of the feature library 1 is input, the knowledge graph is established by the agent model, and the knowledge graph is adjusted in real time according to the running state of the device; the feature library 2 information is collected in real time; the feature library 3 is divided into node temperature zones, and different features can be set, including: the highest temperature of the partition, the position of the highest temperature of the partition, the lowest temperature of the partition, the position of the lowest temperature of the partition, the temperature rise rate, the maximum value position of the temperature rise rate, and the average temperature, wherein the partition can be the target electrical circuit or a pre-divided partition in the target electrical circuit;

[0037] The feature library 4: the environmental temperature and humidity of the house partition. The nodes of the knowledge graph are distribution boxes, sockets, lamps, air conditioners, water heaters, etc., and the node attributes comprise service life Y, cumulative running hours H, historical maintenance times R, and health attenuation coefficient a (YHR).

[0038] S120, input the current multi-source feature information into a preset agent model, and perform spatio-temporal fusion analysis and processing on the current multi-source feature information and historical multi-source feature information of the target electrical circuit within a preset first historical time length through a spatio-temporal attention mechanism in the agent model, to obtain spatio-temporal fusion features of the target electrical circuit.

[0039] In this embodiment, after obtaining the current multi-source feature information, the spatio-temporal attention mechanism in the intelligent agent model is used to perform spatio-temporal fusion analysis and processing on the current multi-source feature information and the historical multi-source feature information of the target electrical circuit within a preset first historical time length, so as to extract the spatio-temporal fusion features of the current target electrical circuit. The first historical time length is a past preset time length from the current time, for example, 24 hours in the past. In addition, the historical multi-source feature information within the first historical time length can be extracted from a preset database, for example, from the above-mentioned feature library 1-feature library 4.

[0040] In addition, in some embodiments, after obtaining the current multi-source feature information of the target electrical circuit, the corresponding knowledge graph is updated according to the current multi-source feature information, and the subsequent intelligent agent model can update the alarm threshold based on the updated intelligent graph.

[0041] In some embodiments, the spatio-temporal attention mechanism includes a time attention module and a space attention module. Specifically, please refer to Figure 3 , step S120 includes:

[0042] S1201, constructing an input sequence according to the current multi-source feature information and the historical multi-source feature information;

[0043] S1202, analyzing the input sequence by the time attention module to determine the time correlation features of each device in the target electrical circuit;

[0044] S1203, analyzing the input sequence by the space attention module to determine the spatial correlation features between different devices in the target electrical circuit;

[0045] S1204, determining the spatio-temporal fusion features according to the time correlation features and the spatial correlation features.

[0046] In this embodiment, the input sequence includes multi-source feature information within N time windows, N is an integer greater than 1, and the N time windows include a current time window and a time window within a preset first historical time length; then the time attention module (Transformer Encoder) is used to analyze the input sequence, extract the time correlation features of each device in the target electrical circuit, the time correlation features are the long-term dependence relationship of the device, reflect the change of the electrical parameters of the device with time, and analyze the spatial correlation features between different devices by analyzing the input sequence, wherein the basic parameters also include the specific position parameters of the device, the spatial correlation relationship reflects the physical correlation such as heat conduction and electromagnetic interference between different intervals (an interval includes one or more devices) or different devices in the target electrical circuit, and the spatio-temporal fusion features include the above-mentioned time correlation features and spatial correlation features.

[0047] S130, input the spatio-temporal fusion feature into an alarm threshold output layer in the agent model, and output a target alarm threshold of the target electrical circuit.

[0048] In this embodiment, after obtaining the spatio-temporal fusion feature, the spatio-temporal fusion feature is input into the alarm threshold output layer, the calculation of the alarm threshold is performed through the alarm threshold output layer, and the target alarm threshold of the target electrical circuit is output. The target alarm threshold of the target electrical circuit includes a circuit alarm threshold and an alarm threshold of each device in the circuit, and the alarm threshold includes a temperature alarm threshold, a current alarm threshold, and a humidity alarm threshold.

[0049] Further, the agent model further includes a residual service life output layer and a risk level output layer. After obtaining the spatio-temporal fusion feature of the target electrical circuit, the method further includes: inputting the spatio-temporal fusion feature into the residual service life output layer, and outputting a residual service life value of each device in the target electrical circuit; inputting the spatio-temporal fusion feature into the risk level output layer, and outputting a risk level of the target electrical circuit.

[0050] It can be seen that the agent model of the embodiment is provided with three output heads, and the three output heads can output different task targets through the same input, wherein the risk level of the target electrical circuit includes a circuit risk level and a risk level of each device in the circuit.

[0051] The risk level is a qualitative and immediate evaluation of the severity of the current system state, and the risk level includes high risk, medium risk, and low risk.

[0052] High risk: means that there is an immediate and serious threat (such as a sharp rise in temperature, excessive leakage current). The system will trigger the highest level of alarm, and may perform forced measures (such as immediately cutting off the power supply), and at the same time notify the user through various channels (App push, SMS, sound and light).

[0053] Medium risk: indicates that the system monitors a continuous deterioration trend or potential defect (such as slow decline in device performance, increasing parameter fluctuations). The system generates a warning work order, recommends checking in the next maintenance cycle, and may automatically fine-tune the alarm threshold to strengthen monitoring.

[0054] Low risk: represents that the system is running in a normal or controllable state. The system remains silent monitoring, and does not need to trigger any alarm to avoid disturbing the user.

[0055] The role of the residual service life value is to quantitatively and prospectively predict the failure time of the device or the circuit.

[0056] Further, the application also provides synchronization operation and maintenance prediction of the expert knowledge base and the agent model. The expert knowledge base is pre-provided with a plurality of early warning rules. In the case of a fault that cannot be touched by the model, the accuracy and compliance of the system are improved.

[0057] For example, early warning rule 1: when any loop temperature > 60°C and lasts more than 10 minutes → force trigger the first level alarm of the loop (high risk);

[0058] Early warning rule 2: residual current > residual current threshold (such as 30mA) → directly trigger high risk, immediately cut off the power supply and push the emergency notification.

[0059] Early warning rule 3: device running time > risk time limit of the corresponding device (such as 10 years) + power fluctuation > 20% (power fluctuation threshold) → automatically reduce the power alarm threshold (such as reduce by 15%).

[0060] Pre-set rule 4: if the increase value of the environmental humidity in the environmental parameters within a pre-set second historical time length (such as 10 minutes) is greater than a pre-set humidity increase threshold (such as 85% relative humidity), then increase the leakage current alarm threshold of the target electric circuit within a pre-set future time length (for example, 2 hours in the future) (for example, increase by 20% tolerance).

[0061] After the spatio-temporal fusion features are input into the risk level output layer of the agent model, the method further comprises:

[0062] If the target electric circuit is continuously determined as a medium risk level by the agent model for a pre-set number of times, then the current alarm threshold of the target electric circuit is adjusted downward according to a pre-set alarm parameter adjustment rule.

[0063] This embodiment combines pre-set early warning rules and the agent model to adjust the alarm threshold, thereby improving the adjustment accuracy.

[0064] S140, monitoring the target electric circuit according to the target alarm threshold.

[0065] In this embodiment, after obtaining the target alarm threshold, the target alarm threshold is updated as the current alarm threshold of the target electric circuit, thereby realizing adaptive adjustment of the alarm threshold.

[0066] Further, the agent model further comprises a residual service life output layer and a risk level output layer. Before the current multi-source feature information is input into the pre-set agent model, the agent model needs to be trained. Please refer to Figure 4 , and the training steps are as follows:

[0067] S150, acquire a training sample set, the training sample set comprising a plurality of training samples, and a true label of each of the training samples, the true label comprising an alarm threshold true value, a remaining useful life true value, and a risk level true value.

[0068] In this embodiment, the acquired training sample can be multi-source feature sample information or spatiotemporal fusion sample features (the sample is information obtained by spatiotemporal fusion analysis and processing of the multi-source feature sample information). When it is multi-source feature sample information, the multi-source feature sample information needs to be spatiotemporally fused and analyzed before the alarm threshold output layer, the remaining useful life output layer, and the risk level output layer are input in step S160, and then the processed spatiotemporal fusion sample features are input into each output layer.

[0069] S160, determine an alarm threshold predicted value of each of the training samples in the training sample set according to the alarm threshold output layer, determine a remaining useful life predicted value of each of the training samples in the training sample set according to the remaining useful life output layer, and determine a risk level predicted value of each of the training samples in the training sample set according to the risk level output layer.

[0070] S170, determine a first loss value according to the first loss function corresponding to the alarm threshold output layer, each of the alarm threshold true values, and the corresponding alarm threshold predicted values.

[0071] In some embodiments, step S170 comprises:

[0072] determine the first loss value based on a preset composite loss formula, the composite loss formula being as follows:

[0073] L thr =λ1·MSE+λ2·Penalty;

[0074] wherein L thr is the first loss value, MSE is a mean square error between each of the alarm threshold true values and the corresponding alarm threshold predicted values determined based on the first loss function, Penalty is a penalty term for each of the alarm threshold predicted values violating a preset physical safety constraint, and λ1 and λ2 are preset weight coefficients.

[0075] S180, determine a second loss value according to a second loss function corresponding to the remaining useful life output layer, each of the remaining useful life true values, and the corresponding remaining useful life predicted values.

[0076] wherein the second loss function is a regression loss function (Huber Loss), and the output is L reg .

[0077] S190, output a third loss function corresponding to the risk level output layer, each risk level real value and the corresponding risk level prediction value according to the risk level, and determine a third loss value.

[0078] The third loss function is a cross-entropy loss function, and the output is L cls .

[0079] S1100, determine a total loss value according to the first loss value, the second loss value and the third loss value.

[0080] The calculation formula of the total loss value is: L total = w1L cls + w2L reg + w3L thr .

[0081] The weights W1, W2 and W3 can be preset fixed weights, or can be dynamically adjusted according to the importance of the sample (for example, a device approaching the expiration date is given a higher weight).

[0082] S1110, train the alarm threshold output layer, the remaining useful life output layer and the risk level output layer according to the total loss value until the model converges.

[0083] Specifically, the training process is an automatic model building process of an agent according to features. The model output includes three sub-tasks, and for each sub-task, there are an alarm threshold output layer, a remaining useful life output layer and a risk level output layer.

[0084] In some embodiments, the method of the present application also provides visual decision tree auxiliary explanation, for example, automatically generating a contribution value (SHAP value) heat map to show which features contribute most to the current risk score; draw a simplified decision path: "device age > 7 years → temperature rise rate > 2℃ / min → trigger high risk (High Risk)"; facilitate non-technical personnel to understand the basis of system judgment, and enhance user trust.

[0085] The present application provides an intelligent monitoring system that integrates multi-source data, constructs a dynamic evolution model and realizes an adaptive early warning mechanism (the intelligent monitoring system is deployed in the computer device provided by the present application), which fundamentally improves the whole life cycle management efficiency of building pipeline systems and improves the accuracy of housing mechanical and electrical operation and maintenance.

[0086] In summary, the current multi-source feature information of the target electrical circuit is obtained, the current multi-source feature information includes basic parameters of each device in the target electrical circuit, real-time electrical parameters of each device, pipeline node temperature and environmental parameters; the current multi-source feature information is input into a preset agent model, the current multi-source feature information and historical multi-source feature information of the target electrical circuit within a preset first historical time length are analyzed and processed by a space-time attention mechanism in the agent model to obtain a space-time fusion feature of the target electrical circuit; the space-time fusion feature is input into an alarm threshold output layer in the agent model to output a target alarm threshold of the target electrical circuit; and the target electrical circuit is monitored according to the target alarm threshold. The alarm threshold of the electrical circuit can be adaptively adjusted by the preset agent model and the multi-source feature information of the electrical circuit, so that the accuracy of the mechanical and electrical equipment monitoring is improved.

[0087] Figure 5 is a schematic block diagram of an embodiment of the mechanical and electrical equipment monitoring device based on an agent provided by the present application. As shown in Figure 5 Corresponding to the above mechanical and electrical equipment monitoring method based on an agent, the present application also provides a mechanical and electrical equipment monitoring device 500 based on an agent. The mechanical and electrical equipment monitoring device 500 based on an agent includes units for executing the above mechanical and electrical equipment monitoring method based on an agent, and the mechanical and electrical equipment monitoring device 500 based on an agent can be configured in a terminal or a server. Specifically, please refer to Figure 5 The mechanical and electrical equipment monitoring device 500 based on an agent includes a transceiver unit 501 and a processing unit 502, wherein:

[0088] The transceiver unit 501 is configured to obtain current multi-source feature information of a target electrical circuit, and the current multi-source feature information includes basic parameters of each device in the target electrical circuit, real-time electrical parameters of each device, pipeline node temperature and environmental parameters.

[0089] The processing unit 502 is configured to input the current multi-source feature information into a preset agent model, analyze and process the current multi-source feature information and historical multi-source feature information of the target electrical circuit within a preset first historical time length by a space-time attention mechanism in the agent model to obtain a space-time fusion feature of the target electrical circuit; input the space-time fusion feature into an alarm threshold output layer in the agent model to output a target alarm threshold of the target electrical circuit; and monitor the target electrical circuit according to the target alarm threshold.

[0090] In some embodiments, the spatio-temporal attention mechanism comprises a time attention module and a space attention module, and the processing unit 502, in the step of performing spatio-temporal fusion analysis and processing on the current multi-source feature information and the historical multi-source feature information of the target electrical circuit within a preset first historical time length by the spatio-temporal attention mechanism in the agent model to obtain the spatio-temporal fusion feature of the target electrical circuit, is specifically configured to:

[0091] construct an input sequence according to the current multi-source feature information and the historical multi-source feature information;

[0092] analyze the input sequence by the time attention module to determine the time correlation feature of each device in the target electrical circuit;

[0093] analyze the input sequence by the space attention module to determine the space correlation feature between different devices in the target electrical circuit;

[0094] determine the spatio-temporal fusion feature according to the time correlation feature and the space correlation feature.

[0095] In some embodiments, the agent model further comprises a remaining useful life output layer and a risk level output layer, and the processing unit 502, after the step of performing spatio-temporal fusion analysis and processing on the current multi-source feature information and the historical multi-source feature information of the target electrical circuit within a preset first historical time length by the spatio-temporal attention mechanism in the agent model to obtain the spatio-temporal fusion feature of the target electrical circuit, is further configured to:

[0096] input the spatio-temporal fusion feature into the remaining useful life output layer to output the remaining useful life value of each device in the target electrical circuit;

[0097] input the spatio-temporal fusion feature into the risk level output layer to output the risk level of the target electrical circuit.

[0098] In some embodiments, the processing unit 502, after the step of inputting the spatio-temporal fusion feature into the risk level output layer in the agent model to output the risk level of each device in the target electrical circuit, is further configured to:

[0099] if the target electrical circuit is continuously determined as a medium risk level by the agent model for a preset number of times, adjust the current alarm threshold of the target electrical circuit downward according to a preset alarm parameter adjustment rule.

[0100] In some embodiments, the environmental parameter comprises environmental humidity, and the processing unit 502, after the step of obtaining the current multi-source feature information of the target electrical circuit, is further configured to:

[0101] if an increase value of ambient humidity in the ambient parameters within a preset second historical time length is greater than a preset humidity increase threshold, increasing an electric leakage current alarm threshold of the target electric circuit within a preset future time length.

[0102] In some embodiments, the agent model further includes a remaining service life output layer and a risk level output layer, and the processing unit 502, before performing the step of inputting the current multi-source feature information into the preset agent model, is further configured to:

[0103] The transceiving unit 501 acquires a training sample set, the training sample set including a plurality of training samples and a true label of each training sample, the true label including an alarm threshold true value, a remaining service life true value, and a risk level true value;

[0104] According to the alarm threshold output layer, an alarm threshold prediction value of each training sample in the training sample set is determined, according to the remaining service life output layer, a remaining service life prediction value of each training sample in the training sample set is determined, and according to the risk level output layer, a risk level prediction value of each training sample in the training sample set is determined;

[0105] According to the first loss function corresponding to the alarm threshold output layer, each alarm threshold true value, and the corresponding alarm threshold prediction value, a first loss value is determined;

[0106] According to the second loss function corresponding to the remaining service life output layer, each remaining service life true value, and the corresponding remaining service life prediction value, a second loss value is determined;

[0107] According to the third loss function corresponding to the risk level output layer, each risk level true value, and the corresponding risk level prediction value, a third loss value is determined;

[0108] According to the first loss value, the second loss value, and the third loss value, a total loss value is determined;

[0109] According to the total loss value, the alarm threshold output layer, the remaining service life output layer, and the risk level output layer are trained until the model converges.

[0110] In some embodiments, when performing the step of determining a first loss value according to the first loss function corresponding to the alarm threshold output layer, each alarm threshold true value, and the corresponding alarm threshold prediction value, the transceiving unit 501 is specifically configured to:

[0111] The first loss value is determined based on a preset composite loss formula, and the composite loss formula is as follows:

[0112] L thr = λ1·MSE + λ2·Penalty.

[0113] wherein, L thr is the first loss value, MSE is a mean square error between each of the alarm threshold real values and the corresponding alarm threshold predicted values determined based on the first loss function, Penalty is a penalty term for each of the alarm threshold predicted values violating a preset physical safety constraint, and λ1 and λ2 are preset weight coefficients.

[0114] In summary, the embodiment of the present application provides the agent-based electromechanical equipment monitoring device 500, which can adaptively adjust the alarm threshold of the electrical circuit through the preset agent model and the multi-source feature information of the electrical circuit, thereby improving the accuracy of electromechanical equipment monitoring.

[0115] It should be noted that the specific implementation process of the agent-based electromechanical equipment monitoring device 500 and each unit can be clearly understood by those skilled in the art, which can be referred to the corresponding description in the foregoing method embodiments. For the convenience and brevity of description, it will not be repeated here.

[0116] The agent-based electromechanical equipment monitoring device described above can be implemented in the form of a computer program, which can run on a computer device as shown in Figure 6 .

[0117] Please refer to Figure 6 , Figure 6 is a schematic block diagram of a computer device provided by an embodiment of the present application. The computer device 600 can be a terminal or a server, wherein the terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, a personal digital assistant, a wearable device, and an electronic device with a communication function. The server can be a stand-alone server or a server cluster composed of multiple servers.

[0118] Referring to Figure 6 , the computer device 600 includes a processor 602, a memory, and a network interface 605 connected through a system bus 601, wherein the memory can include a non-volatile storage medium 603 and an internal memory 604.

[0119] The non-volatile storage medium 603 can store an operating system 6031 and a computer program 6032. The computer program 6032 includes program instructions, which when executed, can cause the processor 602 to execute an agent-based electromechanical equipment monitoring method.

[0120] The processor 602 is configured to provide computing and control capabilities to support the operation of the entire computer device 600.

[0121] The internal memory 604 provides an environment for the operation of the computer program 6032 in the non-volatile storage medium 603, which, when executed by the processor 602, can cause the processor 602 to perform an agent-based electromechanical equipment monitoring method.

[0122] The network interface 605 is configured to communicate with other devices via a network. Those skilled in the art can understand that the network interface 605 can be implemented by a network card, a network adapter, a network chip, or the like. Figure 6 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device 600 to which the scheme of the present application is applied. The specific computer device 600 can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0123] The processor 602 is configured to run the computer program 6032 stored in the memory to implement the following steps:

[0124] Obtain the current multi-source feature information of the target electrical circuit, the current multi-source feature information including basic parameters of each device in the target electrical circuit, real-time electrical parameters of each device, pipeline node temperature, and environmental parameters;

[0125] Input the current multi-source feature information into a preset agent model, perform spatio-temporal fusion analysis and processing on the current multi-source feature information and historical multi-source feature information of the target electrical circuit within a preset first historical time length through a spatio-temporal attention mechanism in the agent model, and obtain a spatio-temporal fusion feature of the target electrical circuit;

[0126] Input the spatio-temporal fusion feature into an alarm threshold output layer in the agent model, and output a target alarm threshold of the target electrical circuit;

[0127] Monitor the target electrical circuit according to the target alarm threshold.

[0128] It should be understood that, in the embodiments of the present application, the processor 602 can be a central processing unit (CPU), and the processor 602 can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0129] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the above-mentioned embodiments.

[0130] Therefore, the present application also provides a storage medium. The storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program includes program instructions. The program instructions are executed by a processor to make the processor perform the following steps:

[0131] obtaining current multi-source feature information of a target electrical circuit, the current multi-source feature information including basic parameters of each device in the target electrical circuit, real-time electrical parameters of each device, pipeline node temperature, and environmental parameters;

[0132] inputting the current multi-source feature information into a preset agent model, performing spatio-temporal fusion analysis and processing on the current multi-source feature information and historical multi-source feature information of the target electrical circuit within a preset first historical time length through a spatio-temporal attention mechanism in the agent model, to obtain spatio-temporal fusion features of the target electrical circuit;

[0133] inputting the spatio-temporal fusion features into an alarm threshold output layer in the agent model, and outputting a target alarm threshold of the target electrical circuit;

[0134] monitoring the target electrical circuit according to the target alarm threshold.

[0135] The storage medium can be a U disk, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disk, and various computer readable storage media that can store program codes.

[0136] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been described in the above description in a general manner. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0137] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of each unit is only a logical function division, and actual implementation can have another division manner. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0138] The steps in the method embodiments of the present application can be adjusted, combined and reduced in sequence according to actual needs. The units in the device embodiments of the present application can be combined, divided and reduced according to actual needs. In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0139] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a storage medium. Based on such understanding, the technical solutions of the present application essentially or say the parts that make contributions to the prior art, or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a terminal or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application.

[0140] The above description is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for monitoring electromechanical equipment based on intelligent agents, characterized in that, The method includes: Obtain the current multi-source feature information of the target electrical circuit. The current multi-source feature information includes the basic parameters of each device in the target electrical circuit, the real-time electrical parameters of each device, the pipeline node temperature and environmental parameters. The basic parameters include: service life, standby power consumption, operating power consumption, cumulative running time and power-on time. The current multi-source feature information is input into a preset intelligent agent model. The spatiotemporal attention mechanism in the intelligent agent model is used to perform spatiotemporal fusion analysis on the current multi-source feature information and the historical multi-source feature information of the target electrical circuit within a preset first historical time period to obtain the spatiotemporal fusion feature of the target electrical circuit. The spatiotemporal fusion features are input into the alarm threshold output layer of the agent model to output the target alarm threshold of the target electrical circuit. Monitor the target electrical circuit according to the target alarm threshold; The agent model further includes a remaining lifetime output layer and a risk level output layer. Before inputting the current multi-source feature information into the preset agent model, the method further includes: Obtain a training sample set, which includes multiple training samples and a real label for each training sample. The real label includes the real value of the alarm threshold, the real value of the remaining service life, and the real value of the risk level. The alarm threshold prediction value for each training sample in the training sample set is determined based on the alarm threshold output layer, the remaining service life prediction value for each training sample in the training sample set is determined based on the remaining service life output layer, and the risk level prediction value for each training sample in the training sample set is determined based on the risk level output layer. The first loss value is determined based on the first loss function corresponding to the alarm threshold output layer, the actual values ​​of each alarm threshold, and the corresponding predicted values ​​of the alarm threshold. The second loss value is determined based on the second loss function corresponding to the remaining lifetime output layer, the actual value of each remaining lifetime, and the corresponding predicted value of the remaining lifetime. The third loss value is determined based on the third loss function corresponding to the risk level output layer, the true value of each risk level, and the predicted value of the corresponding risk level. The total loss value is determined based on the first loss value, the second loss value, and the third loss value. The alarm threshold output layer, the remaining lifespan output layer, and the risk level output layer are trained based on the total loss value until the model converges.

2. The method according to claim 1, characterized in that, The spatiotemporal attention mechanism includes a temporal attention module and a spatial attention module. The spatiotemporal fusion analysis of the current multi-source feature information and the historical multi-source feature information of the target electrical circuit within a preset first historical duration is performed using the spatiotemporal attention mechanism in the agent model to obtain the spatiotemporal fusion features of the target electrical circuit, including: Construct an input sequence based on the current multi-source feature information and the historical multi-source feature information; The input sequence is analyzed by the time attention module to determine the time correlation characteristics of each device in the target electrical circuit. The spatial attention module analyzes the input sequence to determine the spatial correlation characteristics between different devices in the target electrical circuit. The spatiotemporal fusion feature is determined based on the temporal correlation feature and the spatial correlation feature.

3. The method according to claim 1, characterized in that, The agent model further includes a remaining useful life output layer and a risk level output layer. After performing spatiotemporal fusion analysis on the current multi-source feature information and the historical multi-source feature information of the target electrical circuit within a preset first historical duration through the spatiotemporal attention mechanism in the agent model to obtain the spatiotemporal fusion features of the target electrical circuit, the method further includes: The spatiotemporal fusion features are input into the remaining useful life output layer to output the remaining useful life values ​​of each device in the target electrical circuit. The spatiotemporal fusion features are input into the risk level output layer to output the risk level of the target electrical circuit.

4. The method according to claim 3, characterized in that, After inputting the spatiotemporal fusion features into the risk level output layer of the agent model and outputting the risk level of each device in the target electrical circuit, the method further includes: If the target electrical circuit is determined to be of medium risk level by the intelligent agent model for a preset number of consecutive times, the current alarm threshold of the target electrical circuit is adjusted downward according to the preset alarm parameter adjustment rules.

5. The method according to claim 1, characterized in that, The environmental parameters include ambient humidity. After acquiring the current multi-source characteristic information of the target electrical circuit, the method further includes: If the increase in ambient humidity within a preset second historical time period is greater than a preset humidity increase threshold, then the leakage current alarm threshold of the target electrical circuit within a preset future time period is increased.

6. The method according to claim 1, characterized in that, The step of determining the first loss value based on the first loss function corresponding to the alarm threshold output layer, the actual values ​​of each alarm threshold, and the corresponding predicted values ​​of the alarm threshold includes: The first loss value is determined based on a preset composite loss formula, which is as follows: L thr =λ1·MSE+λ2·Penalty; Among them, L thr Let λ1 be the first loss value, MSE be the mean square error between the true value of each alarm threshold determined based on the first loss function and the corresponding predicted value of the alarm threshold, Penalty be the penalty term for each predicted value of the alarm threshold violating the preset physical security constraints, and λ1 and λ2 be preset weighting coefficients.

7. A monitoring device for electromechanical equipment based on intelligent agents, characterized in that, include: The transceiver unit is used to acquire the current multi-source feature information of the target electrical circuit. The current multi-source feature information includes the basic parameters of each device in the target electrical circuit, the real-time electrical parameters of each device, the pipeline node temperature, and the environmental parameters. The basic parameters include: service life, standby power consumption, operating power consumption, cumulative running time, and power-on time. The processing unit is configured to input the current multi-source feature information into a preset agent model, perform spatiotemporal fusion analysis on the current multi-source feature information and the historical multi-source feature information of the target electrical circuit within a preset first historical time period through the spatiotemporal attention mechanism in the agent model, and obtain the spatiotemporal fusion feature of the target electrical circuit; input the spatiotemporal fusion feature into the alarm threshold output layer in the agent model, and output the target alarm threshold of the target electrical circuit; and monitor the target electrical circuit according to the target alarm threshold. The agent model further includes a remaining lifetime output layer and a risk level output layer. Before performing the step of inputting the current multi-source feature information into the preset agent model, the processing unit is also used to: The training sample set is obtained through the transceiver unit. The training sample set includes multiple training samples and a real label for each training sample. The real label includes the real value of the alarm threshold, the real value of the remaining service life, and the real value of the risk level. The alarm threshold prediction value for each training sample in the training sample set is determined based on the alarm threshold output layer, the remaining service life prediction value for each training sample in the training sample set is determined based on the remaining service life output layer, and the risk level prediction value for each training sample in the training sample set is determined based on the risk level output layer. The first loss value is determined based on the first loss function corresponding to the alarm threshold output layer, the actual values ​​of each alarm threshold, and the corresponding predicted values ​​of the alarm threshold. The second loss value is determined based on the second loss function corresponding to the remaining lifetime output layer, the actual value of each remaining lifetime, and the corresponding predicted value of the remaining lifetime. The third loss value is determined based on the third loss function corresponding to the risk level output layer, the true value of each risk level, and the predicted value of the corresponding risk level. The total loss value is determined based on the first loss value, the second loss value, and the third loss value. The alarm threshold output layer, the remaining lifespan output layer, and the risk level output layer are trained based on the total loss value until the model converges.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the agent-based electromechanical equipment monitoring method as described in any one of claims 1-6.

9. A storage medium, characterized in that, The storage medium stores a computer program, which includes program instructions that, when executed by a processor, cause the processor to perform the agent-based electromechanical equipment monitoring method as described in any one of claims 1-6.

Images