A user-side intelligent power distribution system, method, medium, device and terminal

Abstract

The present invention belongs to the field of intelligent power consumption technology, and discloses a user-side intelligent power distribution and utilization system, method, medium, equipment and terminal. The user-side intelligent power distribution and utilization system adopts a B / S architecture and uses an MVC design pattern, including a data acquisition and transmission layer, a data layer, a business layer and an application service layer. The user-side intelligent power distribution and utilization system includes a system setting module, a file management module, a report management module, a map navigation module, a data maintenance module, an alarm management module, an online monitoring module, a power and load prediction module, and an electric energy evaluation module. The user-side intelligent power distribution and utilization system of the present invention integrates four functions of data acquisition and monitoring, power quality and energy efficiency evaluation, data mining and artificial intelligence prediction, and precise optimization and management, establishes full-process, closed-loop data flow supervision, and realizes the integration of various business data, scientific management and information sharing services.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent electricity utilization technology, and in particular relates to a user-side intelligent power distribution system, method, medium, equipment and terminal. Background Art

[0002] In the construction of distribution network automation, communication mode is a very important link and the basis for the implementation of distribution network automation. Therefore, the implementation of distribution network automation must solve the communication problem in order to accurately communicate and exchange data between the distribution terminal equipment and the main station of the distribution automation system. With the development of communication technology, there are many communication means available, mainly optical fiber, distribution line carrier, telephone dedicated line, CATV channel, wireless spread spectrum, microwave communication, dedicated line RS-485, public network GPRS / 3G, etc., which are often used for fault indicators, FTU, TTU communication, load control, low-voltage user meter reading, communication trunk line, etc. The application of these communication technologies has solved the problem of power distribution information transmission to a certain extent. Nowadays, the power distribution method of the power distribution system is to distinguish power distribution according to the power consumption method, that is, residential power consumption, commercial power consumption, etc. The power distribution method is relatively fixed, and it is impossible to adjust the power distribution according to the power consumption demand.

[0003] Through the above analysis, the problems and defects of the existing technology are as follows: the distribution mode of the existing power distribution system is relatively fixed according to the power consumption mode, and the power distribution cannot be adjusted according to the power consumption demand. The power distribution information diagnosis and transmission channels are blocked, and the data information feedback is not timely. The power demand side has uneven power quality, high safety risks and poor low-carbon energy-saving effects. Summary of the invention

[0004] In view of the problems existing in the prior art, the present invention provides a user-side intelligent power distribution and utilization system, method, medium, device and terminal, and more particularly relates to an Internet-based user-side intelligent power distribution and utilization system, method, medium, device and terminal.

[0005] The present invention is implemented as follows: a user-side intelligent power distribution and utilization system, the user-side intelligent power distribution and utilization system adopts a B / S architecture and uses an MVC design pattern, including a data acquisition and transmission layer, a data layer, a business layer, and an application service layer;

[0006] Among them, the data acquisition and transmission layer interacts with data using a specified communication protocol, with the monitoring substation acting as the data sender and the monitoring center acting as the data receiver; the client and server are deployed separately according to the content of the communication protocol; the client sends the specified data content to the receiver periodically according to the communication protocol based on the data sending frequency requirements; the receiver parses the data after receiving it and stores it in a specified database; data reporting methods include socket data reporting and webservice data reporting.

[0007] The data layer and business layer are responsible for the storage and call of monitoring and scheduling data of each subsystem, shielding the differences in the underlying physical database of the platform operation, and providing various general data access services in the form of interfaces; the system is divided into two categories: real-time database and data center. The real-time database is responsible for various real-time data, including the monitoring quantity of subsystem operation; the data center is responsible for the overall management and storage of historical data, and unified management of various detection data is achieved through this platform.

[0008] The application service layer is used to aggregate different types of applications and services and manage them in a unified manner. The applications and services include energy efficiency evaluation, smart energy allocation, intelligent energy consumption adjustment, data interaction services, and mobile applications. The life cycle of each module and the basic operating information of exception handling are managed by a unified microkernel service manager to implement plug-in, dynamically loaded distributed service-oriented component module application management. Through the design of an open distributed service-oriented application model, flexible combination of various energy efficiency management functions and dynamic behavior decision-making are achieved.

[0009] Furthermore, the user-side intelligent power distribution and utilization system also includes a system setting module, a file management module, a report management module, a map navigation module, a data maintenance module, an alarm management module, an online monitoring module, a power and load prediction module, and an electric energy evaluation module.

[0010] Wherein, the system setting module includes a system configuration submodule, a data dictionary maintenance submodule and a menu management submodule, which are used for the system to require users to display differently according to different permissions;

[0011] The archive management module includes an enterprise archive management submodule, a substation archive management submodule, a transformer archive management submodule, a loop archive management submodule and an equipment archive management submodule; the enterprise archive management submodule is used to add, query, modify, delete, import and export key information of the enterprise; the substation archive management submodule is used to add, query, modify, delete, import and export key information of the substation; the transformer archive management submodule is used to add, query, modify, delete, import and export key information of the transformer; the loop archive management submodule is used to add, query, modify, delete, import and export key information of the loop; the equipment archive management submodule is used to add, query, modify, delete, import and export key information of the equipment;

[0012] The report management module includes an original value report management submodule, an electric quantity statistical report management submodule, a load statistical report management submodule and an alarm report management submodule; the original value report management submodule is used to display, import and export the data collected by the communication machine; the electric quantity statistical report management submodule is used to query, import and export the electric quantity data; the load statistical report management submodule is used to query, import and export the load; the alarm report management submodule is used to query, import and export the alarm information;

[0013] The map navigation module is used to observe all running enterprises from a macro perspective through the map, playing the role of positioning and observation;

[0014] The data maintenance module includes a province management submodule, an urban area management submodule, a district and county management submodule, an industrial zone management submodule and a data dictionary maintenance submodule; the province management submodule is used to add, query, modify and delete the province information; the urban area management submodule is used to add, query, modify and delete the urban area information; the district and county management submodule is used to add, query, modify and delete the district and county information; the industrial zone management submodule is used to add, query, modify and delete the industrial zone information; the data dictionary maintenance submodule is used to add, query, modify and delete the data dictionary information;

[0015] The alarm management module includes an alarm object setting submodule and an alarm strategy setting submodule; the alarm object setting submodule is used to add, query, modify and delete alarm object information; the alarm strategy setting submodule is used to add, query, modify and delete alarm strategies;

[0016] The online monitoring module includes a primary wiring diagram submodule, a power consumption parameter submodule, a power analysis submodule, an alarm management submodule and an indicator association submodule; the primary wiring diagram submodule is used to display the enterprise's primary wiring diagram and monitor the data on the wiring diagram in real time; the power consumption parameter submodule is used to display the data collected by the communication machine and perform trend analysis through visual charts; the power analysis submodule is used to display the data of transformers, substations, loops and equipment, and perform trend analysis using visual charts; the alarm management submodule is used to compare and analyze the alarm information and analyze the number of alarm equipment, year-on-year and month-on-month; the indicator association submodule is used to perform coaxial comparative analysis of power consumption parameters;

[0017] The power and load prediction module is used to predict the power load;

[0018] The electric energy evaluation module is used to evaluate the operation mode and economic operation effect of the enterprise.

[0019] Another object of the present invention is to provide a user-side intelligent power distribution method using the user-side intelligent power distribution system, the user-side intelligent power distribution method comprising the following steps:

[0020] Step 1: Analyze the basic unit of the intelligent electric energy interactive terminal AMI and design a multi-protocol data communication gateway;

[0021] Step 2: Build power quality and energy efficiency evaluation models and optimize them respectively;

[0022] Step 3: Power consumption forecasting and analysis based on artificial intelligence methods, and intelligent optimization guidance system analysis;

[0023] Step 4: Optimize the evaluation of energy efficiency evaluation standards and coordinate the optimization of access control of multiple distributed generation units.

[0024] Furthermore, the intelligent electric energy interactive terminal AMI basic unit in step one is composed of a smart meter, an indoor port and display, and a neighboring user data collection and control center; relevant data is collected through the smart meter, and the cluster meter is managed by the communication management machine. The collected data is sent to the measurement data management system and stored in the local real-time database, so as to realize the collection and control of on-site energy data and energy equipment, and calculate, analyze and count the collected data.

[0025] The multi-protocol data communication gateway includes a protocol conversion APP, TCP / IP protocol, ZigBee protocol, Lora protocol, RTOS real-time operating system and a wireless transceiver module driver, which is used to package the collected data and transmit it to the server; as the root node of the Lora or Zigbee network, it is responsible for converting the information of nearby sites and uploading it to the remote server through the GPRS or NBIOT module.

[0026] Furthermore, the construction of the power quality and energy efficiency evaluation model in step 2 includes:

[0027] (1) Start to build the model and analyze the influencing factors of the model and the basic power data of the object;

[0028] (2) Determine whether the indicator and the adaptation data are compatible. If not, return to step (1); if so, use the hierarchical analysis method to determine the subjective weight;

[0029] (3) Determine whether there is an object that meets the subjective weight target set; if not, return to the subjective weight determination step; if so, analyze and preprocess the basic power data;

[0030] (4) Determine the indicator calculation method and data indicator calculation processing, and use genetic algorithms to determine the objective weights of indicators after classifying the data; normalize the data, use BP neural network to optimize and predict the evaluation data, and then use genetic algorithms to determine the objective weights of indicators;

[0031] (5) Determine which of the normalized and graded objective weights is more consistent with the actual application situation; if the graded method is more consistent, use the graded objective weights; if the normalized method is more consistent, use the normalized objective weights;

[0032] (6) The indicator weights are obtained by combining the weights and the model is established.

[0033] Furthermore, the power consumption prediction and analysis based on artificial intelligence method in step 3 includes:

[0034] (1) Using the improved clustering algorithm Ak-means to achieve cross-dataset learning;

[0035] (2) Data preprocessing, including power load data set expansion, power load data digitization, power load data missing value processing, power load data outlier processing, and power load data standardization;

[0036] (3) Model tuning, including optimizing the time step, optimizing the number of neurons, and selecting the activation function;

[0037] (4) Construct a power load combination prediction model based on recurrent neural network.

[0038] Furthermore, the power load combination prediction model based on recurrent neural network includes:

[0039] The data layer is used to integrate the original data sets of multiple different electricity-consuming enterprises;

[0040] The preprocessing layer uses data preprocessing methods to process the data, and uses the Ak-means clustering algorithm to preliminarily extract the features between different sub-datasets and label the data to achieve cross-dataset training;

[0041] The input layer is used to take all available feature data of the preprocessing layer dataset as network input data;

[0042] The hidden layer uses an improved LSTM or GRU network of the recurrent neural network to automatically learn the feature data given by the input layer and automatically learn the useful information hidden in the data;

[0043] The output layer is responsible for converting the information learned by the hidden layer into the final result.

[0044] Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the user-side intelligent power distribution method.

[0045] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the user-side intelligent power distribution method.

[0046] Another object of the present invention is to provide an information data processing terminal, which is used to implement the user-side intelligent power distribution system.

[0047] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0048] The user-side intelligent power distribution and utilization system provided by the present invention is aimed at the current situation that the power demand side management and energy efficiency management system in the market lack intelligent evaluation and intelligent optimization functions. It is based on the basic functions of intelligent online monitoring of power demand side management and energy efficiency management. It uses computer language module development technology to carry out intelligent monitoring, evaluation and optimization modules for the Internet user-side power distribution and utilization system. At the same time, it supports the import of relevant data such as transmission and distribution network and new energy through interfaces. The present invention realizes data sharing, complementary advantages and complementary functions with the existing power distribution and utilization operation collection and monitoring system and the provincial self-power supply service platform through interfaces, safety management and other modules. At the same time, the present invention combines a low-carbon energy-saving smart management system of a certain city to establish a real-time online monitoring network for power supply and utilization, summarizes multiple data such as energy data statistics and energy audits, realizes online monitoring of power consumption of key units, coordinates the reasonable allocation of power resources, and provides real-time information basis and effect verification for the implementation of energy efficiency evaluation and demand response. In addition, the present invention strengthens the level of power demand side management and energy efficiency management through application demonstration, improves the energy efficiency management and intelligent power distribution and utilization technology in the electric energy service industry, and promotes the development of power distribution and utilization related industries.

[0049] The present invention realizes the following key technologies: 1) Analysis of the basic unit of the intelligent electric energy interactive terminal AMI; 2) Analysis of the multi-protocol data communication gateway; 3) Modeling and implementation of power distribution quality and energy efficiency evaluation; 4) Analysis of the intelligent optimization guidance system; 5) Modeling of power consumption prediction and analysis based on artificial intelligence methods; 6) Design and development of a support platform for the intelligent power distribution system based on the Internet of Things user end; 7) Energy efficiency evaluation standards, evaluation and optimization specifications: Establish a complete set of third-party agency-based management evaluation systems; 8) Service power demand side management evaluation system, standards, evaluation methods, and optimization guidance; 9) Access control and coordinated optimization of multiple distributed power generation units.

[0050] The present invention uses new technologies such as embedded technology, mobile communication technology, sensor technology, intelligent control and cloud services. By tackling key technologies such as data collection at the distribution and utilization end, transmission and protocol intelligent interface design, distribution and utilization power and energy efficiency evaluation, power consumption prediction and analysis based on artificial intelligence methods, and distribution and utilization optimization and solutions, a user-end intelligent power distribution and utilization support platform based on the Internet of Things is developed to form a third-party data service system with device association, protocol intercommunication, data sharing, and multi-task management, and the system will be demonstrated in related enterprises.

[0051] The market of this invention mainly faces three directions: informationization of power management, intelligence of Internet users, and industrialization of power services. It comprehensively uses means and measures such as informationization and standardization to continuously improve the scientific level of electricity consumption of enterprises and the whole society. Coordinate the existing power management system and other resources, and gradually build a power service management platform with functions such as online monitoring, decision analysis, implementation management and statistics, orderly power consumption, and demand response in accordance with the principles of unified design, scientific standardization, openness and effectiveness. Strengthen the application of the platform, realize the "digitalization, networking, visualization, and specialization" of power management, and provide users with systematic power management or professional hosting services.

[0052] The product developed by the present invention is an intelligent power distribution and utilization system based on the Internet user end, which has a positive role in promoting the development of energy efficiency management and demand response of the power demand side management platform, cultivating and developing the corresponding power demand side industry, and driving the power evaluation institutions and power service companies to develop in the direction of sustainable development. At the same time, the implementation of the present invention will form an implementation application demonstration, improve the level of power demand side management and energy efficiency management, and improve power quality management. Power quality management can reduce the economic losses caused by poor quality electricity to the power demand side by improving the power supply quality, reduce additional losses, and thus improve energy utilization efficiency. Based on an accurate assessment of the economic cost of power quality and comparing the relative energy-saving benefits of different power quality management schemes, the power demand side can clearly understand the importance of power quality management, actively take measures to improve power quality, and then improve the overall power quality level of the system, so as to achieve an improvement in the overall social benefits.

[0053] The user-side intelligent power distribution and utilization system provided by the present invention integrates four key functions of data collection and monitoring, power quality and energy efficiency evaluation, data mining and artificial intelligence prediction, and precise optimization management, establishes full-process, closed-loop data flow supervision, and realizes the integration of various business data, scientific management and information sharing services.

[0054] The main technical indicators achieved by the user-side intelligent power distribution and utilization system provided by the present invention when completed are as follows: 1) Accurate collection of parameters such as voltage, current, useful power, useless power, harmonics, and electricity; measurement accuracy of 0.5 level, and the ability to network multiple detection devices; 2) The system platform is stable and continuously operates normally for >720h, and the information is retained for more than 2 years; 3) Alarm output time (platform: less than 5S); Alarm code accuracy: 100%; 4) Ensure that user remote operation is stable, reliable, and safe, with a delay of less than 2S. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0056] Figure 1 It is a flow chart of a user-side intelligent power distribution method provided by an embodiment of the present invention;

[0057] Figure 2 It is a topological structure diagram of a user-side intelligent power distribution and utilization system support platform provided by an embodiment of the present invention;

[0058] Figure 3 It is an application topology diagram of an intelligent electric energy interactive terminal AMI basic unit provided by an embodiment of the present invention;

[0059] Figure 4 It is a schematic diagram of the internal software structure of the multi-protocol data communication gateway provided by an embodiment of the present invention;

[0060] Figure 5 It is a logical framework diagram of a dedicated data communication gateway provided by an embodiment of the present invention;

[0061] Figure 6 is a flow chart of establishing an energy efficiency evaluation model provided by an embodiment of the present invention;

[0062] Figure 7 is a hierarchical structure model diagram provided by an embodiment of the present invention;

[0063] Figure 8 It is a schematic diagram of the BP network structure provided by an embodiment of the present invention;

[0064] Fig. 9 is a comparison diagram of normalized data training curves provided by an embodiment of the present invention;

[0065] Fig.10 is a comparison diagram of hierarchical data training curves provided by an embodiment of the present invention;

[0066] Fig.11 It is a schematic diagram of the structure of a power load combination prediction model based on a recurrent neural network provided in an embodiment of the present invention;

[0067] Fig.12 is a network topology diagram of a system support platform provided by an embodiment of the present invention;

[0068] Fig.13 is a system business logic flow chart provided by an embodiment of the present invention;

[0069] Fig.14 It is a design diagram of the supporting platform software architecture provided by an embodiment of the present invention;

[0070] Fig.15 is a structural diagram of a supporting platform software module provided by an embodiment of the present invention;

[0071] Fig.16 It is a schematic diagram of a power quality evaluation system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0073] In view of the problems existing in the prior art, the present invention provides a user-side intelligent power distribution system, method, medium, device and terminal. The present invention is described in detail below with reference to the accompanying drawings.

[0074] In order to enable those skilled in the art to fully understand how to implement the present invention in detail, this section is an explanatory embodiment that expands and describes the technical solution of the claims.

[0075] like Figure 1 As shown, the user-side intelligent power distribution method provided by the embodiment of the present invention includes the following steps:

[0076] S101, analyze the basic unit of the intelligent electric energy interactive terminal AMI and design a multi-protocol data communication gateway;

[0077] S102, respectively constructing power distribution quality and energy efficiency evaluation models and performing model optimization;

[0078] S103, power consumption forecasting and analysis based on artificial intelligence methods, intelligent optimization guidance system analysis;

[0079] S104, energy efficiency evaluation standard evaluation optimization, access control and coordination optimization of multiple distributed power generation units.

[0080] The intelligent electric energy interactive terminal AMI basic unit provided by the embodiment of the present invention is composed of a smart meter, an indoor port and display, and a neighboring user data collection and control center; relevant data is collected through the smart meter, and the cluster meter is managed by the communication management machine. The collected data is sent to the measurement data management system and stored in the local real-time database, so as to realize the collection and control of on-site energy data and energy equipment, and calculate, analyze and count the collected data.

[0081] The multi-protocol data communication gateway provided in the embodiment of the present invention includes a protocol conversion APP, a TCP / IP protocol, a ZigBee protocol, a Lora protocol, an RTOS real-time operating system, and a wireless transceiver module driver, which is used to package the collected data and transmit it to the server; as the root node of the Lora or Zigbee network, it is responsible for converting the information of nearby sites and uploading it to the remote server through the GPRS or NBIOT module.

[0082] The construction of the power quality and energy efficiency evaluation model provided in the embodiment of the present invention includes:

[0083] (1) Start to build the model and analyze the influencing factors of the model and the basic power data of the object;

[0084] (2) Determine whether the indicator and the adaptation data are compatible. If not, return to step (1); if so, use the hierarchical analysis method to determine the subjective weight;

[0085] (3) Determine whether there is an object that meets the subjective weight target set; if not, return to the subjective weight determination step; if so, analyze and preprocess the basic power data;

[0086] (4) Determine the indicator calculation method and data indicator calculation processing, and use genetic algorithms to determine the objective weights of indicators after classifying the data; normalize the data, use BP neural network to optimize and predict the evaluation data, and then use genetic algorithms to determine the objective weights of indicators;

[0087] (5) Determine which of the normalized and graded objective weights is more consistent with the actual application situation; if the graded method is more consistent, use the graded objective weights; if the normalized method is more consistent, use the normalized objective weights;

[0088] (6) The indicator weights are obtained by combining the weights and the model is established.

[0089] The power consumption prediction and analysis based on artificial intelligence method provided by the embodiment of the present invention includes:

[0090] (1) Using the improved clustering algorithm Ak-means to achieve cross-dataset learning;

[0091] (2) Data preprocessing, including power load data set expansion, power load data digitization, power load data missing value processing, power load data outlier processing, and power load data standardization;

[0092] (3) Model tuning, including optimizing the time step, optimizing the number of neurons, and selecting the activation function;

[0093] (4) Construct a power load combination prediction model based on recurrent neural network.

[0094] The power load combination prediction model based on recurrent neural network provided in the embodiment of the present invention includes:

[0095] The data layer is used to integrate the original data sets of multiple different electricity-consuming enterprises;

[0096] The preprocessing layer uses data preprocessing methods to process the data, and uses the Ak-means clustering algorithm to preliminarily extract the features between different sub-datasets and label the data to achieve cross-dataset training;

[0097] The input layer is used to take all available feature data of the preprocessing layer dataset as network input data;

[0098] The hidden layer uses an improved LSTM or GRU network of the recurrent neural network to automatically learn the feature data given by the input layer and automatically learn the useful information hidden in the data;

[0099] The output layer is responsible for converting the information learned by the hidden layer into the final result.

[0100] The user-side intelligent power distribution and utilization system provided in the embodiment of the present invention adopts a B / S architecture and uses an MVC design pattern, including a data acquisition and transmission layer, a data layer, a business layer, and an application service layer;

[0101] Among them, the data acquisition and transmission layer uses the specified communication protocol to interact with data, the monitoring substation is the data sender, and the monitoring center is the data receiver; according to the content of the communication protocol, the client and server are deployed separately; the client sends the specified data content to the receiver according to the communication protocol based on the data sending frequency requirements; the receiver parses the data after receiving it and stores it in the specified database; the data reporting methods include socket data reporting and webservice data reporting.

[0102] The data layer and business layer are responsible for the storage and call of monitoring and scheduling data of each subsystem, shielding the differences in the underlying physical database of the platform operation, and providing various general data access services in the form of interfaces; the system is divided into two categories: real-time database and data center. The real-time database is responsible for various real-time data, including the monitoring volume of subsystem operation; the data center is responsible for the overall management and storage of historical data, and unified management of various detection data is achieved through this platform.

[0103] The application service layer is used to aggregate and uniformly manage different types of applications and services, including energy efficiency evaluation, intelligent energy allocation, intelligent energy consumption adjustment, data interaction services, and mobile applications. The life cycle of each module and the basic operation information of exception handling are managed by a unified microkernel service manager, which implements plug-in, dynamically loaded distributed service-oriented component module application management. Through the design of an open distributed service-oriented application model, flexible combination of various energy efficiency management functions and dynamic behavior decision-making are achieved.

[0104] The user-side intelligent power distribution and utilization system provided in the embodiment of the present invention also includes a system setting module, a file management module, a report management module, a map navigation module, a data maintenance module, an alarm management module, an online monitoring module, a power and load prediction module, and an electric energy evaluation module.

[0105] Among them, the system setting module includes a system configuration submodule, a data dictionary maintenance submodule and a menu management submodule, which are used for the system to require users to display differently according to different permissions;

[0106] The archive management module includes an enterprise archive management submodule, a substation archive management submodule, a transformer archive management submodule, a loop archive management submodule and an equipment archive management submodule; the enterprise archive management submodule is used to add, query, modify, delete, import and export the key information of the enterprise; the substation archive management submodule is used to add, query, modify, delete, import and export the key information of the substation; the transformer archive management submodule is used to add, query, modify, delete, import and export the key information of the transformer; the loop archive management submodule is used to add, query, modify, delete, import and export the key information of the loop; the equipment archive management submodule is used to add, query, modify, delete, import and export the key information of the equipment;

[0107] The report management module includes an original value report management submodule, an electric quantity statistical report management submodule, a load statistical report management submodule and an alarm report management submodule; the original value report management submodule is used to display, import and export the data collected by the communication machine; the electric quantity statistical report management submodule is used to query, import and export the electric quantity data; the load statistical report management submodule is used to query, import and export the load; the alarm report management submodule is used to query, import and export the alarm information;

[0108] The map navigation module is used to observe all running enterprises from a macro perspective through the map, playing the role of positioning and observation;

[0109] The data maintenance module includes a province management submodule, an urban district management submodule, a district and county management submodule, an industrial zone management submodule and a data dictionary maintenance submodule; the province management submodule is used to add, query, modify and delete the province information; the urban district management submodule is used to add, query, modify and delete the urban district information; the district and county management submodule is used to add, query, modify and delete the district and county information; the industrial zone management submodule is used to add, query, modify and delete the industrial zone information; the data dictionary maintenance submodule is used to add, query, modify and delete the data dictionary information;

[0110] The alarm management module includes an alarm object setting submodule and an alarm strategy setting submodule; the alarm object setting submodule is used to add, query, modify and delete alarm object information; the alarm strategy setting submodule is used to add, query, modify and delete alarm strategies;

[0111] The online monitoring module includes a primary wiring diagram submodule, a power consumption parameter submodule, a power analysis submodule, an alarm management submodule and an indicator association submodule; the primary wiring diagram submodule is used to display the enterprise's primary wiring diagram and monitor the data on the wiring diagram in real time; the power consumption parameter submodule is used to display the data collected by the communication machine and perform trend analysis through visual charts; the power analysis submodule is used to display the data of transformers, substations, loops and equipment, and perform trend analysis using visual charts; the alarm management submodule is used to compare and analyze the alarm information and analyze the number of alarm devices, year-on-year and month-on-month; the indicator association submodule is used to perform coaxial comparative analysis of power consumption parameters;

[0112] The power and load forecasting module is used to forecast the power and load;

[0113] The electric energy evaluation module is used to evaluate the operation mode and economic operation effect of the enterprise.

[0114] In order to prove the creativity and technical value of the technical solution of the present invention, this section provides application examples of the claimed technical solution on specific products or related technologies.

[0115] The embodiments of the present invention have achieved some positive effects during the development or use process, and indeed have great advantages over the prior art. The following content is described in conjunction with data, charts, etc. of the test process.

[0116] The user-side intelligent power distribution and utilization system provided by the embodiment of the present invention is composed of data collection and transmission at the power distribution and utilization end, power quality and energy efficiency evaluation of the power distribution and utilization, power consumption prediction and analysis based on artificial intelligence methods, power distribution and utilization optimization solutions, and a user-side intelligent power distribution and utilization support platform based on the Internet of Things. "Intelligence" refers to the proposal of an intelligent evaluation system based on power quality, energy efficiency evaluation and optimization. The key technologies mainly include evaluation specifications, standards and methods, the development of evaluation tools, and the construction of a support platform. The system is a unified, integrated, high-to-low hierarchical system that includes both vertical hierarchical relationships and parallel relationships between indicators. The evaluation content is for each indicator, the test content, the collected data, and the weighted calculation method.

[0117] 1. Analysis of the basic unit of the intelligent electric energy interactive terminal AMI

[0118] AMI is a complete system that connects the following four basic components through a communication network: smart meters, indoor ports and displays, data collection of nearby users, and control centers. The application topology of the designed smart power interactive terminal AMI basic unit is as follows: Figure 3 As shown. Relevant data is collected through smart meters, and cluster meters are managed through communication management machines. Data is collected and sent to the measurement data management system, stored in the local real-time database, and the collection and control of on-site energy data and energy equipment are realized. The collected data is calculated, analyzed, and counted. The system can be configured with two real-time data / data acquisition servers that are mutually master-slave, to realize data collection and control of the entire energy system, and also serve as local application servers.

[0119] The data of the real-time database is imported into the relational database MySQL of the supporting platform through the network interface. Its main technical indicators such as reliability, stability, data compression ratio, response speed, etc. can meet practical requirements.

[0120] Through the intelligent data acquisition module and the intelligent power meter with digital interface, the present invention realizes real-time monitoring of load power consumption and power quality through the graphical configuration function. The monitoring data includes: voltage, current, active power, reactive power, power factor, active and reactive power, harmonics, environment and switch status, event records and other power parameters. The monitoring objects include: medium and low voltage feeder circuits on the power demand side, major energy-consuming electromechanical equipment, and other energy-consuming facilities in the factory (living area). At the same time, non-electrical data, such as water consumption, gas consumption, heat, feed amount, output, etc., can also be collected through on-site intelligent data such as electronic flow meters, electronic heat meters, electronic belt scales, and floor scales. According to the on-site conditions and system application requirements, the collected data can also be taken from the data interface of other intelligent systems of the user.

[0121] Data transmission network: The platform of the present invention can use the enterprise's exclusive intranet, hotspots, networks provided by network operators (broadband, cellular data), etc. for data communication. Through the energy consumption intelligent data gateway installed in the energy consumption monitoring instrument box (cabinet), the data of the energy consumption meter is collected in real time and transmitted to the smart energy management background of the smart enterprise through the TCP / IP network. No long-distance wiring is required, and the construction is simple and reliable. The data gateway of the present invention provides a variety of access methods, currently supporting RS-485 / RS-232 bus, optical fiber, industrial Ethernet, wireless, GPRS / NBIoT / Lora / WiFi / Bluetooth / Zigbee and other network transmission methods.

[0122] 2. Multi-protocol data communication gateway design

[0123] Its main function is not only to package and transmit the collected data to the server, but also to act as the root node of the Lora or Zigbee network, responsible for uploading the information of nearby sites to the remote server through GPRS or NBIOT module after transformation. This 1-to-many topology has good compatibility and can significantly reduce communication service fees. The internal software structure of the multi-protocol data communication gateway is as follows: Figure 4 shown.

[0124] Its logical framework overall framework design, such as Figure 5 shown.

[0125] The dedicated data communication gateway is based on a high-performance microcontroller. The relevant collection items, collection methods, frequencies and other information are configured through the human-machine interface, such as using buttons / LCD screen settings, or through WiFi+mobile phone APP settings, which is convenient for on-site data observation, parameter input and system debugging. The system also supports GPRS / 3G / 4G and other communication methods, and reports the collected power information, environmental information and on-site images to the cloud service support platform. In addition, the dedicated data communication gateway adds a 485 communication interface with the controller.

[0126] 3. Modeling and implementation of power distribution quality and energy efficiency evaluation

[0127] 3.1 Energy efficiency evaluation model establishment process

[0128] By analyzing the influencing factors of the power quality model in various industries, according to the industry universality principle and the data reading items of common domestic electricity meters, the core indicators of the distribution power quality evaluation model are determined to be the following four items: frequency deviation, power factor, voltage fluctuation, and three-phase imbalance.

[0129] The combined weighting method is used to weight the influencing factors of the power quality evaluation model. For the subjective weight, the target set is formed by collecting the recommended weights given in the literature, and then the hierarchical analysis method is further used to finally obtain the most appropriate weight decision according to the nature of the problem and the overall goal to be achieved.

[0130] Because the BP neural network algorithm has high classification accuracy, strong learning ability, associative memory function, and is very objective, it is very suitable for processing and predicting evaluation data. It can enhance the connection between evaluation results and indicators through the learning of evaluation data by the BP neural network, while eliminating other interference factors. In the case of good fit, it can also expand the labeled sample space through prediction, so as to facilitate the subsequent deriving of objective weights.

[0131] Genetic algorithms can obtain the weights of indicators based on learning the results. Compared with other indicator determination algorithms, they have the advantage of being able to further explore the relationship between indicators and evaluation results. Therefore, genetic algorithms are used to obtain the weight combination with the smallest comprehensive deviation from the evaluation result data, that is, to obtain the objective weight, and then the subjective and objective weights are combined and weighted to obtain the combined weight. The specific model establishment flow chart is as follows Figure 6 shown.

[0132] 3.2 Implementation of power quality and energy efficiency evaluation model

[0133] 1) Calculate subjective weights

[0134] The specific steps are as follows:

[0135] A. Building a hierarchical model

[0136] The three weight set candidate targets are:

[0137] Y1=[0.2044,0.1871,0.2074,0.4011];

[0138] Y2=[0.283,0.234,0.267,0.216];

[0139] Y3=[0.2237,0.1956,0.2122,0.3685];

[0140] According to the four influencing factors of subjective weights: frequency deviation, power factor, voltage fluctuation and three-phase unbalance, a hierarchical analysis model is constructed, such as Figure 7 shown.

[0141] B. Constructing a pairwise comparison matrix

[0142]

[0143] Among them, a14=1 / 3 means that the ratio of the importance of three-phase unbalance to voltage fluctuation is 1 / 3, that is, voltage fluctuation is considered to be more important than three-phase unbalance.

[0144] C. Check consistency

[0145] To test the consistency of the pairwise comparison matrix A, we first need to calculate the CI:

[0146]

[0147] Among them, CI represents an indicator to measure the degree of inconsistency of the pairwise comparison matrix.

[0148] CR<0.1, passing the consistency test, means that although A is not a completely consistent matrix, A has appropriate consistency that can satisfy the test.

[0149] D. Hierarchical total ranking and decision making

[0150] To select a candidate target that is generally most suitable for the above four influencing factors from the three candidate targets Y1, Y2, and Y3, the three-phase unbalance (X1), power factor (X2), frequency (X3), and voltage fluctuation (X4) of the three candidate targets Y1, Y2, and Y3 are compared respectively.

[0151] The calculated scores for the three goals are: [0.40554994 0.34072003 0.25373003]

[0153] Y1 in the target set, i.e. [0.2044, 0.1871, 0.2074, 0.4011], should be selected as the subjective weight of the influencing factors.

[0154] 2) Data preprocessing

[0155] The classic process of using KNN is: first, train the data with complete attributes (HR) to obtain the features corresponding to each complete data, then train each missing data, and then select the k value. If the value of k is too small, once there is a noise component, it will have a greater impact on the prediction. For example, when the k value is 1, once the nearest point is noise, there will be a deviation; if the k value is too large, the approximate error of learning will increase, and at this time, instances far away from the input target point will also affect the prediction, causing the prediction to be wrong. The increase in the k value means that the overall model becomes simpler; and if k = N, then all instances are taken, that is, the most points under a certain category in the instance are taken, which has no practical significance for the prediction; and the value of k should be an odd number as much as possible to ensure that more categories will be generated in the final calculation result.

[0156] Use K-Means algorithm to identify and clean abnormal data of basic data. Through analysis, the main abnormalities of the data come from some voltage and current data that are obviously different from others (too large, too small or negative), the power factor is zero, etc. Whether to delete abnormal values ​​can be considered according to the actual situation. According to the purpose of model research and the actual situation of data, the situation that the voltage and current are too large or too small may be the misalignment of the data transmission process. The situation that the current and voltage are zero is the instantaneous failure of the meter itself and the data loss during the transmission process. As for the negative value of some attribute values, there may be a phenomenon when the error of the meter is greater than the measured current. There is no negative value in the real sense. The operation of deleting the records containing abnormal values ​​can be taken for the detected abnormal points.

[0157] 3) Standardize the data

[0158] In layman's terms, data standardization means scaling the data so that the data can be converted into a desired range. This is used in many evaluation methods. By using data standardization to process the indicators accordingly, the unit limits of the indicators can be discarded, and the indicators of different units can be compared and weighted later.

[0159] There are many data standardization methods, the most typical of which are the linear method, the broken line method and the curve method. Obviously, using different data standardization methods can get different processed data, which will have different effects on the evaluation results of the system, but currently, there is no universal rule to follow.

[0160] Existing methods for standardizing and converting index values ​​include data preprocessing based on mathematical transformation of the upper and lower bounds of the index value domain. According to the national standard for power quality, the index evaluation values ​​are standardized using the upper and lower bounds to try to reduce the data to the range of [0, 1].

[0161] The goal of normalization is to convert numbers into decimals between [0, 1] and to convert dimensional expressions into dimensionless expressions. The benefit of normalization is to improve the convergence speed of the model and also improve the accuracy of the model.

[0162] Common data normalization methods include min-max standardization, z-score standardization, log function conversion, and atan function conversion. The most common min-max standardization is selected here. The conversion function is as follows:

[0163]

[0164] Among them, max is the maximum value of the sample data, and min is the minimum value of the sample data.

[0165] Therefore, in order to facilitate the comparison and weighting of indicators of different units or magnitudes and to improve the convergence speed and accuracy of the model, both the indicator calculation values ​​and evaluation values ​​are normalized.

[0166] 4) Model solution and verification

[0167] A. Use BP neural network to train the A part of the data

[0168] First, we need to clarify two theoretical preparations. The first is the goal of network training, which is to find appropriate weights and thresholds to minimize the error E. The formula for error E is:

[0169]

[0170] The second is the sigmoid function, which refers to the threshold function of the neural network, mapping the variable between 0 and 1. The function formula is:

[0171]

[0172] Figure 8 It is the BP neural network structure.

[0173] Therefore, the change in the output layer weight and threshold can be obtained by formula transformation:

[0174]

[0175] At the same time, the changes in hidden layer weights and thresholds can also be obtained:

[0176]

[0177] B. Use genetic algorithms to derive objective weights

[0178] For genetic algorithms, each chromosome during inheritance will correspond to a preset solution. To determine whether the preset solution is excellent, a fitness function can be used to judge based on the value of the fitness function. Therefore, a mapping is formed from a genome to the fitness of its solution. The specific process of the genetic algorithm is to first determine the fitness of the individual according to the strategy, and then select parents based on the fitness. Individuals with high fitness have a high probability of being selected, and individuals with low fitness are eliminated, and then cross-generation is generated. The chromosomes of the offspring are mutated, and a new generation of populations is generated in this cycle. Some parameters that need to be determined in this process are the maximum evolutionary generation and the recombination probability. Through testing, it is determined that the maximum evolutionary generation of 1500 generations and the recombination probability of 0.7 can obtain a relatively fast convergence speed and accuracy.

[0179] For the determination of the fitness function, the normalized data should calculate the error rate between the output value and the actual value because the evaluation result is a certain value, and then the error rate of all training data is accumulated and divided by the number of training data to obtain the average error rate of the data as the value of the fitness function. The evaluation result of the graded data is actually a range, so the fitness function should not calculate the deviation value for the output value within the range, and calculate the deviation value for the output value not in the range. The determined fitness function is the average deviation rate between the evaluation value obtained by the updated weights of all data and the original evaluation value. The function formula is as follows:

[0180]

[0181] Where: w j Represents the weight of the optimized calculation; a j Indicates the index value of the corresponding weight; N j represents the original evaluation value; m represents the total number of training data.

[0182] C. Result inspection

[0183] The non-training data in the electricity consumption data include test data B, test data C and prediction data D. Test data B will test the accuracy of the test data evaluation after BP neural network learning, and test data C will test the accuracy of the objective weights obtained by the genetic algorithm for the test data evaluation. After the weights are obtained by using the subjective and objective combination of weights, the combined weights will be used to predict the evaluation results of the prediction data D.

[0184] 3.3 Evaluation model optimization

[0185] A. Evaluation method optimization

[0186] When using BP neural network to process data, if the processed results can ensure appropriate fit, the connection between indicators and results and between data and data can be strengthened while ensuring the basic characteristics of the result data. This is of great advantage from the perspective of evaluation and has certain benefits for the subsequent exploration of the direct connection between indicators and results. At the same time, under certain circumstances, by using BP neural network to train the already labeled data and obtain a better fit, more unlabeled data can be predicted and labeled, which can save manpower and material resources while obtaining a larger labeled sample space, which can help to analyze the characteristics of certain aspects of the data more deeply, such as further exploring the influence mechanism of a certain indicator on the evaluation results.

[0187] In order to obtain the weight of the indicator based on the learning results, the genetic algorithm is considered. The genetic algorithm can determine the indicator weight that is most suitable for the evaluation environment by calculating the error between the minimum and the actual results. The genetic algorithm can directly operate on structured data and will not be restricted by the continuity of the data. It does not require definite rules and has very good global search capabilities.

[0188] Therefore, for the power quality evaluation model, on the basis of analyzing the advantages and disadvantages of the existing power quality evaluation methods, a new power quality evaluation model based on BP neural network and genetic algorithm is proposed. The objective weights are obtained through the hierarchical analysis method, and finally the weights are combined.

[0189] B. Data processing optimization

[0190] In the process of evaluating the quality of electricity consumption, the evaluation process can also be optimized by using two different data preprocessing methods: ordinary normalization and normalization and re-classification. These two different data pre-processing methods can be used to explore the impact of data deviation values ​​on the prediction of results and the calculation of weights under different circumstances. In fact, for the acquired data, the above-mentioned processing is performed on it. When using the BP neural network, the degree of optimization of the normalized result data is much greater than that of the classified result data. The normalized data can deepen the connection between the studied electric energy data indicators and the results by learning the global training data while ensuring a good fit. Fig. 9 shown.

[0191] The objective weight results are compared, and the weight data obtained by normalization is relatively balanced. The weight obtained by classification will amplify the importance of a certain indicator, making the weight of a certain indicator account for a larger proportion. Both have their own advantages. For the comparison of the error rates of the two, the error rate of the classified data is lower than that of the normalized data. Both have their own suitable environments, such as Fig.10 shown.

[0192] From the analysis of the results of the grading processing, it can be seen that, in fact, for the four indicators and five levels for each indicator, there are only 625 different types of data in total. Therefore, the prediction of the graded data is more intuitive and clear. Only by marking the 625 data, the predicted evaluation value of all subsequent data can be obtained through the classification algorithm. Therefore, using this method to process data can further save manpower and material resources for data marking, and can make the derivation of objective weights relatively simple. It also shows that the data after grading processing has strong anti-interference ability; and the gap between the final indicator weights can reflect the indicators that have a greater impact on the quality of the studied electric energy data. This disparity in the final results is somewhat inspiring for the future electric energy management for a certain indicator; but the large data overlap also means that there are fewer trainable data, and for the evaluation results, there are defects in data segmentation, and the results cannot be further explored.

[0193] By analyzing the results of normalization processing, it can be concluded that in the case of normalization processing, BP neural network can be used to learn more data to avoid subjective evaluation only for current data and ignore its performance in similar situations. At the same time, because its data and results are more detailed, it can conduct in-depth analysis of certain indicators and results based on subjective factors, which is advantageous in the further research direction of electric energy.

[0194] C. Optimization of industry data acquisition

[0195] Using multiple processing methods can only minimize the limitations of labeled data and increase the global connection of the data, but it cannot eliminate the subjectivity brought by the label itself, and cannot reflect the actual impact represented by different labels and evaluation results. There is a lack of corresponding situations in actual production performance. For example, when the power quality evaluation result is relatively poor, there is no prediction of the actual performance of the corresponding power. Therefore, it may not have a great effect on the subsequent precise management. For the optimization of data acquisition, it should be to add typical related performances of the impact of industry power to the data collection. For example, the power of a certain industry will have an impact on the production volume of the industry. The production volume should be included in the data collection items, and this should be used as the impact result of the indicator. Then, the relationship between production volume and power quality can be accurately matched, and the power can be precisely managed with production volume as the guide.

[0196] 4. Electricity consumption forecasting and analysis based on artificial intelligence methods

[0197] The present invention combines the popular neural network learning algorithm in recent years to study the prediction of massive, multi-source, heterogeneous power load data on the power user side. The prediction methods can be roughly divided into two categories: traditional prediction methods and neural network prediction methods. Although the prediction results on a single data set show that the prediction effects of traditional prediction methods and neural network prediction methods are very good, the prediction capabilities in the case of multiple data sets are basically zero. In the prediction problem of multiple power load data sets, due to the differences in the power load change laws of different data sets and the relatively complex power load change laws of each data set, conventional prediction methods are not suitable for the prediction of multiple power load data sets.

[0198] Considering the massive and multi-source characteristics of power load data, the power load forecasting process needs to face the problems of power big data processing optimization and the inability of power load forecasting models to predict across data sets. The following solutions and optimization measures are proposed for these problems:

[0199] 1) Aiming at the problem of cross-dataset learning, a combined prediction model is proposed, which can achieve cross-dataset model training and prediction while ensuring high prediction accuracy for each data set. The combined prediction model is mainly composed of the improved Ak-means clustering model and the recurrent neural network model proposed in the present invention.

[0200] 2) In response to the problem of poor prediction results of the combined prediction model during initial training, a method for optimizing the structural parameters of the combined prediction network model and a method for reducing the risk of overfitting during the model training process were proposed to further improve the prediction accuracy of the combined model. This mainly includes the optimization of the time step, the determination of the number of hidden layer neurons by the local trial and error method, and the comparative analysis and selection of the activation function, so as to improve the prediction accuracy of the combined model.

[0201] 3) Regarding the selection of recurrent neural network types, the working principles of two improved recurrent neural networks, LSTM and GRU, are analyzed and studied in detail, and the prediction effects of various recurrent neural network models are compared and tested to summarize the differences between LSTM and GRU in power load prediction. LSTM has a stronger learning ability when the number of training times is the same; while GRU has a simpler structure and one less gate structure than LSTM, which means fewer matrix multiplications. GRU can save a lot of time when the training data is large, but because GRU has fewer gate structures, it also means that its learning ability is weaker than LSTM under the same number of learning times, so it usually needs more training times to make up for it.

[0202] 4) In view of the massive amount of power load data, the importance of data processing in power load forecasting is pointed out. Different data processing methods may lead to serious deviations in the forecast results. The upper limit of the power load forecasting problem is determined by data and features, and the role of models and algorithms is to continuously approach this upper limit. A data processing scheme suitable for power load forecasting is proposed: including data set expansion, feature digitization, data cleaning and standardization, and other micro-processed and improved data processing steps to further improve the forecasting accuracy of the combined model.

[0203] 4.1 Improved clustering algorithm Ak-means to achieve cross-dataset learning

[0204] Conventional algorithm models and artificial neural network models are trained and predicted for a single data set. These prediction models are basically one-to-one with a single user. If you want to predict the electricity load of each enterprise, you need to train a prediction model for each enterprise separately, and the training process is too complicated. Cluster analysis can classify the electricity load data sets generated by different enterprises, so as to extract the difference features between multiple data sets, and use cluster analysis to extract the difference features between different data sets to characterize different data sets, and then input the characterization features of different data sets into a conventional prediction model to realize cross-dataset training. Since k-means needs to set the K value in advance and is very sensitive to the selection of the initial centroid point, the present invention improves this problem and proposes the AK-means algorithm.

[0205] 4.2 Data Preprocessing Methods

[0206] 1) Expansion of power load dataset

[0207] In order to improve the prediction accuracy, more effective feature data is obtained, including temperature, weather conditions, rainfall, holidays, and wind speed. These characteristics will affect the characteristic information of the enterprise's electricity load.

[0208] 2) Digitalization of power load data

[0209] Some of the feature data types in the extended data features include integers, floating-point numbers, and strings. Non-numeric data cannot be used for model training. For example, data such as weather, wind speed, and holidays need to be digitized.

[0210] 3) Processing of missing values ​​in power load data

[0211] Using a simple fixed value method or mean method to fill missing values ​​can retain data characteristics and does not affect model training.

[0212] 4) Processing of abnormal values ​​of power load data

[0213] The box plot method is used to identify anomalies, and then the average value of several sample data before and after is used to correct the anomaly.

[0214] 5) Standardization of power load data

[0215] For the power load prediction scenario involved in the present invention, after actual comparative test analysis, the two methods of improved log function standardization and z-score standardization are applied to the standardization of power load data.

[0216] 4.3 Model Tuning Methods

[0217] 1) Optimize time step

[0218] The time step determines how long the neuron can absorb and learn historical data information. The appropriate time step can improve the prediction accuracy of the model.

[0219] 2) Optimization of the number of neurons

[0220] The appropriate number of neurons can reduce the computational overhead of model training without affecting the accuracy of model training.

[0221] 3) Activation function selection

[0222] In the experimental test, it was found that when the conventional sigmoid and tanh activation functions were used, the cost function value tended to stabilize easily during the model training process, and the training process stopped. The trained network model basically did not learn any useful information, and the prediction error would be very large. When the ReLU activation function was used, this problem could be effectively improved, and the prediction error of the model was greatly reduced.

[0223] 4.4 Combined prediction model based on neural network

[0224] The structure of the combined power load forecasting model based on recurrent neural network is as follows: Fig.11 The combined model is divided into five parts. The first layer is the data layer (Data), which is the original data set that integrates multiple different electricity-consuming enterprise data sets.

[0225] The second layer is the preprocessing layer (Preprocess), which uses the above data preprocessing method to process the data, and uses the Ak-means clustering algorithm to preliminarily extract the features between different sub-datasets, and label the data to achieve cross-dataset training.

[0226] The third layer is the input layer, which mainly uses all available feature data of the second layer data set as input data for the network.

[0227] The fourth layer is the hidden layer, which is mainly used to automatically learn the feature data given by the input layer and automatically learn the useful information hidden in the data. This layer uses the improved LSTM or GRU network of the recurrent neural network. LSTM / GRU has a very significant effect in time series prediction because it can learn useful information within a specific time.

[0228] The fifth layer is the output layer, which is responsible for converting the information learned by the hidden layer into the final result.

[0229] 4.5 Analysis of combined prediction model results

[0230] Table 1 Comparison of prediction models

[0231]

[0232] As shown in Table 1, the combined prediction model based on recurrent neural network proposed in the embodiment of the present invention not only realizes model training across data sets, but also ensures high prediction accuracy on sub-data sets, and basically achieves the prediction accuracy under single data set training. Among them, the performance of the combined prediction model based on LSTM and the combined prediction model based on GRU are basically equivalent, and their prediction errors are as low as 4.05% and 3.54% respectively, which is a significant improvement compared with linear regression models and random forest models.

[0233] 5. Analysis of intelligent optimization guidance system

[0234] Monitor high-energy-consuming equipment such as motors and air conditioners to evaluate the electricity conversion rate; at the same time, combine the "power distribution quality and energy efficiency evaluation" data mentioned above to establish an expert-assisted analysis system and provide power distribution optimization plans.

[0235] By collecting energy consumption equipment data by category and item, and combining these energy consumption data with the corresponding equipment and energy consumption, managers can understand and grasp the real-time energy consumption status of each link and key equipment in the building, unit energy consumption data, energy consumption change trends and real-time operating parameters. The main research contents are as follows:

[0236] 1) Real-time monitoring, fault alarm and automatic management of medium and low voltage power distribution systems;

[0237] 2) Realize online monitoring, automatic collection and storage of various energy consumption data and energy efficiency indicators, and provide flexible and diverse query methods;

[0238] 3) Establish a management system for statistical analysis and internal accounting of energy consumption based on physical space, key equipment, and classification and sub-items;

[0239] 4) Carry out independent energy consumption accounting for each public area, realize energy consumption cost sharing, implement energy consumption limit standard management, standardize energy use, and control energy expenditure;

[0240] 5) Through the system's data import / export interface, third-party dedicated energy efficiency analysis tools can be used to help managers discover unreasonable energy usage;

[0241] 6) Provide data basis for energy-saving technology transformation and verify the effect of energy-saving transformation.

[0242] 6. Design and development of a support platform for the intelligent power distribution and utilization system based on the Internet of Things user terminal

[0243] 6.1 System structure design

[0244] The present invention is a support platform integrating data collection, online monitoring, electric energy evaluation, and electric energy prediction. The business logic of the platform is: some important data in the circuit is collected through a communication machine, and the data is transmitted to a database with a written table structure through a communication protocol. The platform displays the data in the database to users in a visual chart, which is convenient for users to perform trend analysis and other commercial purposes. The platform also supports online monitoring of the current power system to prevent accidents and observe the operating status of the system in real time. One of the advantages of the platform is that some collected power parameters are analyzed and evaluated to achieve the effect of electric energy evaluation. Another advantage of the platform is that it can predict load and power through existing data. The larger the amount of data, the more accurate the prediction result will be. Therefore, it is recommended to turn on the electric energy prediction module after the system has been running for a period of time.

[0245] The topological diagram of the supporting platform of the present invention is as follows: Fig.12 shown.

[0246] The system adopts the traditional B / S architecture, that is, the system is accessed by a browser. Therefore, the system uses the MVC design pattern. The MVC design pattern divides the system into three layers: V refers to the View layer, which is the page that interacts with the user and the part that displays data; C refers to the Controller layer, which is to process the data after transmitting it to the background and convert the transmitted data into the data required by the system; M refers to the Model layer, which processes the core business logic in the system. The system generally interacts with the database, so in addition to the above-mentioned view layer, model layer, and control layer, a layer is required to connect to the database. The present invention calls the layer that interacts with the database the persistence layer (DAO layer for short). After using the DAO layer, it is easier to maintain data. In a general system, the business logic of the operation is that the user submits some key information to the Controller layer of the background by operating the View; the Controller performs logical business processing on the transmitted data, and the processing method is to call the Model layer to perform business logic; the Model layer generally exchanges business processing with the database, so in general, the Model layer calls the DAO layer for business processing; finally, the Controller returns the result to the View layer.

[0247] The business logic flow of MVC is as follows Fig.13 shown.

[0248] 6.2 Support platform software architecture design

[0249] In order to reduce system coupling, the supporting platform software adopts a layered design solution, and its software architecture design is as follows: Fig.14 shown.

[0250] 1) Data collection and transmission layer

[0251] The platform uses data integration to integrate the enterprise data collected by various service providers and display them uniformly on the platform according to needs. The connected service providers need to report data according to the development specifications. Currently, the platform provides two data reporting methods: socket data reporting and webservice data reporting.

[0252] The specified communication protocol is used for data exchange, and the monitoring substation is used as the data sender, and the monitoring center is used as the data receiver. According to the content of the communication protocol, the client and the server are deployed respectively. According to the requirements of the data upload frequency, the client sends the specified data content to the receiver according to the communication protocol at a fixed time. After receiving the data, the receiver parses the data and stores it in the specified database.

[0253] To ensure stable and reliable data transmission, the communication protocol meets the following conditions:

[0254] Support real-time data transmission and regular reporting

[0255] Supports breakpoint data transmission

[0256] Support heartbeat packets to judge the communication status in time

[0257] Support identity authentication and filter illegal data

[0258] In this implementation, the mature network communication protocol solution of industrial configuration is used as the designated protocol for data communication between the monitoring subsystem and the monitoring center.

[0259] 2) Data layer and business layer

[0260] It is mainly responsible for the storage and call of monitoring and scheduling data of each subsystem. It can shield the differences in the underlying physical database of the platform operation, provide various general data access services in the form of interfaces, and provide data source support for the pre-release of platform data analysis. The system is mainly divided into two categories: real-time database and data center. The former is mainly responsible for various real-time data, such as the monitoring volume of subsystem operation, etc. The latter is mainly responsible for the overall management and storage of historical data. The unified management of various detection data can be achieved through this platform.

[0261] 3) Application Service Layer

[0262] Different types of applications and services, such as energy efficiency assessment, intelligent energy allocation, intelligent energy adjustment, data interaction services, and mobile applications, are all managed in this application service layer. The life cycle of each module, exception handling and other basic operating information are all managed by the agreed microkernel service manager, realizing plug-in, dynamically loaded distributed service-oriented component module application management. Through the design of an open distributed service-oriented application model, the flexible combination of various energy efficiency management functions and dynamic behavior decision-making are guaranteed.

[0263] Technology Selection

[0264] The front-end and back-end separation was adopted in the development process. The data format for exchange between the front-end and the back-end was JSON, which is the most common data format in web development. The back-end uniformly converts the data into JSON format and transmits it to the front-end for display. The advantage of the front-end and back-end separation is that if the display layer page of the front-end needs to be replaced, the business logic of the back-end does not need to be modified; if the back-end language needs to be replaced, the front-end page does not need to be modified. The selection of software languages ​​is shown in Table 2.

[0265] Table 2 Software language selection

[0266]

[0267]

[0268] 6.3 Support platform software module structure design

[0269] The specific module function table design of the support platform is shown in Table 3.

[0270] Table 3 Function table of specific modules of the supporting platform

[0271]

[0272]

[0273] 7. Energy efficiency evaluation standards, assessment and optimization specifications

[0274] Since energy efficiency evaluation involves a wide range and the funding of the present invention is limited, the present invention only studies the power quality evaluation system standards, evaluation methods, and optimization guidance in the distribution power energy efficiency evaluation.

[0275] 7.1 The significance of establishing power quality evaluation standards

[0276] We should first conduct a reasonable and comprehensive evaluation of the power quality, and then conduct an in-depth analysis of its influencing factors to ensure the high efficiency of power management.

[0277] Before forming these energy efficiency assessment standards, evaluating the power quality and analyzing the indicators of various industries are indispensable preliminary work. Only by accurately evaluating the power quality in the actual environment and then analyzing the indicators accordingly can a reasonable and effective indicator management method be formulated.

[0278] 7.2 Design of power quality evaluation system and evaluation methods and optimization guidance for various indicators

[0279] Here, power quality specifically refers to the quality of the power supply voltage to power users at the public connection point of the distribution grid and the interference level of power users on the public grid at the public connection point of the grid. Special power quality testing is required for typical power quality interference sources within the enterprise. Use a portable power quality analyzer to conduct actual on-site measurements and issue a relevant test report. The technical indicators of the special test include: harmonic current, harmonic voltage, three-phase voltage imbalance, voltage fluctuation and flicker, power factor, frequency deviation, voltage deviation, etc. According to the various power quality indicators, the evaluation system is as follows: Fig.16 shown.

[0280] Power Factor

[0281] Evaluation basis

[0282] The power factor standard is 0.90, which is applicable to industrial users with high voltage power supply of 160 kVA or more (including industrial users of communes and teams), high voltage power supply users equipped with load voltage adjustment devices, and high voltage power supply drainage and irrigation stations with 3200 kVA or more;

[0283] The power factor standard is 0.85, which is applicable to other industrial users (including industrial users of communes and teams) with a power of 100 kVA (kW) or more, non-industrial users with a power of 100 kVA (kW) or more, and electric drainage and irrigation stations with a power of 100 kVA (kW) or more;

[0284] The power factor standard of 0.80 is applicable to agricultural users and wholesale users of 100 kilovolt-amperes (kilowatts) and above. However, for large industrial users who are not classified as wholesale users directly managed by the power company, the power factor standard should be 0.85.

[0285] harmonic

[0286] Evaluation points

[0287] The harmonic voltage (current) should be measured in the minimum operating mode that may occur when the power grid is in normal operation, and should be measured during the period when the harmonics generated during the working cycle of the harmonic source are large (for example, an arc steelmaking furnace should be measured during the melting period).

[0288] The measurement time should include the full production cycle of the harmonic source load operation, which should be no less than 24 hours. For harmonic sources with fast load changes (for example: steelmaking arc furnaces, rolling mills powered by thyristor converters, and electric locomotives), the measurement time window should be no more than 2 minutes.

[0289] Evaluation basis:

[0290] The harmonic voltage (phase voltage) limits of the public power grid are shown in Table 4.

[0291] Table 4 Harmonic voltage limits of public power grid

[0292]

[0293]

[0294] The harmonic current components (root mean square value) injected into the common connection point by all users of the point should not exceed the allowable values ​​specified in Table 5.

[0295] Table 5 Allowable values ​​of harmonic current injected into the common connection point

[0296]

[0297] The permissible value of harmonic current injected into the power grid by each user at the same public connection point is allocated according to the ratio of the agreed capacity of the user at that point to the capacity of the power supply equipment at its public connection point. The allocation calculation method is shown in the following formula.

[0298]

[0299] Among them, I hi It indicates the permissible value of harmonic current injected into the power grid by each user, in ampere (A); I h It represents the allowable value of the hth harmonic current converted for the first time, in ampere (A); α represents the phase superposition coefficient, which can be taken according to Table 6.

[0300] Table 6 Phase superposition coefficient values

[0301] h 3 5 7 11 13 9,>13,even times α 1.1 1.2 1.4 1.8 1.9 2.0

[0302] Data processing and analysis

[0303] The sampling window width of harmonic measurement is 10 power frequency cycles. The root mean square value of j measured values ​​within 3s is taken as the measurement result of the hth harmonic (voltage or current), which should be calculated as follows:

[0304]

[0305] Where: j represents the number of measurements evenly spaced within 3 seconds, which is 15; u h,k represents the hth harmonic value obtained by the kth measurement; U h Indicates the measurement result of the hth harmonic within 3s.

[0306] The data for harmonic measurement should take the largest phase value among the 95% probability maximum values ​​of the actual measured values ​​of each phase within the effective measurement period as the basis for judging whether the harmonic exceeds the allowable value.

[0307] Note: For practical convenience, the 95% probability maximum value of the measured values ​​can be obtained by arranging the measured values ​​in descending order, discarding the first 5% of the large values ​​and taking the maximum value of the remaining measured values.

[0308] Interharmonics

[0309] Evaluation points

[0310] The interharmonic evaluation measurement requires that the system be measured at the minimum normal operation mode and when the interharmonic occurs at the maximum. When the system conditions do not meet the requirements (greater than the normal minimum mode), the result can be converted according to the short-circuit capacity (i.e., the measurement result is multiplied by the ratio of the actual short-circuit capacity to the minimum short-circuit capacity). The measurement time should include the full operating cycle of the harmonic source load, which should not be less than 24 hours.

[0311] Evaluation basis

[0312] The interharmonic voltage content rate at the common connection point (PCC) of the power system of 220kV and below should not exceed the limit values ​​in Table 7.

[0313] Table 7 Interharmonic voltage content rate limit (%)

[0314]

[0315] The interharmonic voltage content rate caused by a single user connected to the PCC shall generally not exceed the limit value in Table 8. This limit value may be appropriately changed according to the load status of the connection point, but it must meet the power factor requirement ①.

[0316] Table 8 Limits of harmonic voltage content rate between single users (%)

[0317]

[0318] Data processing and analysis

[0319] The frequency resolution of interharmonic measurement should be 5Hz, the measurement sampling window width should be 10 power frequency cycles, and the root mean square value of the m-th measurement value within 3s is taken as a measurement result of the ih-th interharmonic voltage, which can be calculated as follows:

[0320]

[0321] Where: U ih represents the measurement result of the ihth harmonic within 3s; m represents the number of measurements evenly spaced within 3s, and m=15 represents seamless sampling; u ih,k It represents the ih-th interharmonic voltage value obtained by the k-th measurement.

[0322] The interharmonic measurement can be based on the 3s measurement results to synthesize the 3min measurement value, which can be calculated as follows:

[0323]

[0324] Where: Indicates the 3-minute measurement value of the ih-th interharmonic; u ih,k Indicates the ihth interharmonic value obtained by the kth measurement. The larger single-phase value among the 95% probability maximum values ​​of the three-phase comprehensive value within the effective evaluation period is used as the evaluation basis.

[0325] Frequency Deviation

[0326] Evaluation points

[0327] Measure the fundamental frequency of the power grid, and take the ratio of the integer cycles counted in the 1s, 3s or 10s interval to the accumulated time of the integer cycles each time (the single cycle overlapping with the 1s, 3s or 10s clock should be discarded). The measurement time intervals cannot overlap, and each 1s, 3s or 10s interval should be counted at the beginning of the 1s, 3s or 10s clock. This basis does not exclude the use of more advanced frequency measurement methods.

[0328] Evaluation basis

[0329] The frequency deviation limit under normal power system operation is ±2% Hz. When the system capacity is small, the deviation limit can be relaxed to ±5% Hz;

[0330] The system frequency change caused by impact load is ±2% Hz, which may also be appropriately changed according to the nature and size of the impact load and the conditions of the system, but the safety, stable operation and normal power supply of the nearby power grid, generator sets and users should be guaranteed.

[0331] Voltage deviation

[0332] Evaluation points

[0333] Changes in system operation mode and changes in user production cycle (load changes) will affect voltage deviation. The monitoring time for voltage deviation assessment should be continuous and should not be less than 24 hours.

[0334] The basic measurement time window for obtaining the voltage RMS value should be 10 cycles, and each measurement time window should be close to the adjacent measurement time window without overlapping. The average value of the voltage RMS value is continuously measured and calculated, and finally the power supply voltage deviation value is calculated. The calculation formula is as follows:

[0335] Voltage deviation (%) = (voltage measurement value - system nominal voltage) / system nominal voltage × 100 (%)

[0336] Evaluation basis

[0337] The sum of the absolute values ​​of positive and negative deviations of the supply voltage of 35kV and above shall not exceed 10% of the nominal voltage; Note: If the upper and lower deviations of the supply voltage have the same sign (both positive or negative), the larger absolute value of the deviation shall be used as the basis for measurement;

[0338] "The three-phase power supply voltage deviation of 20kV and below is ±7% of the nominal voltage";

[0339] “220V single-phase power supply voltage deviation is +7%, -10% of the nominal voltage;

[0340] For users with small short-circuit capacity at the power supply point, long power supply distance, or special requirements for power supply voltage deviation, the requirements shall be determined by agreement between the power supply and the power user.

[0341] Data processing and analysis

[0342] The basic measurement time window for obtaining the voltage effective value should be 10 cycles. The measurement time window should be selected at equal intervals, and the voltage effective value should be continuously measured and calculated to finally obtain the power supply voltage deviation value. The monitoring data should be stored in one group every minute, and the 1-minute voltage effective value should be calculated according to Formula 6.

[0343]

[0344] Where: U min -1 minute voltage effective value; U i -10 cycle voltage effective value; the number of Ui within N-1min is 300.

[0345] Voltage fluctuations and flickers

[0346] Evaluation points

[0347] For the common connection point of the power system, the voltage fluctuation value and the maximum long-term flicker value of at least one week (168h) shall be measured under the minimum mode of normal system operation. For the voltage fluctuation caused by the fluctuating load users at the common connection point of the power system, when they are in normal and continuous working state and the maximum working cycle of the fluctuating load is included, the voltage fluctuation value and the maximum long-term flicker value of at least one day (24h) shall be measured, as well as the background flicker value when the fluctuating load is withdrawn. The observation period of voltage fluctuation should be an integer multiple of 1 hour.

[0348] Evaluation basis

[0349] The voltage fluctuations generated by any fluctuating load user at the common connection point of the power system have limits related to the voltage change frequency and voltage level. The voltage fluctuation limits are shown in Table 9.

[0350] Table 9 Voltage Fluctuation Limits

[0351]

[0352] At the common connection point of the power system, under the minimum mode of normal system operation, with one week (168h) as the measurement period, all long-term flicker values ​​P lt All should meet the flicker limit requirements in Table 10.

[0353] Table 10 Flicker Limits

[0354]

[0355] The flicker value caused by fluctuating load alone is regulated and handled differently at three levels according to the size of the user's load, the proportion of its agreed power consumption capacity to the total power supply capacity, and the status of the public connection point of the power system.

[0356] The first level stipulates that if the requirements of this level are met, it is allowed to connect to the grid without flicker calculation. The first level limit values ​​of LV and MV users are shown in Table 11.

[0357] For HV users, (ΔS / S ac ) max <0.1%.

[0358] Satisfy P lt <0.25 for a single volatile user.

[0359] Low voltage electrical equipment that complies with GB 17625.2 and GB / Z 17625.3.

[0360] Level 2 regulations: The long-term flicker value caused by the fluctuating load alone must be less than the flicker limit value of the load user.

[0361] Level 3 regulations: For individual fluctuating users that do not meet the second level regulations and still exceed their flicker limits after treatment, the limits can be appropriately relaxed based on the actual flicker conditions at the PCC point and the development forecast of the power grid, but the flicker limits at the PCC point must meet the requirements of level 1.

[0362] Table 11 LV and MV user first level limit

[0363]

[0364]

[0365] Data processing and analysis

[0366] Voltage fluctuation is described by the voltage RMS curve U(t). Voltage variation d and voltage variation frequency r are the main indicators to measure the magnitude and speed of voltage fluctuation. The definition of voltage variation d can be expressed as

[0367]

[0368] Where: ΔU is the difference between two adjacent extreme voltages on the voltage RMS curve; U N -Nominal system voltage.

[0369] The voltage RMS value should adopt the effective value algorithm of the half-cycle data between two adjacent zero-crossing points of the waveform, and calculate a RMS value for each half cycle;

[0370] Calculate the voltage variation frequency r and voltage variation d, and take the maximum value of d measured corresponding to r variations as the value of the measurement time window. The time stamp of the data is the end time of the measurement time window.

[0371] The measurement time window for voltage fluctuations should be an integer multiple of 1 hour;

[0372] For the common connection point of the power system, the maximum value of all flicker measurement values ​​shall not exceed the specified value;

[0373] The flicker value caused by the fluctuating load user at the common connection point of the power system is calculated by formula (11), and the maximum value should not be greater than the flicker limit value Ei calculated by formula (3).

[0374]

[0375] Where: P lt2 - Long-term flicker value caused by fluctuating load alone; P lt1 - Long-term flicker measurement value when fluctuating load is put into use; P lt0 -Background flicker value is the long-term flicker measurement value during a period of time when the fluctuating load is withdrawn.

[0376] Calculation method of flicker limit for a single user:

[0377] First, find the total limit value G of flicker generated by all loads connected to the PCC point;

[0378]

[0379] Where: L p ——Long-term flicker value P corresponding to the voltage level at PCC point It Limit value; L H ——Long-term flicker value P of the previous voltage level It Limit value; T - flicker transfer coefficient of the previous voltage level to the next voltage level, recommended to be 0.8. Flicker transfer from the extra-high voltage (EHV) system to the next voltage system is not considered.

[0380] Flicker limit value E for a single user i for:

[0381]

[0382] Where: F is the simultaneous coefficient of fluctuating load, with a typical value of F = 0.2 to 0.3 (but must meet S i / F≤S t ).

[0383] Three-phase unbalance

[0384] Evaluation points

[0385] The measurement should be carried out under the minimum mode (or smaller mode) of normal operation of the power system, when the unbalanced load is in normal and continuous working state, and ensure that the maximum working cycle of the unbalanced load is included.

[0386] For the common connection point of the power system, the measurement duration is one week (168 hours), and the measurement interval of each imbalance can be an integer multiple of 1 minute; for fluctuating loads, continuous measurement can be taken for 24 hours on normal working days, and the measurement interval of each imbalance is 1 minute.

[0387] Evaluation basis

[0388] When the power grid operates normally, the negative sequence voltage imbalance at the public connection point shall not exceed 2%, and shall not exceed 4% in a short time.

[0389] The permissible value of negative sequence voltage unbalance caused by each user connected to the common connection point is generally 1.3%, and not more than 2.6% for a short time. This permissible value can be appropriately changed according to the load conditions of the connection point and the safe operation requirements of the adjacent generators, relay protection and automatic devices.

[0390] Data processing and analysis

[0391] For the common connection point of the power system, the 95% probability maximum value of the 10-minute root mean square value of the negative sequence unbalanced value of the supply voltage should not be greater than 2%, and the maximum value of all measured values ​​should not be greater than 4%. For daily fluctuating unbalanced loads, the 95% probability maximum value of the 1-minute root mean square value of the negative sequence unbalanced value of the supply voltage should not be greater than 2%, and the maximum value of all measured values ​​should not be greater than 4%.

[0392] For each user connected to the common connection point, the 95% probability maximum value of the 1-minute root mean square value of the negative sequence imbalance of the power supply voltage caused should not be greater than 1.3%, and the maximum value of all measured values ​​should not be greater than 2.6%. For practical convenience, the 95% probability maximum value and maximum value of the 1-minute root mean square value of the negative sequence current measurement value injected into the system can be taken, and their values ​​should not be greater than the allowable value of the negative sequence current injected into the system obtained by conversion according to regulations.

[0393] 7.3 Application of Power Quality Evaluation System

[0394] The application of the power quality evaluation system in typical power distribution and consumption industries is shown in Table 12.

[0395] Table 12 Application of power quality index allocation in different industries

[0396]

[0397]

[0398] Design of power quality classification scheme. Power quality is divided into 5 levels: excellent, good, general, qualified, and unqualified. Each indicator of power quality is also divided into the above 5 levels. The specific limit values ​​are evenly divided according to the actual situation. The values ​​of each indicator are excellent, good, general, qualified, and indicators exceeding the limit values ​​are considered unqualified.

[0399] After the levels of each power quality indicator are determined, the final power quality level needs to be determined based on the weights of different power quality indicators in various industries. In fact, in the evaluation and research of power quality, on the basis of calculating the objective weights of each determined indicator, it is also possible to abandon the indicators with too low weights and reselect the indicators. That is, if a certain indicator shows that the impact on the result is almost negligible in the actual calculation, the indicator can be abandoned, and other suitable indicators can be selected for calculation. In this way, the indicators and their influence coefficients that really affect the power quality evaluation results under study can be obtained, which has certain inspiration for further realizing the formulation of industry standards.

[0400] The classification scheme for each power quality indicator is shown in Table 13.

[0401] Table 13 Classification of power quality indicators

[0402]

[0403] 8. Access control and coordination optimization of multiple distributed generation units

[0404] By installing distributed power sources such as photovoltaic power generation, geothermal power generation, wind power generation, and energy storage devices in the system, and deploying control interfaces and monitoring software, bidirectional metering of distributed power sources, monitoring of the operating status of distributed power sources on the user side, and grid-connected control can be achieved; comprehensive energy demand, electricity prices, fuel consumption, power quality requirements, etc., combined with energy storage devices, local consumption and optimized coordinated control of distributed power sources in the park can be achieved, and distributed power sources can participate in peak shifting of the power grid. At the same time, safe monitoring of the power supply and operation status of the park and real-time monitoring of power quality can be achieved. Develop and deploy intelligent detection and isolation equipment for power distribution lines and equipment faults in the park, and respond quickly to faults to achieve self-healing power supply. Relying on the power distribution SCADA master station system, remote monitoring and control of the power supply equipment status can be achieved; support information integration with the master station of the park management center to improve fault response and processing speed.

[0405] 8.1 Wind power generation energy supply control model

[0406] Comprehensively monitor the real-time operation information and alarm information of wind turbine power generation, and conduct statistics and analysis on wind turbine power generation in various aspects to achieve comprehensive control of wind turbine power generation. Wind power generation energy management provides the following specific functions: real-time display of the current total power generation, daily total power generation, cumulative total power generation, cumulative CO2 emission reduction and daily power generation curve of wind power generation; collect wind turbine operation status data including: three-phase voltage, three-phase current, grid frequency, power factor, output power, etc.; real-time monitoring and analysis of the power quality of the grid connection point.

[0407] 8.2 Photovoltaic power supply control model

[0408] Photovoltaic power generation monitoring can display the following information: it can display the current total power generation of photovoltaics, the total daily power generation, the cumulative total power generation, and the daily power generation curve in real time. It monitors the operating status of all inverters, uses sound and light alarms to indicate equipment failures, and can monitor and analyze the power quality of the grid connection point in real time.

[0409] 8.3 Energy storage supply control model

[0410] Comprehensively monitor the real-time operation information and alarm information of the energy storage battery, and conduct statistics and analysis on energy storage in many aspects to achieve comprehensive control of energy storage. Energy storage monitoring can display the following information: 1) It can display the current dischargeable amount, chargeable amount, maximum discharge power, current discharge power, dischargeable time, total charge amount today, and total discharge amount today of the energy storage in real time. 2) It can remotely signal the operation status and alarm information of the AC / DC bidirectional converter, among which the protection signals include: low voltage protection, overvoltage protection, phase loss protection, low frequency protection, overfrequency protection, overcurrent protection, device abnormality protection, battery pack abnormal condition protection, and overtemperature protection. 3) Telemetry: It can remotely measure the battery voltage, battery charge and discharge current, AC voltage, input and output power, etc. of the AC / DC bidirectional converter. 4) Remote adjustment: It can remotely adjust the battery charge and discharge time, charge and discharge current, and battery protection voltage to achieve remote adjustment of the relevant parameters of the AC / DC bidirectional converter. 5) Remote control: It can remotely control the battery charging and discharge of the AC / DC bidirectional converter.

[0411] 8.4 Multi-energy complementary energy supply control model

[0412] On the basis of completing load monitoring, distributed energy monitoring and management, and integrated energy management, a unified multi-energy complementary high-efficiency energy conversion and control model is established: 1) Monitor the voltage, frequency, and phase angle of the distribution network at the grid entrance PCC, as well as the power of the distribution network, the voltage, frequency, and phase angle of the grid system. 2) Real-time statistics of the total power generation output of the microgrid, the remaining capacity of energy storage, the total active load of the microgrid, the total reactive load, the total active power of sensitive loads, the total active power of controllable loads, and the total active power and reactive power of fully removable loads.

[0413] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0414] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A user-side intelligent power distribution system, characterized in that: The user-side intelligent power distribution and utilization system adopts B / S architecture and uses MVC design mode, including data collection and transmission layer, data layer, business layer and application service layer; The data acquisition and transmission layer uses a specified communication protocol to exchange data, and the monitoring substation acts as the data transmitter, and the monitoring center acts as the data receiver; the client and the server are deployed respectively according to the content of the communication protocol; the client sends the specified data content to the receiver according to the communication protocol according to the data transmission frequency requirement; the receiver parses the data after receiving it and stores it in a specified database; the data reporting methods include socket data reporting and webservice data reporting; The data layer and business layer are responsible for the storage and call of monitoring and scheduling data of each subsystem, shielding the differences of the underlying physical database of the platform operation, and providing various general data access services in the form of interfaces; the system is divided into two categories: real-time database and data center. The real-time database is responsible for various real-time data, including the monitoring amount of subsystem operation; the data center is responsible for the overall management and storage of historical data, and realizes the unified management of various detection data through this platform; The application service layer is used to aggregate and uniformly manage different types of applications and services, including energy efficiency evaluation, intelligent energy allocation, intelligent energy adjustment, data interaction services, and mobile applications. The life cycle of each module and the basic operation information of exception handling are managed by a unified microkernel service manager, realizing plug-in, dynamically loaded distributed service-oriented component module application management. Through the design of an open distributed service-oriented application model, flexible combination of various energy efficiency management functions and dynamic behavior decision-making are realized. The user-side intelligent power distribution and utilization method using the user-side intelligent power distribution and utilization system comprises the following steps: Step 1: Analyze the basic unit of the intelligent electric energy interactive terminal AMI and design a multi-protocol data communication gateway; Step 2: Build power quality and energy efficiency evaluation models and optimize them respectively; Step 3: Power consumption forecasting and analysis based on artificial intelligence methods, and intelligent optimization guidance system analysis; Step 4: Optimize the evaluation of energy efficiency evaluation standards and coordinate the optimization of access control of multiple distributed generation units; The construction of the power distribution quality and energy efficiency evaluation model in step 2 includes: (1) Start to build the model and analyze the influencing factors of the model and the basic power data of the object; (2) Determine whether the indicator and the adaptation data are compatible. If not, return to step (1); if so, use the hierarchical analysis method to determine the subjective weight; (3) Determine whether there is an object that meets the subjective weight target set; if not, return to the subjective weight determination step; if so, analyze and preprocess the basic power data; (4) Determine the indicator calculation method and data indicator calculation processing, and use genetic algorithms to determine the objective weights of indicators after classifying the data; normalize the data, use BP neural network to optimize and predict the evaluation data, and then use genetic algorithms to determine the objective weights of indicators; (5) Determine which of the normalized and graded objective weights is more consistent with the actual application; if the graded method is more consistent, use the graded objective weights; if the normalized method is more consistent, use the normalized objective weights; (6) Combine and weight the indicators to obtain their weights, and the model is established; The user-side intelligent power distribution and utilization system also includes a system setting module, a file management module, a report management module, a map navigation module, a data maintenance module, an alarm management module, an online monitoring module, a power and load prediction module, and an electric energy evaluation module; Wherein, the system setting module includes a system configuration submodule, a data dictionary maintenance submodule and a menu management submodule, which are used for the system to require users to display differently according to different permissions; The archive management module includes an enterprise archive management submodule, a substation archive management submodule, a transformer archive management submodule, a loop archive management submodule and an equipment archive management submodule; the enterprise archive management submodule is used to add, query, modify, delete, import and export key information of the enterprise; the substation archive management submodule is used to add, query, modify, delete, import and export key information of the substation; the transformer archive management submodule is used to add, query, modify, delete, import and export key information of the transformer; the loop archive management submodule is used to add, query, modify, delete, import and export key information of the loop; the equipment archive management submodule is used to add, query, modify, delete, import and export key information of the equipment; The report management module includes an original value report management submodule, an electric quantity statistical report management submodule, a load statistical report management submodule and an alarm report management submodule; the original value report management submodule is used to display, import and export the data collected by the communication machine; the electric quantity statistical report management submodule is used to query, import and export the electric quantity data; the load statistical report management submodule is used to query, import and export the load; the alarm report management submodule is used to query, import and export the alarm information; The map navigation module is used to observe all running enterprises from a macro perspective through the map, playing the role of positioning and observation; The data maintenance module includes a province management submodule, an urban area management submodule, a district and county management submodule, an industrial zone management submodule and a data dictionary maintenance submodule; the province management submodule is used to add, query, modify and delete the province information; the urban area management submodule is used to add, query, modify and delete the urban area information; the district and county management submodule is used to add, query, modify and delete the district and county information; the industrial zone management submodule is used to add, query, modify and delete the industrial zone information; the data dictionary maintenance submodule is used to add, query, modify and delete the data dictionary information; The alarm management module includes an alarm object setting submodule and an alarm strategy setting submodule; the alarm object setting submodule is used to add, query, modify and delete alarm object information; the alarm strategy setting submodule is used to add, query, modify and delete alarm strategies; The online monitoring module includes a primary wiring diagram submodule, a power consumption parameter submodule, a power analysis submodule, an alarm management submodule and an indicator association submodule; the primary wiring diagram submodule is used to display the enterprise's primary wiring diagram and monitor the data on the wiring diagram in real time; the power consumption parameter submodule is used to display the data collected by the communication machine and perform trend analysis through visual charts; the power analysis submodule is used to display the data of transformers, substations, loops and equipment, and perform trend analysis using visual charts; the alarm management submodule is used to compare and analyze the alarm information and analyze the number of alarm equipment, year-on-year and month-on-month; the indicator association submodule is used to perform coaxial comparative analysis of power consumption parameters; The power and load prediction module is used to predict the power load; The electric energy evaluation module is used to evaluate the operation mode and economic operation effect of the enterprise; Power quality and energy efficiency evaluation model implementation: 1) Calculate subjective weights The specific steps are as follows: A. Building a hierarchical model The three weight set candidate targets are: Y1＝[0.2044,0.1871,0.2074,0.4011]； Y2＝[0.283,0.234,0.267,0.216]； Y3＝[0.2237,0.1956,0.2122,0.3685]； According to the four influencing factors of subjective weights: frequency deviation, power factor, voltage fluctuation and three-phase unbalance, a hierarchical analysis model is constructed; B. Constructing a pairwise comparison matrix where a 14 =1 / 3 means that the ratio of the importance of three-phase unbalance to voltage fluctuation is 1 / 3, that is, voltage fluctuation is considered to be more important than three-phase unbalance; C. Check consistency To test the consistency of the pairwise comparison matrix A, we first need to calculate the CI: Among them, CI represents an indicator to measure the degree of inconsistency of the pairwise comparison matrix; CR<0.1, passing the consistency test, which means that although A is not a completely consistent matrix, A has appropriate consistency that can satisfy the test; D. Hierarchical total ranking and decision making To select a candidate target that is generally most suitable for the above four influencing factors from the three candidate targets Y1, Y2, and Y3; to this end, the three candidate targets Y1, Y2, and Y3 are compared with their three-phase unbalance X1, power factor X2, frequency X3, and voltage fluctuation X4 respectively; After calculation, the scores of the three goals are: [0.40554994 0.34072003 0.25373003] Y1 in the target set, i.e. [0.2044, 0.1871, 0.2074, 0.4011], should be selected as the subjective weight of the influencing factors; 2) Data preprocessing The classic process of using KNN is: first, train the data with complete attribute HR to obtain the features corresponding to each complete data, then train the data for each missing data, and then select the k value. If the value of k is too small, once there is a noise component, it will have a greater impact on the prediction. When the k value is 1, once the nearest point is noise, there will be a deviation; if the value of k is too large, the approximate error of learning will increase, and at this time, it will also affect the prediction when it is far away from the input target point, causing the prediction to be wrong; the increase in the k value means that the overall model becomes simpler; and if k = N, then all instances are taken, that is, the most points under a certain category in the instance are taken, which has no practical significance for the prediction; and the value of k should be an odd number as much as possible to ensure that more categories will be generated in the final calculation result; Use K-Means algorithm to identify and clean abnormal data of basic data; through analysis, the data anomaly comes from some voltage and current data that are too large, too small or negative, and the power factor is zero. Whether to delete abnormal values ​​can be considered according to the actual situation. According to the purpose of model research and the actual situation of data, the situation that the voltage and current are too large or too small is the misalignment of the data transmission process. The situation that the current and voltage are zero is the instantaneous failure of the meter itself and the data loss during the transmission process. As for the negative value of some attribute values, when the error of the meter is greater than the measured current, this phenomenon will occur. There is no negative value in the real sense. For the detected abnormal points, the operation of deleting the records containing abnormal values ​​is taken; 3) Standardize the data In layman's terms, data standardization means scaling the data in proportion, so that the data can be converted into a desired range; this is used in many evaluation methods. By using data standardization to process the indicators accordingly, the unit limits of the indicators can be discarded, and the indicators of different units can be compared and weighted later; There are many data standardization methods, the most typical of which are the linear method, the broken line method and the curve method. Obviously, using different data standardization methods can get different processed data, which will have different impacts on the evaluation results of the system. However, there is no universal rule to follow. The existing method of standardizing and converting the index value is to preprocess the data based on the mathematical transformation of the upper and lower bounds of the index value domain; according to the national standard of power quality, the index evaluation value is standardized by the upper and lower bounds, so as to try to summarize the data into the range of [0, 1]; The goal of normalization is to convert numbers into decimals between [0, 1] and to convert dimensional expressions into dimensionless expressions. The benefit of normalization is to improve the convergence speed of the model and also improve the accuracy of the model. Common data normalization methods include min-max standardization, z-score standardization, log function conversion, and atan function conversion. The most common min-max standardization is selected here. The conversion function is as follows: Where max is the maximum value of the sample data, and min is the minimum value of the sample data; Therefore, in order to facilitate the comparison and weighting of indicators of different units or magnitudes and to improve the convergence speed and accuracy of the model, both the indicator calculation value and the evaluation value are normalized; 4) Model solution and verification A. Use BP neural network to train the A part of the data First, we need to clarify two theoretical preparations. The first is the goal of network training, which is to find appropriate weights and thresholds to minimize the error E. The formula for error E is: The second is the sigmoid function, which refers to the threshold function of the neural network, mapping the variable to between 0 and 1. The function formula is: Therefore, the change in the output layer weight and threshold can be obtained by formula transformation: At the same time, the changes in hidden layer weights and thresholds can also be obtained: B. Use genetic algorithms to derive objective weights For genetic algorithms, each chromosome during inheritance will correspond to a preset solution. To determine whether the preset solution is excellent, a fitness function can be used to judge based on the value of the fitness function; so a mapping is formed from a genome to the fitness of its solution; the specific process of the genetic algorithm is to first determine the fitness of the individual according to the strategy, and then select parents based on the fitness. Individuals with high fitness have a high probability of being selected, and individuals with low fitness are eliminated, and then cross-generation is generated. The chromosomes of the offspring are mutated, and this cycle is repeated to generate a new generation of populations; some parameters that need to be determined in this process are the maximum evolutionary generation and the recombination probability. Through testing, it is determined that the maximum evolutionary generation of 1500 generations and the recombination probability of 0.7 can obtain a relatively fast convergence speed and accuracy; For the determination of the fitness function, the normalized data should calculate the error rate between the output value and the actual value because the evaluation result is a certain value, and then the error rate of all training data is accumulated and divided by the number of training data to obtain the average error rate of the data as the value of the fitness function. The evaluation result of the graded data is actually a range, so the fitness function should not calculate the deviation value for the output value within the range, and calculate the deviation value for the output value not in the range. The determined fitness function is the average deviation rate between the evaluation value obtained by the updated weights of all data and the original evaluation value. The function formula is as follows: Where: w i Represents the weight of the optimized calculation; a i Indicates the index value of the corresponding weight; N j represents the original evaluation value; m represents the total number of training data; C. Result inspection The non-training data in the electricity consumption data include test data B, test data C and prediction data D. Test data B will test the accuracy of the test data evaluation after BP neural network learning, and test data C will test the accuracy of the objective weights obtained by the genetic algorithm for the test data evaluation. After the weights are obtained by using the subjective and objective combination of weights, the combined weights will be used to predict the evaluation results of the prediction data D.

2. The user-side intelligent power distribution system according to claim 1, characterized in that: The smart electric energy interactive terminal AMI basic unit in step 1 is composed of a smart meter, an indoor port and display, and a neighboring user data collection and control center; the smart meter collects relevant data, the communication management machine manages the cluster meter, collects data and sends it to the measurement data management system, stores it in the local real-time database, realizes the collection and control of on-site energy data and energy equipment, and calculates, analyzes and statistics the collected data; The multi-protocol data communication gateway includes a protocol conversion APP, TCP / IP protocol, ZigBee protocol, Lora protocol, RTOS real-time operating system and wireless transceiver module driver, which is used to package the collected data and transmit it to the server; As the root node of the Lora or Zigbee network, it is responsible for converting the information of nearby sites and uploading it to the remote server through the GPRS or NBIOT module.

3. The user-side intelligent power distribution system according to claim 1, characterized in that: The power consumption prediction and analysis based on artificial intelligence method in step 3 includes: (1) Using the improved clustering algorithm Ak-means to achieve cross-dataset learning; (2) Data preprocessing, including power load data set expansion, power load data digitization, power load data missing value processing, power load data outlier processing, and power load data standardization; (3) Model tuning, including optimizing the time step, optimizing the number of neurons, and selecting the activation function; (4) Construct a power load combination prediction model based on recurrent neural network.

4. The user-side intelligent power distribution system as claimed in claim 3, characterized in that: The power load combination prediction model based on recurrent neural network includes: The data layer is used to integrate the original data sets of multiple different electricity users; The preprocessing layer uses data preprocessing methods to process the data, and uses the Ak-means clustering algorithm to preliminarily extract the features between different sub-datasets and label the data to achieve cross-dataset training; The input layer is used to take all available feature data of the preprocessing layer dataset as network input data; The hidden layer uses an improved LSTM or GRU network of the recurrent neural network to automatically learn the feature data given by the input layer and automatically learn the useful information hidden in the data; The output layer is responsible for converting the information learned by the hidden layer into the final result.

5. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the user-end smart power distribution method of the user-end smart power distribution system according to any one of claims 1 to 4.

6. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the user-side intelligent power distribution and utilization method of the user-side intelligent power distribution and utilization system according to any one of claims 1 to 4.

7. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the user-side intelligent power distribution system as described in any one of claims 1 to 4.

Images