Multi-scale energy consumption model of urban energy system and its energy flow monitoring method

Abstract

The present invention proposes a multi-scale energy consumption model of an urban energy system and an energy flow monitoring method thereof. By collecting relevant data, energy flow monitoring indicators and methods for the urban energy system are constructed according to the energy consumption and energy flow conditions of the urban energy system. Based on deep learning, key links and problems in urban energy operations are explored and warned, and solutions are proposed to help promote the green and sustainable development of urban energy security and economy, provide low-carbon, efficient, economical and convenient energy guarantees for the high-quality development of the urban economy, and help accelerate the realization of the city's carbon peak and carbon neutrality process.

Description

Technical Field

[0001] The present invention belongs to the technical fields of big data information processing, energy flow monitoring of urban energy systems, and in particular to a multi-scale energy consumption model of an urban energy system and an energy flow monitoring method thereof. Background Art

[0002] Urban energy transformation is the key to achieving a national energy revolution. Urban energy monitoring and panoramic depiction are the basis for mastering the characteristics of urban energy operation and formulating policies and implementing management in a more targeted manner. At present, there is no effective energy monitoring at the city level. It is necessary to carry out research on urban energy monitoring and panoramic depiction to grasp the characteristics of urban energy operation and provide a basis for management departments, power grid companies and other energy companies to formulate policies and implement management in a more targeted manner.

[0003] Urban energy monitoring and panoramic depiction are the basis for carrying out energy revolution. At the same time, energy transformation is a powerful engine for improving the quality of future cities and an important breakthrough for breaking the bottleneck of urban development.

[0004] However, in the current urban energy system, there are problems of isolation between different energy types and between different links of the same energy types, and serious information barriers, which lead to unclear overall situation of urban energy operation, unclear interaction mechanism between energy sources, lack of effective coordination between energy sources, and a large gap between energy utilization efficiency and the world's advanced level. The urban energy system involves many types of energy such as electricity, gas, cold, and heat. The characteristics of energy supply and use of different energy types are different, and the line management is obvious, which is not conducive to the overall energy efficiency improvement. Carrying out urban energy system monitoring is the basis for improving urban energy efficiency and realizing urban energy transformation. At present, there is no effective energy monitoring in the city. It is necessary to carry out urban energy monitoring and panoramic description research to grasp the characteristics of urban energy operation and provide a basis for management departments, power grid companies and other energy enterprises to formulate policies and implement management in a more targeted manner.

[0005] From the perspective of basic conditions, the level of informatization of urban energy systems continues to improve, and various types of energy data are rapidly generated and collected, providing a physical and information basis for conducting relevant research.

[0006] At present, the construction of smart cities has greatly improved the level of urban information networks and has initially established a foundation for an integrated urban energy information interaction network, providing a good physical basis for monitoring urban energy systems and optimizing and controlling urban energy systems.

[0007] At the same time, after the construction of smart city smart energy has achieved the upgrading of the basic information network, the energy information network has been expanded to various users and energy providers. Due to its operation and collection, an explosive amount of data has been generated. These data have typical big data characteristics. The large amount of energy data has provided new solutions and paths for monitoring the operation of urban energy systems and analyzing the optimization operation strategies of urban energy systems through big data and artificial intelligence. Summary of the invention

[0008] In view of this, in order to overcome the defects and shortcomings of the prior art, the purpose of the present invention is to provide a multi-scale energy consumption model of an urban energy system and an energy flow monitoring method thereof. By collecting relevant data, according to the energy consumption and energy flow conditions of the urban energy system, energy flow monitoring indicators and methods of the urban energy system are constructed. Based on deep learning, key links and problems in urban energy operations are explored and warned, and solutions are proposed to help the safe, green and sustainable development of urban energy economy, provide low-carbon, efficient, economical and convenient energy guarantees for the high-quality development of the urban economy, and help accelerate the realization of urban carbon peak and carbon neutrality.

[0009] Based on the above research and design, the present invention specifically adopts the following technical solutions:

[0010] A multi-scale energy consumption model of an urban energy system, characterized in that the data preparation steps in the construction process are:

[0011] According to the characteristic analysis of energy consumption objects, historical energy consumption data, environmental data and time data are selected as model inputs; after data integration, combined with the actual load at the target time, a training data set is formed. Where X i represents the model input vector consisting of historical energy consumption data, environmental data and time data, y i is the actual load at the target time, and N is the size of the data set. Aiming at the needs of multiple spatial scales, based on the basic data set, combined with the spatial distribution of energy-consuming objects and energy consumption characteristics, it is aggregated according to the required scale to obtain energy consumption data sets at different spatial scales.

[0012] The steps of model construction and training are as follows: according to the complexity of the actual energy consumption object, the depth of the neural network is selected, and the model input is the corresponding input vector X i , the expected output is y i The real output is The loss function of the constructed model is:

[0013]

[0014] Where θ is the model parameter; the model is trained using the stochastic gradient descent method:

[0015]

[0016] Until the model converges or iterates to the maximum number of iterations, the quantitative relationship between load demand and its influence vector is obtained.

[0017] Furthermore, in the data preparation process, according to the requirements of multiple time scales, the simultaneous load 1 to 4 weeks before the target moment, the simultaneous load for the previous 7 consecutive days, and the load for the previous 23 consecutive hours are selected as the main historical energy consumption data; the meteorological data corresponding to each historical energy consumption is used as environmental data; the season, week and time of the target moment are used as time data.

[0018] And one of the energy flow monitoring methods based on the above multi-scale energy consumption model of urban energy system:

[0019] Step A1: Select the model input according to the energy consumption object: historical energy consumption data x1, environmental factors x2 and social factors x3, and send them to the multi-scale energy consumption model of the urban energy system to obtain the predicted load y at the future moment;

[0020] Step A2: Determine the parameters including the capacity and energy consumption threshold of the energy system in the area where the energy-consuming object is located, and monitor the energy flow operation of the energy system in real time by comparing the values ​​of each parameter with the set safety alarm limits.

[0021] And the second energy flow monitoring method based on the above multi-scale energy consumption model of the urban energy system: when monitoring the abnormal operating status of objects or equipment in the system, directly build an abnormal state discrimination data model based on the energy object operation data and deep learning methods, that is: select relevant energy data as input according to the characteristics of the target object and the characteristics of the abnormal state, and then extract data features based on prior knowledge, and then send them to the deep network to extract the deep semantic features in the data, and finally use a linear classifier to complete the abnormal state judgment.

[0022] And the third energy flow monitoring method based on the above multi-scale energy consumption model of urban energy system: using a semi-supervised learning framework to use a large amount of normal operation data to form a reliable feature representation; and on the basis of this feature representation, combined with a small amount of abnormal data, training the final classifier of energy flow monitoring, specifically including the following steps:

[0023] Step B1: Based on prior knowledge, construct a set of standard operating data of energy consumption objects, and on this basis, combine the actual operating data to form a set of data pairs. The distribution is sent to the feature extraction network F(·) to obtain the corresponding feature vector The difference between the two is used as the final feature of the time running data

[0024]

[0025] Step B2: For normal operation data or unmarked abnormal data, when the data fluctuates within a small range or shifts in time series, the actual operating state of the energy consumption object does not change, so the representation in the feature space should be basically consistent; based on this assumption, a consistency loss function is constructed:

[0026]

[0027] Where f' is the feature representation after the data is offset or time-shifted;

[0028] For labeled anomaly data, in addition to the consistency loss, the loss function of the anomaly classifier constructed based on the feature representation should also be considered:

[0029] L cls (x,y)=CE(y,y')

[0030] Where y is the label of the abnormal data, y' is the classifier output, and CE(·) is the cross entropy function;

[0031] Based on these two, the overall loss function of model training is constructed:

[0032] L=L cls (x,y)+λL con (x)

[0033] Where λ is the preset proportionality coefficient;

[0034] During the training process of the model, the parameters of the feature extraction network F(·) are constantly changing, especially at the beginning of the model training. This leads to a large volatility in the consistency loss. In order to improve the stability of the model training, the feature extraction network after parameter sliding average is used as the teacher model to obtain f', thereby guiding the learning ability of the original feature extraction network F(·) as the student model.

[0035] Furthermore, the execution of energy flow monitoring of the urban energy system specifically includes the following steps:

[0036] Step S1: Design of energy flow monitoring service for urban energy system

[0037] According to the monitoring requirements of urban energy flow, summarize the urban energy flow monitoring business needs, classify the differences of urban energy flow monitoring business according to the structural characteristics of energy subsystems, and establish urban energy flow monitoring business rules and mapping relationships;

[0038] Step S2: Preparation of urban energy system energy flow monitoring data

[0039] Sort out the city energy flow monitoring business data, trace the source business system, source channel, data table, corresponding field for each city energy flow monitoring business data item in the data demand table, clarify the association and matching relationship between data items, form the monitoring business data demand table of each city energy subsystem, and merge the detailed business data required for city energy flow monitoring through the interfaces of each business subsystem of city energy flow monitoring and other data channels;

[0040] Step S3: Urban energy system energy flow monitoring data processing

[0041] In view of the source business system access and subsystem data of urban energy flow monitoring, data quality verification is carried out on the extracted detailed data related to urban energy flow monitoring from the aspects of data integrity, standardization, rationality, accuracy and consistency to verify the availability and validity of the data; based on the actual business and data requirements of urban energy flow monitoring, data cleaning and conversion rules are formed to clean invalid data and form a valid data set of detailed business data; combined with the association mapping relationship between data tables and data items, detailed business data tables and monitoring-level data corresponding to the monitoring rules are formed;

[0042] Step S4: Construction of multi-scale energy consumption model of urban energy system

[0043] Based on the business content, rules and data requirements of urban energy flow monitoring, a business data model suitable for each energy subsystem is constructed, a certain proportion of data is extracted, and the data is substituted into the urban energy flow monitoring data model for training. Based on the accuracy, fit and other parameters of the training results, the feasibility, rationality and accuracy of the model are verified, and the business model parameters are adaptively adjusted to meet the business needs of urban energy flow monitoring;

[0044] Step S5: Calculation and analysis of urban energy system energy flow monitoring data

[0045] Utilizing the multi-scale energy consumption model of the urban energy system, data calculation and correlation mining are carried out to form monitoring results for each energy subsystem and urban energy flow.

[0046] Furthermore, the execution of energy flow monitoring of urban energy systems also includes step S6: solidification of energy flow monitoring results of urban energy systems: optimizing and improving the model according to the deviation between the urban energy flow monitoring results and the expected values ​​of urban energy flow; setting the monitoring results presentation form according to different needs including those of urban energy authorities, energy operating companies, energy-consuming companies and the general public, and solidifying the urban energy flow monitoring report model using data processing and data mining tools.

[0047] Compared with the existing technology, the present invention and its preferred scheme collect relevant data, construct energy flow monitoring indicators and methods for urban energy systems according to the energy consumption and energy flow conditions of urban energy systems, and based on deep learning, explore and warn of key links and problems in urban energy operations, propose solutions, and help promote the green and sustainable development of urban energy security and economy, provide low-carbon, efficient, economical and convenient energy guarantees for the high-quality development of the urban economy, and help accelerate the realization of urban carbon peak and carbon neutrality. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0049] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0050] Figure 1 This is a schematic diagram of the main design and workflow of the multi-scale energy consumption model of the urban energy system and its energy flow monitoring method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the features and advantages of this patent more obvious and easy to understand, the following embodiments are specifically described in detail as follows:

[0052] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. The components described and shown in the drawings generally herein can be designed in combination with different configurations. Therefore, the following detailed description of the selected embodiments of the present invention provided in the drawings is not intended to limit the scope of the invention claimed for protection, but only represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0053] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0054] like Figure 1 As shown, this embodiment further introduces the design process and working mechanism of the multi-scale energy consumption model of the urban energy system and the energy flow monitoring method thereof proposed by the present invention from the perspective of working mechanism and process, which is implemented in the form of a computer program:

[0055] Embodiment 1

[0056] Embodiment 1 of the present invention discloses a host platform (hardware) that can be relied on for energy flow monitoring of an urban energy system based on a multi-scale energy consumption model, which can utilize the energy monitoring system of energy enterprises such as power grid companies or urban competent departments. The energy monitoring system includes a monitoring terminal and at least two collection terminals, including:

[0057] Data collection at the collection end. The collection end is deployed in the operation monitoring system of different energy subsystems (such as electricity, gas, etc.). The subsystem monitoring equipment automatically collects or manually enters data such as the current type of energy consumption, supply, unit energy consumption, cost, emissions, reserves, future production plans, etc.

[0058] The data collection end calculates the data collected by the collection end, and calculates the gap between the current energy subsystem operation index and the total expected value of urban energy based on the summary index issued by the monitoring end, and provides an adjustment plan for the current energy subsystem operation index.

[0059] The communication between the collection end and the monitoring end adopts the existing public communication networks and communication protocols such as telecommunications, including encoding, parsing, security protection, communication protocols, etc., to receive and send statistical and analytical data of energy subsystems of one monitoring end and at least two collection ends.

[0060] The monitoring end calculates data. The monitoring end is deployed in the operation monitoring system of large energy enterprises such as power grid companies or urban energy management departments. The monitoring end performs combined calculations based on the urban energy subsystem operation statistical data received from at least two collection ends, and gives the operation evaluation results of each urban energy subsystem, as well as the main influencing factors and impact range of urban energy flow.

[0061] The monitoring end displays data. The monitoring end displays the summary statistics and analysis results on the monitoring screen. The display page includes an overview, an overview of each collection end, current values, cumulative values, main factors affecting urban energy supply and demand, comprehensive evaluation indicators and improvements, etc. This will help urban energy operation authorities and related energy companies to timely understand the evolution of urban energy flows and structures, and help implement the national dual-carbon goals and energy dual-control goals.

[0062] Embodiment 2

[0063] In combination with the platform provided in the first embodiment, the energy flow monitoring design of the urban energy system is performed, including the following steps:

[0064] (1) Design of energy flow monitoring service for urban energy system

[0065] In combination with the monitoring requirements of the municipal government and energy enterprises such as the power grid company for urban energy flow, the urban energy flow monitoring business needs of relevant stakeholders are summarized, the urban energy flow monitoring business is differentiated according to the structural characteristics of the energy subsystem, and the urban energy flow monitoring business rules and mapping relationships are established.

[0066] (2) Preparation of energy flow monitoring data for urban energy systems

[0067] Sort out the urban energy flow monitoring business data, trace the source business system, source channel, data table, and corresponding fields for each urban energy flow monitoring business data item in the data requirement table, clarify the correlation and matching relationship between the data items, and form a data requirement table for monitoring business of each urban energy subsystem. Through the interfaces of each business subsystem of urban energy flow monitoring, offline collection and other methods, merge and form the detailed business data required for urban energy flow monitoring.

[0068] (3) Data processing of energy flow monitoring in urban energy systems

[0069] In view of the source business system access and subsystem data of urban energy flow monitoring, data quality verification is carried out on the extracted detailed data related to urban energy flow monitoring from the aspects of data integrity, standardization, rationality, accuracy, consistency, etc., using R, Python, SPSS, EXCEL and other tools to verify the availability and validity of the data. Based on the actual business and data requirements of urban energy flow monitoring, data cleaning and conversion rules are formed to clean invalid data and form a valid data set of detailed business data. Combined with the association mapping relationship between data tables and data items, detailed business data tables and monitoring-level data corresponding to the monitoring rules are formed.

[0070] (4) Construction of energy flow monitoring model for urban energy system

[0071] Based on the business content, rules and data requirements of urban energy flow monitoring, a business data model suitable for each energy subsystem is constructed. A certain proportion of data is extracted and substituted into the urban energy flow monitoring data model for training. Based on the accuracy, fit and other parameters of the training results, the feasibility, rationality and accuracy of the model are verified. At the same time, the business model parameters are adaptively adjusted to meet the business needs of urban energy flow monitoring.

[0072] (5) Calculation and analysis of energy flow monitoring data of urban energy systems

[0073] Utilize the urban energy flow monitoring model to carry out data calculation and correlation mining, generate monitoring results of various energy subsystems and urban energy flows, including result charts, monitoring reports and other achievements, and carry out PC terminal display content design and configuration.

[0074] (6) Consolidation of the results of energy flow monitoring in urban energy systems

[0075] According to the deviation between the urban energy flow monitoring results and the expected values ​​of urban energy flow, the model is optimized and improved; the monitoring results presentation form is set according to the different needs of urban energy management departments, energy operating companies, energy-consuming companies and the general public; and the urban energy flow monitoring report model is solidified using tools such as data processing and data mining.

[0076] Embodiment 3

[0077] Combined with the urban energy system energy flow monitoring design provided by the second embodiment of the present invention, an algorithm model is further designed, including: Figure 1 The technical route of the urban energy system energy flow monitoring method based on the multi-scale energy consumption model shown includes the following steps:

[0078] S1: Select the energy monitoring type. According to the commonly used classification method in the energy field of the China Statistical Yearbook, the energy of the urban energy system can be divided into six categories: coal, oil, natural gas, electricity, heat, and cooling. On this basis, the energy types are further subdivided according to the specific conditions of each energy category, such as variety, source, and use, among which:

[0079] (1) Coal is divided into three types according to international coal classification standards: lignite, bituminous coal, and anthracite;

[0080] (2) Petroleum is divided into four main energy types: gasoline, diesel, kerosene and liquefied petroleum gas;

[0081] (3) The sources of coal, oil and natural gas supply are divided into local production and imported;

[0082] (4) Electricity includes conventional energy electricity and renewable energy electricity, including: coal-fired power (pure condensing coal-fired power units, combined heat and power coal-fired power units), gas-fired power generation, hydropower, wind power, nuclear power, solar power generation, biomass power generation, etc. The power supply is divided into locally generated and externally received electricity;

[0083] (5) Heat includes heating technologies such as cogeneration units, coal-fired boilers, gas boilers, and heat pump floor heating, with local supply being the main consideration.

[0084] (6) Cooling capacity includes cooling by combined heat and power units, electric refrigeration units, heat pumps and other cooling technologies, with the main consideration being local supply.

[0085] The basic links of the urban energy system include resource mining and collection, processing and conversion, transportation and distribution, demand sectors, energy-using facilities, and final use / useful energy demand. Primary energy is obtained after energy resources are mined and collected; secondary energy is obtained after primary energy is processed and converted. Secondary energy is transported and distributed to terminal energy demand sectors such as industry, agriculture, transportation, commerce, and urban and rural life, and various energy-using facilities provide users with useful energy that meets various final uses such as process heat, power, lighting, hot water, cooking, heating, and refrigeration. Energy flow will have certain losses when passing through each link, such as energy production loss, processing and conversion loss, transportation and distribution loss (transportation loss of petroleum products, transmission and transformation loss of electricity, etc.), and terminal consumption loss. Various environmental impact issues may also occur in the process of energy development, conversion, processing, transportation, storage, and utilization, such as the emission of PM10, TSP, SO2, NO2, petroleum hydrocarbons and other atmospheric pollutants, and the emission of greenhouse gases such as CO2, which will affect global climate change.

[0086] S2: Construct a multi-scale energy consumption model for deep learning;

[0087] (1) Data preparation. According to the analysis of the characteristics of energy consumption objects, the main factors affecting energy consumption are historical energy consumption data, environmental data and time data. Therefore, these three aspects of data are selected as model input. Specifically, according to the requirements of multiple time scales, the load at the same time of 1 to 4 weeks before the target time, the load at the same time of the previous 7 consecutive days and the load of the previous 23 consecutive hours are selected as the main historical energy consumption data; the meteorological data corresponding to each historical energy consumption, such as temperature, humidity, etc., are used as environmental data; the season, week and time of the target time are used as time data; the above data are integrated and combined with the actual load at the target time to form a training data set. Where X i Indicates that historical energy consumption data, environmental data and time data constitute the model input vector, y i is the actual load at the target time, and N is the size of the data set. Aiming at the needs of multiple spatial scales, based on the basic data set, combined with the spatial distribution of energy-consuming objects, energy consumption characteristics and other factors, it is aggregated according to the required scale to obtain energy consumption data sets at different spatial scales.

[0088] (2) Model construction and training. According to the complexity of the actual energy consumption object, the depth of the neural network is selected, usually 3-5 layers. The model input is the corresponding input vector X i , the expected output is y i The real output is The loss function of the constructed model is:

[0089]

[0090] Where θ is the model parameter. The model is trained using stochastic gradient descent:

[0091]

[0092] Until the model converges or iterates to the maximum number of iterations, the quantitative relationship between load demand and its influence vector is obtained.

[0093] S3: Establish a method for monitoring energy flow data using deep neural networks and semi-supervised learning;

[0094] Energy system monitoring refers to real-time monitoring of the operating status of the energy system and guiding the operation and maintenance scheduling of the energy system through distributed energy data collection, data preprocessing, energy data analysis and other steps based on physical mechanisms and energy consumption models. Common monitoring indicators include energy system characteristics such as energy flow, energy structure, and energy efficiency. These indicators can be calculated and processed in real time based on the collected energy data. Peak load is also an important task in energy system monitoring. It is necessary to first predict the peak load at a specified time based on the constructed energy consumption model, and then estimate and monitor the system operation safety based on information such as system capacity and margin.

[0095] (1) According to the multi-scale model constructed in this embodiment, the model input is selected according to the determined energy-consuming object: historical energy consumption data x1, environmental factors x2 and social factors x3 are sent to the multi-scale energy consumption model to obtain the predicted load y at the future time.

[0096] (2) Determine the capacity and energy consumption threshold of the energy system in the area where the energy-consuming object is located. Taking the power system as an example, the installed capacity of the energy-consuming object is recorded as C. According to its energy consumption characteristics and expert knowledge, the energy consumption threshold parameter τ corresponding to the category to which the object belongs is determined. Then, the three data y, C, and τ are sent to the constructed energy flow monitoring platform.

[0097] (3) On the upper-level energy flow monitoring platform, the energy flow operation of the energy system is monitored in real time by comparing the numerical relationship between the predicted load y, installed capacity C, and threshold parameter τ, combined with the set safety warning limit.

[0098] Another important monitoring task is to monitor the abnormal operating status of objects or equipment in the system. This task has many coupling factors and random factors, and it is difficult to accurately describe this situation using physical mechanisms. Therefore, an abnormal state discrimination data model is directly constructed based on energy object operation data and deep learning methods. The specific process is to select relevant energy data as input based on the characteristics of the target object and the characteristics of the abnormal state, and then extract data features based on prior knowledge, and then send it to the deep network to extract the deep semantic features in the data, and finally use a linear classifier to complete the abnormal state judgment.

[0099] In actual scenarios, energy systems are in normal operation most of the time, which results in a serious imbalance in the distribution of normal and abnormal energy data. The scarce abnormal state is the real focus of attention for the energy system. On the other hand, compared with normal data, the distribution of abnormal data is more scattered, with many patterns, which is difficult to represent with unified features, and requires expert knowledge to determine specific anomalies. These two characteristics determine that energy flow monitoring of energy systems is essentially a multi-classification problem with serious sample imbalance.

[0100] In order to solve such problems, this embodiment adopts a semi-supervised learning framework to make full use of a large amount of normal operating data to condense a reliable feature representation; and based on this feature representation, combined with a small amount of abnormal data, the final classifier for energy flow monitoring is trained.

[0101] First, based on prior knowledge, a set of standard operating data of energy consumption objects is constructed. On this basis, combined with the actual operating data, a set of data pairs is formed. The distribution is sent to the feature extraction network F(·) to obtain the corresponding feature vector The difference between the two is used as the final feature of the time running data

[0102]

[0103] Secondly, for normal operation data or unmarked abnormal data, when the data is slightly shifted or time-series shifted, the actual operation state of the energy consumption object does not change, so the representation in the feature space should be basically consistent. Based on this assumption, a consistency loss function is constructed.

[0104]

[0105] Where f' is the feature representation after the data is offset or time-shifted.

[0106] For labeled anomaly data, in addition to consistency loss, the loss function of the anomaly classifier constructed based on feature representation should also be considered.

[0107] L cls (x,y)=CE(y,y')

[0108] Where y is the label of the abnormal data, y' is the classifier output, and CE(·) is the cross entropy function.

[0109] Based on these two, the overall loss function of model training is constructed

[0110] L=L cls (x,y)+λL con (x)

[0111] Where λ is the preset proportionality factor.

[0112] During the model training process, the parameters of the feature extraction network F(·) are constantly changing, especially at the beginning of the model training. This leads to a large volatility in the consistency loss. In order to improve the stability of the model training, the feature extraction network after the sliding average of the parameters is used as the teacher model to obtain f', thereby guiding the learning ability of the original feature extraction network F(·) as the student model.

[0113] S4: Analyze and evaluate monitoring results.

[0114] This embodiment can evaluate the urban energy flow from multiple dimensions such as safety, economy, greenness, efficiency, convenience, and sustainability.

[0115] Safety is expressed as a comprehensive weighted assessment of indicators such as the probability of abnormal energy operation events occurring in the city, the expected value of economic losses, and the self-sufficiency rate of local energy.

[0116] The economic expression is a comprehensive weighted assessment of indicators such as the current average cost per unit of energy consumption and the annual rate of decline in the average cost.

[0117] Green represents the comprehensive weighted assessment value of indicators such as conventional environmental pollutants, total greenhouse gas emissions, emission intensity, and emission reduction rate of current energy consumption.

[0118] Efficiency is the comprehensive weighted assessment value of current energy consumption indicators such as energy efficiency and energy efficiency improvement.

[0119] Convenience is a comprehensive weighted assessment of indicators such as the availability, coverage, and service satisfaction of current energy consumption.

[0120] Sustainability is a comprehensive weighted assessment of indicators such as the supply and demand gap of future energy consumption, economic fluctuations, and emission trends.

[0121] After weighted aggregation of the above indicators, we can obtain the comprehensive evaluation results of energy flow between different cities.

[0122] The above program design scheme provided in this embodiment can be stored in a computer-readable storage medium in a coded form and implemented in the form of a computer program, and the basic parameter information required for calculation is input through computer hardware, and the calculation result is output.

[0123] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0124] The present invention is described with reference to flowcharts of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each process in the flowchart, as well as the combination of processes in the flowchart, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the flowchart. Figure 1 A device that specifies functions in a process or multiple processes.

[0125] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple functions specified in a flowchart.

[0126] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 The steps of a specified function in a process or multiple processes.

[0127] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

[0128] This patent is not limited to the above-mentioned optimal implementation mode. Anyone can derive other forms of multi-scale energy consumption models of urban energy systems and energy flow monitoring methods under the inspiration of this patent. All equal changes and modifications made according to the scope of the patent application of the present invention should be covered by this patent.

Claims

1. An energy flow monitoring method for a multi-scale energy consumption model of an urban energy system, characterized by: The data preparation steps in the construction process of the multi-scale energy consumption model of the urban energy system are as follows: According to the characteristic analysis of energy consumption objects, historical energy consumption data, environmental data and time data are selected as model inputs; after data integration, combined with the actual load at the target time, a training data set is formed. Where X i represents the model input vector consisting of historical energy consumption data, environmental data and time data, y i is the actual load at the target time, and N is the size of the data set. Aiming at the needs of multiple spatial scales, based on the basic data set, combined with the spatial distribution of energy-consuming objects and energy consumption characteristics, it is aggregated according to the required scale to obtain energy consumption data sets at different spatial scales. The steps of model construction and training are as follows: according to the complexity of the actual energy consumption object, the depth of the neural network is selected, and the model input is the corresponding input vector X i , the expected output is y i The real output is The loss function of the constructed model is: Where θ is the model parameter; the model is trained using the stochastic gradient descent method: Until the model converges or iterates to the maximum number of iterations, the quantitative relationship between the load demand and its influence vector is obtained; A semi-supervised learning framework is used to form a reliable feature representation using a large amount of normal operation data. Based on this feature representation, a small amount of abnormal data is combined to train the final classifier for energy flow monitoring, which specifically includes the following steps: Step B1: Based on prior knowledge, construct a set of standard operating data of energy consumption objects, and on this basis, combine the actual operating data to form a set of data pairs. The distribution is sent to the feature extraction network F(·) to obtain the corresponding feature vector The difference between the two is used as the final feature of the time running data Step B2: For normal operation data or marked abnormal data, when the data fluctuates within a small range or shifts in time series, the actual operation state of the energy consumption object does not change, so the representation in the feature space should be basically consistent; Based on this assumption, a consistency loss function is constructed: Where f' is the feature representation after the data is offset or time-shifted; For labeled anomaly data, in addition to the consistency loss, the loss function of the anomaly classifier constructed based on the feature representation should also be considered: L cls (x,y)=CE(y,y') Where y is the label of the abnormal data, y' is the classifier output, and CE(·) is the cross entropy function; Based on these two, the overall loss function of model training is constructed: L=L cls (x,y)+λL con (x) Where λ is the preset proportionality coefficient; During the training process of the model, the parameters of the feature extraction network F(·) are constantly changing, especially at the beginning of the model training. This leads to a large volatility in the consistency loss. In order to improve the stability of the model training, the feature extraction network after parameter sliding average is used as the teacher model to obtain f', thereby guiding the learning ability of the original feature extraction network F(·) as the student model.

2. The energy flow monitoring method of the multi-scale energy consumption model of the urban energy system according to claim 1 is characterized by: During the data preparation process, according to the requirements of multiple time scales, the simultaneous loads 1 to 4 weeks before the target time, the simultaneous loads for the previous 7 consecutive days, and the loads for the previous 23 consecutive hours were selected as the main historical energy consumption data; the meteorological data corresponding to each historical energy consumption was used as environmental data; the season, week, and time of the target time were used as time data.

3. The energy flow monitoring method of the multi-scale energy consumption model of the urban energy system according to claim 1 is characterized by: Step A1: Select the model input according to the energy consumption object: historical energy consumption data x1, environmental factors x2 and social factors x3, and send them to the multi-scale energy consumption model of the urban energy system to obtain the predicted load y at the future moment; Step A2: Determine the parameters including the capacity and energy consumption threshold of the energy system in the area where the energy-consuming object is located, and monitor the energy flow operation of the energy system in real time by comparing the values ​​of each parameter with the set safety alarm limits.

4. The energy flow monitoring method of the multi-scale energy consumption model of the urban energy system according to claim 1 is characterized by: When monitoring the abnormal operating status of objects or equipment within the system, an abnormal state discrimination data model is constructed directly based on the energy object operation data and deep learning methods, that is: relevant energy data is selected as input according to the characteristics of the target object and the abnormal state characteristics, and then data features are extracted based on prior knowledge, and then sent to the deep network to extract the deep semantic features in the data, and finally a linear classifier is used to complete the abnormal state judgment.

5. The energy flow monitoring method of the multi-scale energy consumption model of the urban energy system according to claim 1 is characterized by: The implementation of energy flow monitoring in urban energy systems includes the following steps: Step S1: Design of energy flow monitoring service for urban energy system According to the monitoring requirements of urban energy flow, summarize the urban energy flow monitoring business needs, classify the differences of urban energy flow monitoring business according to the structural characteristics of energy subsystems, and establish urban energy flow monitoring business rules and mapping relationships; Step S2: Preparation of urban energy system energy flow monitoring data Sort out the city energy flow monitoring business data, trace the source business system, source channel, data table, corresponding field for each city energy flow monitoring business data item in the data demand table, clarify the association and matching relationship between data items, form the monitoring business data demand table of each city energy subsystem, and merge the detailed business data required for city energy flow monitoring through the interfaces of each business subsystem of city energy flow monitoring and other data channels; Step S3: Urban energy system energy flow monitoring data processing In view of the source business system access and subsystem data of urban energy flow monitoring, data quality verification is carried out on the extracted detailed data related to urban energy flow monitoring from the aspects of data integrity, standardization, rationality, accuracy and consistency to verify the availability and validity of the data; based on the actual business and data requirements of urban energy flow monitoring, data cleaning and conversion rules are formed to clean invalid data and form a valid data set of detailed business data; combined with the association mapping relationship between data tables and data items, detailed business data tables and monitoring-level data corresponding to the monitoring rules are formed; Step S4: Construction of multi-scale energy consumption model of urban energy system Based on the business content, rules and data requirements of urban energy flow monitoring, a business data model suitable for each energy subsystem is constructed, a certain proportion of data is extracted, and the data is substituted into the urban energy flow monitoring data model for training. Based on the accuracy and fit parameters of the training results, the feasibility, rationality and accuracy of the model are verified, and the business model parameters are adaptively adjusted to meet the business needs of urban energy flow monitoring; Step S5: Calculation and analysis of urban energy system energy flow monitoring data Utilizing the multi-scale energy consumption model of the urban energy system, data calculation and correlation mining are carried out to form monitoring results for each energy subsystem and urban energy flow.

6. The energy flow monitoring method of the multi-scale energy consumption model of the urban energy system according to claim 5 is characterized by: The execution of energy flow monitoring of the urban energy system also includes step S6: solidification of the results of the energy flow monitoring of the urban energy system: optimizing and improving the model according to the deviation between the urban energy flow monitoring results and the expected value of the urban energy flow; setting the form of monitoring results according to different needs including those of urban energy authorities, energy operating companies, energy-consuming companies and the general public, and solidifying the urban energy flow monitoring report model using data processing and data mining tools.

Images