CN118312860B

CN118312860B - Park energy monitoring method and system based on artificial intelligence

Info

Publication number: CN118312860B
Application number: CN202410742627.8A
Authority: CN
Inventors: 贺敬川; 徐斌; 康凤珠; 李勇; 吕楠; 李宏; 付国龙; 杨光熙
Original assignee: New Yingshun Information Technology Co ltd
Current assignee: New Yingshun Information Technology Co ltd
Priority date: 2024-06-11
Filing date: 2024-06-11
Publication date: 2024-08-23
Anticipated expiration: 2044-06-11
Also published as: CN118312860A

Abstract

The embodiment of the application relates to the technical field of artificial intelligence, in particular to a park energy monitoring method and system based on artificial intelligence, which are used for acquiring a plurality of batches of network debugging example tuples according to a plurality of first park energy monitoring data samples; vector mining is respectively carried out on the park energy monitoring data unit sample and the event hidden danger priori label in the network debugging example binary group through the energy hidden danger event judging network to obtain a park energy monitoring quantization vector and an event hidden danger labeling quantization vector, and a commonality score between the park energy monitoring quantization vector and the event hidden danger labeling quantization vector is determined; and circularly debugging the energy hidden danger event to judge the network according to the common scores and the set common thresholds of the multiple batches of network debugging example binary groups. Through the innovative data processing flow and debugging method, the performance of the energy hidden danger event discrimination network is obviously improved, and powerful technical support is provided for energy management and safety guarantee of a park.

Description

Park energy monitoring method and system based on artificial intelligence

Technical Field

The application relates to the technical field of artificial intelligence, in particular to a park energy monitoring method and system based on artificial intelligence.

Background

In the field of monitoring energy sources in parks, along with the development of intelligent technologies, the monitoring requirements on the energy source using efficiency and potential safety hazards are higher and higher. However, conventional energy monitoring systems often provide only basic data acquisition and simple threshold alarm functions, and have limited capability to identify complex energy change events and their potential hazards. This results in the fact that in practical applications, potential energy hazards often cannot be found and handled in time, increasing the risk and cost of park operations.

In order to solve this problem, in recent years, a method for discriminating an energy hidden trouble event based on deep learning has been attracting attention. However, the existing method often faces the problems of low accuracy, low efficiency and the like when processing complex energy monitoring data. This is mainly because existing methods lack the steps of deep mining and refinement of data while processing the data, and also lack efficient network debugging and optimization mechanisms.

Therefore, how to improve the accuracy and efficiency of judging the hidden energy events becomes a technical problem to be solved in the field of energy monitoring in the current park.

Disclosure of Invention

In order to overcome at least the above-mentioned shortcomings in the prior art, one of the purposes of the present application is to provide a method and a system for monitoring energy of a campus based on artificial intelligence.

The embodiment of the application provides a park energy monitoring method based on artificial intelligence, which is applied to a park energy monitoring system and comprises the following steps:

Acquiring a plurality of first park energy monitoring data samples, wherein the first park energy monitoring data samples all comprise target energy change events, each set of network monitoring data samples comprises a park energy monitoring data unit sample and event hidden danger priori labels, the park energy monitoring data unit sample comprises event state information of the target energy change events, and the event hidden danger priori labels are used for representing potential hidden danger characteristics of the event state information in the park energy monitoring data unit samples;

For each batch of network debugging example tuples, respectively carrying out vector mining on a park energy monitoring data unit sample and event hidden danger priori labels in the network debugging example tuples through an energy hidden danger event judging network to obtain a park energy monitoring quantization vector of the park energy monitoring data unit sample and an event hidden danger labeling quantization vector of the event hidden danger priori labels, determining a commonality score between the park energy monitoring quantization vector and the event hidden danger labeling quantization vector, wherein the energy hidden danger event judging network is used for judging the event state information of the target energy change event in the input park energy monitoring data unit;

and circularly debugging the energy hidden danger event discrimination network according to the common scores and the set common threshold values of the multiple batches of network debugging example binary groups.

Preferably, the obtaining multiple batches of network debugging example tuples according to multiple first campus energy monitoring data samples includes:

for each first campus energy monitoring data sample, respectively disassembling the first campus energy monitoring data samples according to a plurality of scale data capture cores to obtain respective campus energy monitoring data unit sets of the plurality of scales, wherein the plurality of campus energy monitoring data unit samples included in each of the plurality of scale campus energy monitoring data unit sets are of the scale;

And obtaining the multiple batches of network debugging example binary groups according to the park energy monitoring data unit samples in the park energy monitoring data unit set of the multiple first park energy monitoring data samples and event hidden danger priori labels corresponding to each park energy monitoring data unit sample.

Preferably, each park energy monitoring data unit sample corresponds to a plurality of event hidden danger priori labels, and the event hidden danger priori labels describe potential hidden danger characteristics of event state information in the park energy monitoring data unit sample by different event hidden danger attributes respectively;

vector mining is carried out on event hidden danger priori labels in the network debugging example binary group through the energy hidden danger event discrimination network to obtain event hidden danger label quantized vectors of the event hidden danger priori labels, and the method comprises the following steps:

And respectively carrying out vector mining on a plurality of event hidden danger priori labels in the network debugging example binary group through the energy hidden danger event judging network to obtain basic event hidden danger label quantized vectors respectively corresponding to the event hidden danger priori labels, and determining average vectors of the basic event hidden danger label quantized vectors to obtain the event hidden danger label quantized vectors.

Preferably, the method further comprises:

and updating a plurality of labeling information according to the event state information and the potential hidden danger trend value of the event state information in the park energy monitoring data unit sample to obtain a plurality of event hidden danger priori labels, wherein the potential hidden danger trend value is used for representing potential hidden danger characteristics of the event state information, and the plurality of labeling information are different.

Preferably, the method further comprises:

Acquiring a plurality of second-park energy monitoring data samples, wherein the second-park energy monitoring data samples all comprise the target energy change event;

And for each second-park energy monitoring data sample, disassembling the target energy change event from the second-park energy monitoring data sample to obtain the first-park energy monitoring data sample.

Preferably, the method further comprises:

acquiring a plurality of third-park energy monitoring data samples, wherein the plurality of third-park energy monitoring data samples all comprise the target energy change event;

And for each third campus energy monitoring data sample, correlating the distribution characteristics of the target energy change events in the third campus energy monitoring data sample based on the distribution characteristics of the target energy change events in the campus energy monitoring reference data to obtain the first campus energy monitoring data sample, wherein the distribution characteristics of the target energy change events in the first campus energy monitoring data sample are matched with the distribution characteristics of the target energy change events in the campus energy monitoring reference data.

Preferably, the method further comprises:

A plurality of park energy monitoring data units for acquiring park energy monitoring data, wherein the park energy monitoring data comprise target energy change events, and each park energy monitoring data unit comprises event state information of the target energy change events;

For each park energy monitoring data unit, determining a target mark corresponding to the park energy monitoring data unit through the energy hidden danger event discrimination network, wherein the target mark is used for representing potential hidden danger characteristics of event state information of the target energy change event in the park energy monitoring data unit;

And determining potential hidden danger trend values of the target energy change events in the park energy monitoring data according to the target labels respectively corresponding to the park energy monitoring data units, wherein the potential hidden danger trend values are used for representing potential hidden danger characteristics of the target energy change events.

Preferably, the determining the potential hidden danger trend value of the target energy change event in the park energy monitoring data according to the target labels corresponding to the plurality of park energy monitoring data units respectively includes:

Determining that a target energy change event in the park energy monitoring data has hidden danger risk and determining event state information of the target energy change event has hidden danger risk on the basis that a target label corresponding to at least one park energy monitoring data unit in the plurality of park energy monitoring data units indicates that the event state information of the target energy change event has hidden danger risk;

And determining that the target energy change event in the park energy monitoring data does not have hidden danger risk on the basis that the target label corresponding to each park energy monitoring data unit in the park energy monitoring data units reflects that the event state information of the target energy change event does not have hidden danger risk.

Preferably, the determining, for each energy monitoring data unit of the park, the target label corresponding to the energy monitoring data unit of the park through the energy hidden trouble event discrimination network includes:

For each park energy monitoring data unit, extracting park energy monitoring quantized vectors of the park energy monitoring data units through the energy hidden danger event discrimination network, determining commonality scores between the park energy monitoring quantized vectors and a plurality of set event hidden danger labeling quantized vectors respectively, and determining target event hidden danger labeling quantized vectors with commonality scores meeting setting requirements from the set event hidden danger labeling quantized vectors, wherein the set event hidden danger labeling quantized vectors respectively correspond to set labels, and the target labels are set labels corresponding to the target event hidden danger labeling quantized vectors.

Preferably, the plurality of campus energy monitoring data units for acquiring the campus energy monitoring data include:

And respectively disassembling the park energy monitoring data according to the data capturing cores of a plurality of scales to obtain park energy monitoring data unit sets of the respective scales, wherein the park energy monitoring data unit sets of each scale comprise a plurality of park energy monitoring data units of the scale.

Preferably, each campus energy monitoring data unit includes a plurality of monitoring fields, and determining a potential hidden danger trend value of a target energy change event in the campus energy monitoring data according to target labels corresponding to the plurality of campus energy monitoring data units, including:

For each park energy monitoring data unit of each scale, adding a commonality score between a park energy monitoring quantized vector of the park energy monitoring data unit and a corresponding target event hidden danger labeling quantized vector to a plurality of monitoring fields in the park energy monitoring data unit on the basis that a target label corresponding to the park energy monitoring data unit indicates that event state information of the target energy change event in the park energy monitoring data unit has hidden danger risk;

For each monitoring field, obtaining a discrimination weight of the monitoring field according to the commonality scores of the monitoring field under the scales, wherein the discrimination weight is used for indicating the possibility that the monitoring field has hidden danger;

And determining potential hidden danger trend values of the target energy change events in the park energy monitoring data according to the discrimination weight of each monitoring field in the park energy monitoring data, wherein the potential hidden danger trend values of the target energy change events comprise at least one of the distribution characteristics and hidden danger grades of the monitoring fields of the target energy change events with hidden danger risks.

The embodiment of the application also provides a park energy monitoring system, which comprises a processor, and a memory and a bus which are connected with the processor; wherein the processor and the memory complete communication with each other through the bus; the processor is configured to invoke the program instructions in the memory to perform the artificial intelligence based campus energy monitoring method described above.

The embodiment of the application also provides a computer readable storage medium, wherein a program is stored on the computer readable storage medium, and the program is executed by a processor to realize the park energy monitoring method based on artificial intelligence.

Advantageous effects

According to the embodiment of the application, a plurality of batches of network debugging example tuples are obtained according to a plurality of first park energy monitoring data samples; vector mining is respectively carried out on the park energy monitoring data unit sample and the event hidden danger priori label in the network debugging example binary group through the energy hidden danger event judging network, so that a park energy monitoring quantized vector of the park energy monitoring data unit sample and an event hidden danger labeling quantized vector of the event hidden danger priori label are obtained, and a commonality score between the park energy monitoring quantized vector and the event hidden danger labeling quantized vector is determined; and circularly debugging the energy hidden danger event to judge the network according to the common scores and the set common thresholds of the multiple batches of network debugging example binary groups. Through the innovative data processing flow and debugging method, the performance of the energy hidden danger event discrimination network is obviously improved, and powerful technical support is provided for energy management and safety guarantee of a park.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flowchart of a method for monitoring energy of a campus based on artificial intelligence according to an embodiment of the present application.

Fig. 2 is a block diagram of a campus energy monitoring system according to an embodiment of the present application.

Icon:

100-park energy monitoring system;

A 101-processor; 102-memory; 103-bus.

Detailed Description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the application to those skilled in the art.

In order to better understand the above technical solutions, the following detailed description of the technical solutions of the present application is made by using the accompanying drawings and specific embodiments, and it should be understood that the specific features of the embodiments and the embodiments of the present application are detailed descriptions of the technical solutions of the present application, and not limiting the technical solutions of the present application, and the technical features of the embodiments and the embodiments of the present application may be combined with each other without conflict.

Fig. 1 is a flowchart of an artificial intelligence based method for monitoring energy of a campus, which is applied to a system for monitoring energy of a campus, according to an embodiment of the present application, and includes S101-S103.

S101, acquiring a plurality of batches of network debugging example tuples according to a plurality of first campus energy monitoring data samples, wherein the plurality of first campus energy monitoring data samples all comprise target energy change events, each batch of network debugging example tuples comprises a campus energy monitoring data unit sample and event hidden danger priori labels, the campus energy monitoring data unit sample comprises event state information of the target energy change events, and the event hidden danger priori labels are used for representing potential hidden danger characteristics of the event state information in the campus energy monitoring data unit sample.

S102, for each batch of network debugging example tuples, vector mining is respectively carried out on a park energy monitoring data unit sample and event hidden danger priori labels in the network debugging example tuples through an energy hidden danger event judging network to obtain a park energy monitoring quantization vector of the park energy monitoring data unit sample and an event hidden danger labeling quantization vector of the event hidden danger priori labels, a commonality score between the park energy monitoring quantization vector and the event hidden danger labeling quantization vector is determined, and the energy hidden danger event judging network is used for judging the energy hidden danger event according to event state information of the target energy change event in the input park energy monitoring data unit.

S103, circularly debugging the energy hidden danger event discrimination network according to the common scores and the set common thresholds of the multiple batches of network debugging example binary groups.

The park energy monitoring system in the embodiment of the application is a set of highly integrated and intelligent management platform which is specially designed for monitoring various energy use conditions of the park in real time and identifying and preventing potential energy use hidden danger through an advanced data analysis technology. The following is a specific scenario of the system in practical applications.

First, in S101, the campus energy monitoring system obtains a plurality of first campus energy monitoring data samples including the target energy change event from the history database. These data samples may include grid voltage fluctuations, water line pressure changes, abrupt changes in fuel gas usage, etc. The system may extract a plurality of network debug example doublets from the data. Each binary group consists of a park energy monitoring data unit sample and corresponding event hidden danger priori labels. The data unit sample records the specific state information of the target energy change event, such as the occurrence time, the change amplitude and the like; the prior labeling of the hidden danger of the event is to pre-label hidden danger features possibly hidden in the state information according to historical experience and expert knowledge.

Next, in S102, the campus energy monitoring system uses a deep learning model called "energy hidden trouble event discrimination network". For each batch of network debugging example doublets, the model can respectively carry out depth vector mining on the park energy monitoring data unit samples and event hidden danger priori labels. This process converts these data and labels into a mathematical vector form that is easier for a computer to understand and compare. Through excavation, the system can obtain a park energy monitoring quantization vector and an event hidden danger labeling quantization vector. The system then calculates a commonality score between the two vectors that reflects the similarity or association between the event state information and the hidden danger features in the data unit samples.

Finally, in S103, the campus energy monitoring system performs cyclic debugging on the energy hidden danger event discrimination network according to the previously calculated commonality score and a set commonality threshold. If the commonality score is lower than the set threshold, the identification capacity of the network is still to be improved, and the system can adjust the parameters and the structure of the network according to the difference between the score and the threshold, so that potential energy use hidden danger can be identified more accurately. This process is repeated until the performance of the network reaches a satisfactory level.

Through the process, the energy monitoring system of the park can continuously self-optimize and promote, so that the energy service condition of the park can be monitored and managed more intelligently, potential energy hidden danger events can be found and prevented in time, and the energy use safety and high efficiency of the park are ensured.

In combination with the above technical solution, the first campus energy monitoring data sample refers to a specific data sample collected from the campus energy monitoring system, and records the energy use condition of the garden within a certain period of time. The data samples comprise various energy change events, such as increase and decrease of power load, consumption change of water resources and the like, and are basic data for system analysis and debugging.

An exemplary tuple of network debug is a data pair consisting of two parts: firstly, a park energy monitoring data unit sample and secondly, event hidden danger priori labeling corresponding to the park energy monitoring data unit sample. The binary set is used for training and optimizing the energy hidden trouble event discrimination network so as to improve the identification accuracy of the energy hidden trouble event.

The target energy change event refers to a change in energy status of particular interest in the campus energy monitoring data, such as a sudden power peak, abnormal fluctuation of water flow, etc. These events may be indicative of equipment failure, energy waste, or safety concerns.

The campus energy monitoring data unit sample is a representative data segment extracted from the first campus energy monitoring data sample, and the representative data segment records the specific states of the target energy change event, such as time, amplitude, duration and the like, and is an important information source for analyzing the energy use condition and identifying hidden danger.

Event hidden danger priori labeling is pre-labeling of hidden danger which may exist in a data unit sample based on expert knowledge or historical experience. These labels help the system identify and understand which energy change events may hide potential problems or risks.

The event state information refers to specific state descriptions of the target energy source change event, such as specific values of voltage, flow rate of water flow and the like. The information reflects the real-time condition of energy use and is an important basis for judging whether hidden danger exists or not.

Potential hidden trouble features refer to some features hidden in energy change events that are not easily directly observed, and may be indicative of equipment failure, energy inefficiency, or safety issues, etc. The system needs to identify these features through learning and analysis in order to take timely action to prevent potential problems.

The energy hidden trouble event judging network is a deep learning model and is specially used for analyzing and judging whether hidden trouble events exist in the incoming park energy monitoring data. Through extensive data training, the network is able to automatically identify and classify various energy change events and predict the risk that they may carry.

The campus energy monitoring quantization vector is a mathematical vector form that converts the campus energy monitoring data into a computer understandable and operable. By this quantitative representation, the system is able to more conveniently perform data analysis and comparison, thereby more accurately identifying potential hidden danger events.

Event hidden danger labeling quantization vector: similar to the campus energy monitoring quantization vector, this is the conversion of event hidden danger a priori labels into a form of mathematical vectors. The quantitative representation is beneficial to matching and comparing the labeling information with the actual energy monitoring data by the system, and improves the accuracy of hidden danger identification.

The commonality score is an index for measuring the similarity or the relevance between the park energy monitoring quantized vector and the event hidden danger labeling quantized vector. The higher the score, the greater the commonality between the two, namely the more the actually monitored energy change event is matched with the pre-marked hidden danger feature.

The energy hidden trouble event judgment refers to the process of analyzing and judging the energy monitoring data of the park which is input by utilizing the energy hidden trouble event judgment network. Through the process, the system can automatically identify and classify various potential hidden danger events, and provides important basis for subsequent early warning and countermeasure.

In detail, the campus energy monitoring system first retrieves a plurality of first campus energy monitoring data samples containing the target energy change event from its database. These data samples may come from different time periods, different monitoring points, but all contain valuable energy change information. The system then extracts critical information from these data samples to construct a network debug example doublet. Each binary group consists of a park energy monitoring data unit sample and an event hidden danger priori label corresponding to the park energy monitoring data unit sample. The data unit sample records specific state information of a target energy change event in detail, such as abrupt change of voltage, abnormality of water flow and the like; the event hidden danger priori labeling is pre-judging the hidden danger possibly existing in the state information based on expert knowledge or historical experience. After the network debug example doublet is built, the campus energy monitoring system can utilize the energy hidden trouble event discrimination network to conduct deep analysis on the data. First, the network will vector mine the data unit samples and event hidden danger prior labels, i.e., convert them into a mathematical vector form that is easier for the computer to process. These vectors not only contain all the information of the original data, but also highlight the key features therein by specific algorithms. The system then calculates a commonality score between the two vectors. This score is actually an indicator of the similarity or relevance of the two: if the score is high, the fact that the actually monitored energy change event is highly consistent with the pre-marked hidden danger feature is indicated, and the network has strong identification capability on the event; otherwise, it is stated that the network requires further optimization and tuning. And according to the calculated commonality score, the park energy monitoring system can carry out cyclic debugging on the energy hidden danger event discrimination network. Specifically, the system sets a commonality threshold as the target for debugging: if the commonality score of a certain binary group is below this threshold, it is stated that the network's ability to identify such energy change events remains to be improved. At this time, the system adjusts the parameters and structure of the network according to the difference between the score and the threshold value, so that the system can more accurately identify the potential energy hidden trouble event. This process is repeated a number of times until the performance of the network reaches a satisfactory level. The energy monitoring system of the park can continuously self-optimize and promote in a circulating debugging mode, so that the energy service condition of the park can be monitored and managed more intelligently, potential energy hidden danger events can be found and prevented in time, and the energy use safety and high efficiency of the park are ensured.

According to the embodiment of the application, the accuracy and the efficiency of distinguishing the energy hidden danger event are obviously improved through a refined data processing flow. Firstly, a plurality of first park energy monitoring data samples containing target energy change events are collected to construct a plurality of network debugging example doublets, and rich and targeted learning materials are provided for subsequent network training. Each binary group fuses actual monitoring data and priori labels of potential hidden danger features, so that the model can more accurately capture subtle relations between energy change events and hidden troubles in the learning process.

Further, the embodiment of the application utilizes the energy hidden trouble event discrimination network to deeply mine the carefully prepared binary data. And vector mining is carried out on the data unit sample and the event hidden danger priori label through a network, so that a quantized vector capable of accurately representing the energy monitoring data and hidden danger features is obtained. The quantization mode not only enables comparison and calculation between data to be more convenient, but also greatly improves the sensitivity of hidden danger identification.

In addition, by calculating the commonality score between the park energy monitoring quantization vector and the event hidden danger labeling quantization vector, an objective and quantifiable index is provided for debugging and optimizing the discrimination network. The debugging method based on the commonality score ensures that the optimization process of the network is more scientific and efficient, thereby ensuring the accuracy of judging the hidden energy events.

Finally, through continuous cyclic debugging, the performance of the energy hidden danger event discrimination network can be continuously perfected and improved according to the commonality score and the set commonality threshold, so that the energy hidden danger identification capability can still be maintained efficiently and accurately when the energy hidden danger event discrimination network faces to complex and changeable energy monitoring data.

In summary, the embodiment of the application remarkably improves the accuracy and efficiency of the network for judging the potential energy events through the innovative data processing flow and debugging method, and provides powerful technical support for energy management and safety guarantee of the park.

In some alternative embodiments, the obtaining a plurality of network debug example doublets from a plurality of first campus energy monitoring data samples includes: for each first campus energy monitoring data sample, respectively disassembling the first campus energy monitoring data samples according to a plurality of scale data capture cores to obtain respective campus energy monitoring data unit sets of the plurality of scales, wherein the plurality of campus energy monitoring data unit samples included in each of the plurality of scale campus energy monitoring data unit sets are of the scale; and obtaining the multiple batches of network debugging example binary groups according to the park energy monitoring data unit samples in the park energy monitoring data unit set of the multiple first park energy monitoring data samples and event hidden danger priori labels corresponding to each park energy monitoring data unit sample.

In this embodiment, in the optimization process of performing the energy hidden trouble event discrimination, the campus energy monitoring system first obtains a plurality of network debug example tuples according to a plurality of first campus energy monitoring data samples. This process is elaborate and critical, involving the disassembly of the raw data and the combination with a priori knowledge.

Specifically, for each first campus energy monitoring data sample, the campus energy monitoring system may utilize multiple data capture cores of different sizes for disassembly. These data capture kernels, like a pair of "rules" of varying precision, are capable of parsing data samples on different data scales or time scales, respectively. In this way, the system can obtain a plurality of individual campus energy monitoring data unit sets of scale. The "scale" herein is understood to mean the length of time of data, the number of data points, or the degree of refinement of data, etc.

In each scale of the campus energy monitoring data unit set, a plurality of campus energy monitoring data unit samples are included, and the samples are unified into the scale. This means that the system can capture corresponding energy change events, both on a larger time scale and on a smaller time scale, providing a rich data base for subsequent analysis.

Next, the campus energy monitoring system constructs a network debugging example binary set according to the disassembled data unit samples and the prior labels of the event hidden dangers corresponding to the disassembled data unit samples. These prior annotations are derived based on expert knowledge or historical data analysis, which provide potential hidden danger features for event state information in data unit samples. In this way, each binary group contains actual monitoring data and corresponding hidden danger feature labels, and an accurate reference is provided for subsequent network debugging and optimization.

In this way, the campus energy monitoring system can obtain multiple batches of targeted network debugging example doublets. The binary groups not only reflect the actual condition of energy use, but also integrate the prior knowledge of experts, and provide powerful data support for the training and optimization of the subsequent energy hidden trouble event discrimination network.

Thus, the park energy monitoring system can realize the deep disassembly of the original energy monitoring data and the effective combination of priori knowledge. In the process, the data is captured and checked by utilizing a plurality of scales of data to disassemble, so that the system can capture the detailed characteristics of the energy change event on different data scales. Meanwhile, by combining with the prior labeling of the event hidden danger, the system can more accurately understand hidden danger features possibly hidden behind the change events. The comprehensive processing method remarkably improves the accuracy and efficiency of distinguishing the energy hidden danger event. Firstly, through multi-angle and multi-scale data disassembly, the system can more comprehensively understand the actual situation of energy use; secondly, the introduction of priori labels enables the system to integrate the knowledge and experience of an expert in the judging process, so that the judging accuracy is improved; finally, by means of the combination of the refined data processing and the labeling, the energy monitoring system of the park can train and optimize the energy hidden danger event discrimination network more effectively, and further the identification and early warning capability of the energy hidden danger monitoring system to the potential energy hidden danger is improved.

In some alternative embodiments, each campus energy monitoring data unit sample corresponds to a plurality of event hidden danger prior labels, and the plurality of event hidden danger prior labels describe potential hidden danger characteristics of event state information in the campus energy monitoring data unit sample with different event hidden danger attributes respectively; vector mining is carried out on event hidden danger priori labels in the network debugging example binary group through the energy hidden danger event discrimination network to obtain event hidden danger label quantized vectors of the event hidden danger priori labels, and the method comprises the following steps: and respectively carrying out vector mining on a plurality of event hidden danger priori labels in the network debugging example binary group through the energy hidden danger event judging network to obtain basic event hidden danger label quantized vectors respectively corresponding to the event hidden danger priori labels, and determining average vectors of the basic event hidden danger label quantized vectors to obtain the event hidden danger label quantized vectors.

In some exemplary applications of the campus energy monitoring system, each of the campus energy monitoring data unit samples corresponds not only to one event hidden danger apriori label, but to a plurality of event hidden dangers priors labels. These labels describe the potential hidden danger features of event state information in a data unit sample from different angles or attributes. For example, a data unit sample may be simultaneously labeled with "voltage fluctuation anomaly", "temperature rise trend", and other hidden trouble attributes.

When the campus energy monitoring system processes the data unit samples with a plurality of event hidden danger priori labels, the data unit samples can utilize the energy hidden danger event discrimination network to deeply mine vectors of the labels. This process aims to translate these text-based annotations into a mathematical form that the computer can understand and manipulate, i.e., a quantized vector.

Specifically, the energy hidden danger event discrimination network may first perform vector mining on prior labels of each hidden danger event, and generate corresponding quantized vectors of labels of hidden danger of basic events. These basic quantization vectors represent the feature direction and intensity of each hidden danger attribute in a multidimensional space. The system then calculates the average of these basic quantized vectors to obtain a comprehensive event hidden danger labeling quantized vector. The comprehensive quantization vector fuses information of a plurality of hidden danger attributes, and can reflect potential hidden danger characteristics in the data unit sample more comprehensively.

By the mode, the park energy monitoring system can process complex and variable energy monitoring data and extract rich hidden danger information from the complex and variable energy monitoring data. The system not only improves the monitoring capability of the system on the energy use state, but also provides more accurate data support for subsequent hidden danger early warning and fault processing.

Therefore, the energy monitoring system for the park remarkably improves the capability of the energy monitoring system for the park for processing complex energy data by allowing each sample of the energy monitoring data unit for a plurality of event hidden danger priori labels and utilizing the energy hidden danger event discrimination network to conduct deep vector mining on the labels. The multi-label and multi-attribute data processing mode enables the system to more comprehensively capture and understand various potential hidden dangers in energy use, and accordingly accuracy and sensitivity of hidden danger identification are improved. In addition, the average value of the plurality of basic event hidden danger labeling quantization vectors is calculated to obtain the comprehensive event hidden danger labeling quantization vector, and the system further enhances the data representation capability. The comprehensive quantization vector not only fuses information of a plurality of hidden danger attributes, but also reduces misjudgment risks possibly brought by a single attribute, so that hidden danger identification results of the system are more stable and reliable. In summary, the embodiment remarkably improves the performance of the energy monitoring system of the park in the aspect of energy hidden danger identification by introducing a multi-label and multi-attribute data processing mode and a comprehensive quantization vector calculation method, and provides more powerful technical support for energy management and safety guarantee of the park

In some alternative embodiments, the method further comprises: for each park energy monitoring data unit sample, updating a plurality of labeling information according to event state information and potential hidden danger trend values of the event state information in the park energy monitoring data unit sample to obtain a plurality of event hidden danger priori labels, wherein the potential hidden danger trend values are used for representing potential hidden danger characteristics of the event state information, and the plurality of labeling information are different;

Or the method further comprises: acquiring a plurality of second-park energy monitoring data samples, wherein the second-park energy monitoring data samples all comprise the target energy change event; for each second-park energy monitoring data sample, disassembling the target energy change event from the second-park energy monitoring data sample to obtain the first-park energy monitoring data sample;

Or the method further comprises: acquiring a plurality of third-park energy monitoring data samples, wherein the plurality of third-park energy monitoring data samples all comprise the target energy change event; and for each third campus energy monitoring data sample, correlating the distribution characteristics of the target energy change events in the third campus energy monitoring data sample based on the distribution characteristics of the target energy change events in the campus energy monitoring reference data to obtain the first campus energy monitoring data sample, wherein the distribution characteristics of the target energy change events in the first campus energy monitoring data sample are matched with the distribution characteristics of the target energy change events in the campus energy monitoring reference data.

Based on this embodiment, the functionality of the campus energy monitoring system is further extended and enhanced. The embodiments not only cover the data processing and hidden trouble recognition flow described before, but also introduce new data updating and expanding mechanisms to continuously improve the accuracy and adaptability of the system.

First, the system updates a plurality of labeling information according to the event status information and the potential hidden trouble trend value of each campus energy monitoring data unit sample. The potential risk trend value is a key indicator used to quantify hidden risk features in the event state information. Based on this trend value, the system can dynamically adjust or update the prior event hidden danger prior label. The dynamic updating mechanism enables the system to adapt to the change of the energy use state in real time, so that the timeliness of hidden danger identification is improved.

Second, the system may also expand its data set by obtaining multiple second campus energy monitoring data samples. These second data samples also contain target energy change events. For each second data sample, the system disassembles the target energy change event therein to obtain a new first campus energy monitoring data sample. The method for disassembling and expanding the data not only increases the data volume of the system, but also improves the diversity of the data, and is beneficial to the system to better learn and identify various energy change modes.

In addition, the system may also obtain and process third campus energy monitoring data samples by correlating to the target energy change event distribution characteristics in the campus energy monitoring reference data. Specifically, the system analyzes the distribution characteristics of the target energy change events in the reference data and based thereon de-correlates and processes the corresponding events in the third data sample. Through this correlation process, the system is able to extract a first data sample that matches the reference data distribution characteristics. The method not only improves the accuracy of the data, but also is helpful for the system to better understand and simulate the actual energy use scene.

By introducing the embodiment, the park energy monitoring system realizes dynamic update of data annotation, expansion of a data set and association processing with reference data, thereby remarkably improving the performance and accuracy of the system. The mechanism for dynamically updating the labeling information enables the system to capture and adapt to the change of the energy state in real time, and the timeliness and accuracy of hidden danger identification are improved. Meanwhile, by disassembling the second data sample and processing the third data sample in an associated manner, the system not only increases the data quantity and diversity, but also improves the accuracy and representativeness of the data, and provides a richer and reliable data basis for the subsequent discrimination of the hidden energy events. In summary, these embodiments collectively enhance the capability of the campus energy monitoring system in terms of energy management and safety assurance.

In other possible embodiments, the method further comprises: a plurality of park energy monitoring data units for acquiring park energy monitoring data, wherein the park energy monitoring data comprise target energy change events, and each park energy monitoring data unit comprises event state information of the target energy change events; for each park energy monitoring data unit, determining a target mark corresponding to the park energy monitoring data unit through an energy hidden danger event discrimination network, wherein the target mark is used for representing potential hidden danger characteristics of event state information of the target energy change event in the park energy monitoring data unit; and determining potential hidden danger trend values of the target energy change events in the park energy monitoring data according to the target labels respectively corresponding to the park energy monitoring data units, wherein the potential hidden danger trend values are used for representing potential hidden danger characteristics of the target energy change events.

In other possible embodiments, the campus energy monitoring system further refines the process and analysis flow of the energy data. This process begins with the acquisition of raw campus energy monitoring data, which typically includes various energy change events such as voltage fluctuations, power changes, etc.

First, the system will obtain a plurality of campus energy monitoring data units for these raw data. Each data unit covers event status information of a target energy change event, which is the basis for understanding energy usage and identifying potential hazards.

Next, for each acquired energy monitoring data unit of the campus, the system determines its corresponding target label using the trained energy hidden trouble event discrimination network. This target annotation is an important indicator for revealing potentially hidden features behind event state information of a target energy change event in a data unit. For example, if a data cell shows a sudden drop in voltage, its target label may point to a "voltage anomaly", indicating that this may be a potential point of hidden danger.

After the target labels for each data unit are determined, the campus energy monitoring system further analyzes potential trend values of the target energy change events in the whole campus energy monitoring data according to the labels. The trend value is a comprehensive index, and the potential hidden danger characteristics of the target energy change event are reflected on the basis of target labels of a plurality of data units. For example, if the target labels of multiple data units all point to a voltage problem, the potential hazard trend value may indicate that there may be a greater potential hazard in terms of voltage.

And finally, the potential hidden danger trend value can be used for subsequent hidden danger early warning, fault removal and other links by the system, and powerful support is provided for the safe use of energy sources in the park.

In this way, the campus energy monitoring system can effectively identify and analyze potential hazards in energy data through refined data processing flows. From the acquisition of the original data unit, the determination of target labels by utilizing a discrimination network, and the calculation of potential hidden danger trend values based on the labels, each step is closely connected, so that an efficient and accurate energy hidden danger recognition system is formed. The method not only improves the safety and efficiency of energy use in the park, but also provides powerful data support for timely preventing and solving the energy problem. Overall, these embodiments significantly enhance the capability of the campus energy monitoring system in terms of hidden trouble identification and analysis.

In some preferred embodiments, the determining the potential hidden danger trend value of the target energy change event in the campus energy monitoring data according to the target labels corresponding to the plurality of campus energy monitoring data units respectively includes: determining that a target energy change event in the park energy monitoring data has hidden danger risk and determining event state information of the target energy change event has hidden danger risk on the basis that a target label corresponding to at least one park energy monitoring data unit in the plurality of park energy monitoring data units indicates that the event state information of the target energy change event has hidden danger risk; and determining that the target energy change event in the park energy monitoring data does not have hidden danger risk on the basis that the target label corresponding to each park energy monitoring data unit in the park energy monitoring data units reflects that the event state information of the target energy change event does not have hidden danger risk.

In some more preferred embodiments, the campus energy monitoring system uses a more detailed and accurate approach in determining potential trend values for target energy change events in the campus energy monitoring data. The process mainly involves comprehensive analysis of multiple campus energy monitoring data units and their corresponding target labels.

First, the system will check the target labels for each of the campus energy monitoring data units. The target labels are obtained through an energy hidden danger event judging network and can indicate whether event state information of a target energy change event in the data unit has hidden danger or not.

If the system finds that, among the plurality of campus energy monitoring data units, the target label for at least one data unit indicates that the event status information thereof is at risk, the system further analyzes the data units. By comprehensively considering these data units with risk labels, the system can determine that the target energy change event in the campus energy monitoring data does have a risk of hidden danger, and can explicitly indicate which specific event state information is at risk.

In contrast, if the system detects that the target labels corresponding to all the campus energy monitoring data units indicate that the event state information of the target labels does not have hidden danger risks, the system can determine that the target energy change event in the campus energy monitoring data is safe and does not have hidden danger risks.

The comprehensive analysis method based on the multiple data units and the target labels not only improves the accuracy of hidden danger identification, but also provides a more detailed risk assessment result for the system.

Therefore, the park energy monitoring system can accurately judge potential hidden danger trend values of the target energy change event through deep analysis of the plurality of park energy monitoring data units and the target labels thereof. The method not only can timely find and identify hidden danger risks, but also can provide clear directions and targets for subsequent hidden danger treatment. Overall, these embodiments significantly improve the accuracy and reliability of the campus energy monitoring system in terms of risk assessment and hidden trouble identification, thereby more effectively guaranteeing the safety and stability of the use of the campus energy.

In some preferred embodiments, the determining, for each energy monitoring data unit of the campus, the target label corresponding to the energy monitoring data unit of the campus through the energy hidden trouble event discrimination network includes: for each park energy monitoring data unit, extracting park energy monitoring quantized vectors of the park energy monitoring data units through the energy hidden danger event discrimination network, determining commonality scores between the park energy monitoring quantized vectors and a plurality of set event hidden danger labeling quantized vectors respectively, and determining target event hidden danger labeling quantized vectors with commonality scores meeting setting requirements from the set event hidden danger labeling quantized vectors, wherein the set event hidden danger labeling quantized vectors respectively correspond to set labels, and the target labels are set labels corresponding to the target event hidden danger labeling quantized vectors.

In some more specific preferred embodiments, the campus energy monitoring system uses an energy risk event discrimination network to determine target labels for each of the campus energy monitoring data units, a process that involves a number of critical steps.

First, for each campus energy monitoring data unit, the system will input it into the energy hidden trouble event discrimination network. This discrimination network is trained to identify potential hidden trouble events in the energy monitoring data.

Then, the discrimination network extracts the campus energy monitoring quantization vector of the campus energy monitoring data unit. This quantization vector is a characteristic representation of the data unit that captures critical information in the data, such as changes in voltage, current, power, etc.

The system then determines a commonality score between the campus energy monitoring quantization vector and a plurality of preset event hidden danger tagging quantization vectors. The set event hidden danger labeling quantization vectors are predefined and represent the characteristics of various possible hidden danger events. The commonality score reflects the similarity between the quantized vectors of the data units and the preset hidden danger labeling quantized vectors.

Then, the system can find out a target event hidden danger labeling quantization vector with the commonality score meeting the setting requirement from a plurality of setting event hidden danger labeling quantization vectors. The set requirement may be a threshold value, and the corresponding set event hidden danger tagging quantization vector may be selected as the target event hidden danger tagging quantization vector only if the commonality score exceeds the threshold value.

Finally, as each set event hidden danger labeling quantization vector corresponds to one set label, the system can take the set label corresponding to the selected target event hidden danger labeling quantization vector as the target label of the energy monitoring data unit of the park.

Thus, the campus energy monitoring system can accurately determine the target label of each of the campus energy monitoring data units by using the energy hidden trouble event discrimination network. The method captures the core characteristics of the data in a mode of quantization vectors, and accurately identifies potential hidden danger events in the data by comparing the core characteristics with the preset hidden danger labeling quantization vectors. The hidden danger identification method not only improves the accuracy and efficiency of hidden danger identification, but also provides powerful data support for subsequent hidden danger processing. Overall, the embodiments significantly enhance the capability of the energy monitoring system in hidden danger identification and risk assessment, and provide powerful guarantee for energy safety and stable operation of the park.

In some preferred embodiments, the plurality of campus energy monitoring data units for acquiring the campus energy monitoring data includes: and respectively disassembling the park energy monitoring data according to the data capturing cores of a plurality of scales to obtain park energy monitoring data unit sets of the respective scales, wherein the park energy monitoring data unit sets of each scale comprise a plurality of park energy monitoring data units of the scale.

In some more detailed and efficient preferred embodiments, the campus energy monitoring system employs a method based on multi-scale data capture when acquiring multiple units of campus energy monitoring data.

Specifically, the system first uses a plurality of scale data capture check park energy monitoring data to disassemble according to different data scales. The term "data size" as used herein refers to the magnitude or range of data processing, which may be adjusted to accommodate different data processing scenarios, depending on the actual requirements.

During the disassembly process, each scale of data capture kernel is dedicated to capturing and disassembling data of a corresponding scale. In this way, the system is able to obtain a plurality of individual campus energy monitoring data unit sets of scale. Each data unit set contains a plurality of campus energy monitoring data units, and the size or range (i.e., the "size") of these data units is uniform.

For example, the system sets three scale data capture kernels: small scale, medium scale and large scale. A small-scale data capture kernel may focus on capturing energy usage data for a short period of time, a medium-scale may focus on data changes during one day or week, and a large-scale may cover data trends for longer periods of time. In this way, the system is able to analyze the energy usage of the campus comprehensively and carefully.

The multi-scale data dismantling method not only improves the flexibility and efficiency of data processing, but also enables the system to more accurately identify and analyze energy use modes and potential hazards in different time scales or data ranges.

Thus, the energy monitoring system of the park can acquire and analyze the energy monitoring data of the park more comprehensively and more finely by adopting a multi-scale data capturing method. The method not only improves the efficiency and flexibility of data processing, but also enhances the capability of the system for identifying and analyzing the energy use modes and potential hazards under different data scales. Overall, these embodiments significantly optimize the data processing and analysis flows of the campus energy monitoring system, providing more powerful and accurate support for the energy management and hidden trouble prevention of the campus.

In some preferred embodiments, each of the campus energy monitoring data units includes a plurality of monitoring fields, and determining a potential hidden danger trend value of a target energy change event in the campus energy monitoring data according to target labels corresponding to the plurality of the campus energy monitoring data units, including: for each park energy monitoring data unit of each scale, adding a commonality score between a park energy monitoring quantized vector of the park energy monitoring data unit and a corresponding target event hidden danger labeling quantized vector to a plurality of monitoring fields in the park energy monitoring data unit on the basis that a target label corresponding to the park energy monitoring data unit indicates that event state information of the target energy change event in the park energy monitoring data unit has hidden danger risk; for each monitoring field, obtaining a discrimination weight of the monitoring field according to the commonality scores of the monitoring field under the scales, wherein the discrimination weight is used for indicating the possibility that the monitoring field has hidden danger; and determining potential hidden danger trend values of the target energy change events in the park energy monitoring data according to the discrimination weight of each monitoring field in the park energy monitoring data, wherein the potential hidden danger trend values of the target energy change events comprise at least one of the distribution characteristics and hidden danger grades of the monitoring fields of the target energy change events with hidden danger risks.

In some preferred embodiments, the processing and analysis of energy monitoring data by the campus energy monitoring system proceeds to a more refined stage. In this process, each campus energy monitoring data unit contains a plurality of monitoring fields, which record different energy parameters, such as voltage, current, power factor, etc., and provide comprehensive energy use status information for the system.

When determining potential hidden danger trend values of the target energy change event, the system first focuses on the target labels corresponding to each park energy monitoring data unit. If the target label indicates that the target energy change event in the data unit has hidden danger risk, the system adds a commonality score between the park energy monitoring quantized vector of the data unit and the corresponding target event hidden danger label quantized vector to each monitoring field of the data unit. This is done to quantify the degree of association between each field and risk of hidden danger.

Next, the system will integrate the commonality scores of each monitoring field at multiple data scales to calculate the discrimination weight of the field. This discriminant weight actually indicates the size of the likelihood that the field is at risk. For example, if a field shows a high commonality score with risk of hidden danger across multiple data units and multiple scales, its discrimination weight increases accordingly.

Finally, the system determines potential hidden danger trend values of the target energy change events in the whole park energy monitoring data according to the judging weight of each monitoring field. This trend value includes not only the distribution characteristics of the monitored field at risk of the hidden trouble, but may also contain the level information of the hidden trouble. Thus, the system can provide a comprehensive and detailed energy use hidden danger analysis report for the user.

In this manner, the campus energy monitoring system is able to accurately determine potential hidden danger trend values for a target energy change event through in-depth analysis and weight discrimination for each monitored field. The method not only improves the accuracy of hidden danger identification, but also provides a more detailed and specific hidden danger risk assessment report for the user. Overall, these embodiments significantly promote the level of refinement and intelligence of the energy monitoring system in the aspect of risk assessment and hidden trouble recognition, and provide more powerful data support for energy safety and stable operation of the campus.

Embodiments of the present application provide a computer readable storage medium having stored thereon a program which when executed by a processor implements the artificial intelligence based campus energy monitoring method.

The embodiment of the application provides a processor which is used for running a program, wherein the program runs to execute the park energy monitoring method based on artificial intelligence.

In an embodiment of the present application, as shown in fig. 2, the campus energy monitoring system 100 includes at least one processor 101, and at least one memory 102 and a bus 103 connected to the processor 101; wherein, the processor 101 and the memory 102 complete communication with each other through the bus 103; the processor 101 is configured to invoke program instructions in the memory 102 to perform the artificial intelligence based campus energy monitoring method described above.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, campus energy monitoring systems (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In one typical configuration, the campus energy monitoring system includes one or more processors (CPUs), memory, and buses. The campus energy monitoring system may also include input/output interfaces, network interfaces, etc.

The memory may include volatile memory, random Access Memory (RAM), and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM), among other forms in computer readable media, the memory including at least one memory chip. Memory is an example of a computer-readable medium.

Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, read only compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage computer readable storage media, or any other non-transmission media, which can be used to store information that can be accessed by the campus energy monitoring system. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article of manufacture, or computer readable storage medium that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article of manufacture, or computer readable storage medium. Without further limitation, an element defined by the phrase "comprising one does not exclude the presence of other like elements in a process, method, article of manufacture, or computer readable storage medium comprising the element.

It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and variations of the present application will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the application are to be included in the scope of the claims of the present application.

Claims

1. A method for monitoring energy of a campus based on artificial intelligence, which is applied to a system for monitoring energy of the campus, the method comprising:

Circularly debugging the energy hidden danger event discrimination network according to the common scores and the set common threshold values of the multiple batches of network debugging example binary groups;

Each park energy monitoring data unit sample corresponds to a plurality of event hidden danger priori labels, and the event hidden danger priori labels describe potential hidden danger characteristics of event state information in the park energy monitoring data unit sample by different event hidden danger attributes respectively;

Vector mining is respectively carried out on a plurality of event hidden danger priori labels in the network debugging example binary group through the energy hidden danger event judging network to obtain basic event hidden danger label quantized vectors respectively corresponding to the event hidden danger priori labels, and average vectors of the basic event hidden danger label quantized vectors are determined to obtain the event hidden danger label quantized vectors;

2. The method of claim 1, wherein obtaining a plurality of network debug example doublets from a plurality of first campus energy monitoring data samples comprises:

3. The method according to claim 1, wherein the method further comprises:

4. The method according to claim 1, wherein the method further comprises:

5. The method according to claim 1, wherein the method further comprises:

6. The method of claim 5, wherein determining potential risk trend values for the target energy change event in the campus energy monitoring data based on the target labels corresponding to the plurality of campus energy monitoring data units, respectively, comprises:

7. The method of claim 5, wherein for each campus energy monitoring data unit, determining, by the energy hidden danger event discrimination network, a target label corresponding to the campus energy monitoring data unit, comprises:

For each park energy monitoring data unit, extracting park energy monitoring quantized vectors of the park energy monitoring data unit through the energy hidden danger event discrimination network, determining commonality scores between the park energy monitoring quantized vectors and a plurality of set event hidden danger labeling quantized vectors respectively, and determining target event hidden danger labeling quantized vectors with commonality scores meeting setting requirements from the set event hidden danger labeling quantized vectors, wherein the set event hidden danger labeling quantized vectors respectively correspond to set labels, and the target labels are set labels corresponding to the target event hidden danger labeling quantized vectors;

wherein, obtain a plurality of garden energy monitoring data units of garden energy monitoring data, include: respectively disassembling the park energy monitoring data according to a plurality of scale data capturing cores to obtain a park energy monitoring data unit set of each scale, wherein the plurality of park energy monitoring data units included in each scale park energy monitoring data unit set are of the scale;

Each of the campus energy monitoring data units comprises a plurality of monitoring fields, and the potential hidden danger trend value of the target energy change event in the campus energy monitoring data is determined according to the target labels respectively corresponding to the plurality of the campus energy monitoring data units, and the method comprises the following steps: for each park energy monitoring data unit of each scale, adding a commonality score between a park energy monitoring quantized vector of the park energy monitoring data unit and a corresponding target event hidden danger labeling quantized vector to a plurality of monitoring fields in the park energy monitoring data unit on the basis that a target label corresponding to the park energy monitoring data unit indicates that event state information of the target energy change event in the park energy monitoring data unit has hidden danger risk; for each monitoring field, obtaining a discrimination weight of the monitoring field according to the commonality scores of the monitoring field under the scales, wherein the discrimination weight is used for indicating the possibility that the monitoring field has hidden danger; and determining potential hidden danger trend values of the target energy change events in the park energy monitoring data according to the discrimination weight of each monitoring field in the park energy monitoring data, wherein the potential hidden danger trend values of the target energy change events comprise at least one of the distribution characteristics and hidden danger grades of the monitoring fields of the target energy change events with hidden danger risks.

8. A campus energy monitoring system, comprising a processor, and a memory and a bus connected with the processor; wherein the processor and the memory complete communication with each other through the bus; the processor is configured to invoke program instructions in the memory to perform the artificial intelligence based campus energy monitoring method of any of claims 7.