Distributed multi-element equipment agent aggregation double-layer management and control system in regional power distribution network

Abstract

The invention discloses a distributed multi-element equipment agent aggregation double-layer management and control system in a regional power distribution network, and belongs to the technical field of active power distribution network intelligent management and control. The invention aims to establish a physical-information mapping agent of multi-element controllable equipment in a power distribution network, and provides a distributed multi-element equipment agent aggregation double-layer management and control system in a regional power distribution network, which is based on a software information model fused with an IEC61850 protocol and a multi-agent general language and is combined with hierarchical and partitioned intelligent aggregation management. According to the method, an IEC61850-based information description model and an agent language description model combined with an MAS technology are designed, and then on the basis of the model, bottom-layer optimal energy management of the multi-agent intelligent aggregation system and maximum income regulation and control of an upper-layer multi-agent cooperative game of the multi-agent intelligent aggregation system are designed. According to the method, the interactive behavior characteristics of the multi-element controllable equipment are described through effective software information of the active power distribution network, and the intelligent control effect of the multi-element controllable equipment in the active power distribution network is remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent control of active power distribution networks. Background Art

[0002] Various distributed multi-devices are constantly connected to the distribution network, such as distributed photovoltaic, wind power, distributed energy storage, electric vehicle charging piles, etc. However, these distributed power devices are usually scattered in location, large in number, and have impact and random power, which brings new problems to the operation of the distribution network. The volatility of distributed power sources may affect the voltage stability of the power grid. The access of distributed power sources will change the distribution of power flow and affect the network loss of the power grid. At the same time, these devices have better flexibility and controllability, and have the underlying foundation of intelligent management and control. Effective management and control of massive distributed multi-devices has become an important problem for the safe operation of new distribution networks.

[0003] However, currently distributed multi-devices come from different manufacturers, and the communication protocols of the devices of various manufacturers are inconsistent, the communication methods are also different, and there is a lack of unified communication access equipment. Furthermore, as a power source, distributed multi-devices currently only perform electricity metering, lack an information description model, cannot reflect their operating behavior, and cannot give full play to the power flexible control capabilities of distributed multi-devices. In order to realize the intelligent management and control of distributed multi-devices, a more advanced management and control method is to perform physical-information mapping on distributed multi-controllable devices and describe the operating characteristics of distributed multi-devices through software agent models.

[0004] Single distributed devices have low power, scattered locations, and large numbers. The separate management of massive small-capacity distributed power sources will increase the complexity of grid management and control, and may increase the management burden and cost of grid operators. At present, the power company's operating guidelines require at least 1MW of capacity to participate in power market transactions. It is difficult for individual small-capacity distributed power sources to enter the power market and lack effective means of profit. Therefore, it is a more reasonable and feasible way to aggregate multiple distributed power sources with similar geographical locations and electrical connections into an overall system and connect them to the power market in a unified manner through an aggregate system.

[0005] Patent CN117094843 discloses a method for mapping real-time data and event information of distribution network. The invention obtains real-time data and event information of distribution network in distribution network system, pre-processes data and event information, establishes a mapping model of real-time data and event information of distribution network, monitors data and event information in distribution network system in real time, and feeds back to the mapping model in real time. Data and event information are analyzed through the mapping model to realize the management and processing of data and event information in distribution network system. However, the accuracy of the mapping model depends on the quality of input data. If there are errors or incompleteness in real-time data or event information, the accuracy of the mapping result may be affected.

[0006] Patent CN117094843A discloses a distribution physical-information system optimization method, system, storage medium and computing device. The invention achieves dual optimization of the reliability and economy of the distribution system by establishing a distribution physical-information system collaborative planning logical architecture that integrates the target layer, planning layer, and operation layer. The method first establishes a unified description model based on the logical architecture, and performs a double-layer coupling reliability and comprehensive economic analysis and evaluation on the distribution physical-information system. According to the evaluation results, a hierarchical collaborative optimization model is determined, and an elite genetic algorithm is used to solve it to achieve a balance between system reliability and economy. However, in practical applications, it may take a long time to calculate, especially when facing large-scale distribution systems.

[0007] Patent CN117808244A discloses a distributed resource data processing method, device and electronic device. The invention constructs a distributed power supply model and an adjustable load model, aggregates these models, obtains a distributed resource monomer model, and then constructs a scheduling model and an operation model, and finally obtains the target aggregation result through aggregation processing. This method solves the problem of distributed resource storage, thereby alleviating the difficulties in carrying out distribution network control business. However, the invention relies on specific technologies or algorithms, which limits its applicability in different technical backgrounds.

[0008] In summary, although the currently disclosed patents have involved solutions for distributed multi-device physical-information mapping dual-layer optimization aggregation, they are basically rule-based design methods. These methods have limited applicability and flexibility in different technical backgrounds, making it difficult to achieve optimal intelligent management and market profitability. Summary of the invention

[0009] The purpose of the present invention is to establish a physical-information mapping intelligent agent for multiple controllable devices in a distribution network, and propose a software information model based on the IEC61850 protocol and the multi-agent systems (MAS) common language (Agent Communication Language, ACL), combined with a distributed multi-device intelligent agent aggregation double-layer management and control system in a regional distribution network with hierarchical and partitioned intelligent aggregation management.

[0010] The steps of the present invention are: S1. Design a local power grid management framework system for intelligent aggregation of multiple controllable devices in active distribution networks: Based on the geographical area and management level of the equipment, it is divided into a distributed controllable equipment agent layer, a multi-agent aggregation layer, and a multi-aggregation layer to form a regional layer; S2. Design based on IEC61850 information description: Based on the information protocol framework of the international standard IEC6185ED.20 in the field of power system automation and the characteristics of multiple controllable devices in active distribution networks, a standardized IEC61850 XML extensible markup language XML model file is established. S3. Design an agent language description model based on the IEC61850 information description model and MAS technology: It integrates IEC61850 information description and agent common language ACL to describe the physical information mapping model, and describes intelligent behaviors such as agent information sharing, dynamic characteristics and interactive control; S4. Design the underlying optimal energy management of the multi-agent intelligent aggregation system based on the agent information model: It uses a two-layer optimization aggregation technology, with the bottom layer using optimal energy management based on predictive control; S5. Based on the agent information model, design the maximum benefit regulation of the multi-agent cooperative game in the upper layer of the multi-agent intelligent aggregation system: A cooperative game algorithm is used to achieve the high efficiency of power resources in the active distribution network in the multi-aggregate optimal configuration and the fairness of the aggregate system in the distribution of power market benefits.

[0011] The multi-agent intelligent aggregation system of the present invention is a dual-layer management and control system of multi-agents at the aggregation layer: optimal operation cost management within the lower layer within the day; Multi-resource energy aggregation layer model: in represents the electricity imported from the grid by the aggregator at time t, Indicates the RES power in maximum power point tracking mode; respectively represents the natural gas consumed by CHP and GF; and are the gas-to-electricity efficiency and gas-to-heat efficiency of CHP; Indicates the amount of electricity consumed by EB; and are the efficiencies of G and EB, respectively; and Represent the charging power and discharging power of BESS respectively; and are the charge and discharge amounts of TES respectively; and Indicates the value of electrical and thermal energy losses; and Represents movable electrical and thermal loads; L E,t and L H,t represent fixed electrical load and thermal load respectively; Aggregation layer energy management objective function and constraints: Operating costs are divided into two parts: electricity costs and natural gas costs where μ e,t and μ g,t They represent the electricity price and natural gas price for time period t given by the superior respectively; Other constraints include: in and Indicates the maximum exchange power through the connecting line; and represents the installed capacity of the combined heat and power pump, GF and EB respectively; P CHP , P GF and P EB Indicates the lower limit of the corresponding power; and Indicates the maximum charge and discharge power of BESS; and Respectively represent the maximum value of TES charging and discharging; Use MPC to correct the deviation between the actual daily operation plan and the previous day's scheduling plan: The state variables in the model are: x(k)=[P e P RES P CHP P EB P BESS H GF H EB H TES ] (9) The control variables in the model are: u(k)=[ΔP e ΔP RES ΔP CHP ΔP EB ΔP BESS ΔH GF ΔH EB ΔH TES ] (10) The disturbance variable in the model is: r(k)=[ΔP Load ΔH Load ΔP RES ] (11) Create the following objective function: The power control quantity of the aggregation layer is obtained after calculating the target (9) and sent to the upper regional controller. The regional controller will calculate the power balance of the distribution network system and correct the market price.

[0012] The present invention's regional layer multi-aggregate optimal dispatch: participating in the optimal dispatch of the demand-side power market on the day before Optimization problem modeling: Where n is the index of the aggregator; N is the total number of aggregators; is the input power of the main transformer; is the maximum exchange power; The relaxed Lagrangian dual problem based on equation (13) is: Given λ t Formula (15) is decomposed into N+1 sub-problems, corresponding to N aggregators: Transformer for If the Lagrange multiplier λ t In Eqs. (16) and (17), it is interpreted as the price, the local price signal λ e,t Defined as: λ e,t =μ e,t +λ t . (18) The optimization problem of formula (13) is solved in a two-layer framework, that is, in the upper-level region, the controller adjusts the local price vector to achieve an overall balance between supply and demand, and in the lower-level region, each aggregator autonomously minimizes its cost under the price vector; in each iteration, each aggregator solves subproblem (16) with the provided price vector and submits the optimal power vector as a bid, while the transformer obtains the optimal power vector by solving subproblem (17); the regional controller calculates the power balance after receiving all bids and updates the price vector; these steps are repeated until a balance is reached.

[0013] The aggregate of the present invention adopts two-layer intelligent management and control: the bottom layer adopts model predictive control for bottom-level energy management and the upper layer adopts game theory-related theories for economic regulation, so as to realize economical and efficient aggregation and management of multiple controllable devices in the distribution network. The present invention is aimed at the intelligent management and control of multiple controllable devices in the active distribution network, and proposes an intelligent agent software description model of multiple controllable devices based on physical-information fusion and an optimized aggregation management and control system based on two-layer optimized multi-agents. The present invention realizes the interactive behavior characteristics of multiple controllable devices in the active distribution network with effective software information description, realizes the aggregation controllability of multiple controllable devices and the ability to participate in the market behavior of the distribution network, and significantly improves the intelligent management and control effect of multiple controllable devices in the active distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a three-tier management system architecture diagram; Figure 2 It is a diagram of the physical device of a multi-protocol edge gateway; Figure 3 It is the logical equipment and node diagram of the photovoltaic system; Figure 4 It is the model diagram of charging pile logic equipment and logic nodes; Figure 5 It is the energy storage system logical device and logical node diagram; Figure 6 It is a fusion diagram of IEC601850 and ACL; Figure 7 It is a two-tier control framework diagram of the regional power grid; Figure 8 It is a schematic diagram of a typical polymer layer (body) system; Fig. 9 It is a schematic diagram of energy management of a regional power grid containing three aggregates; Fig.10 is the photovoltaic and wind power curve in each aggregate; Fig.11 is the electrical load curve in each aggregate; Fig.12 is the heat load curve in each aggregate; Fig.13 is the transformer power diagram in two operating modes; Fig.14 is the input and output power diagram of the aggregate; Fig.15 is the electrical power result graph for each aggregate; Fig.16 Resulting plot of thermal power for each aggregate. DETAILED DESCRIPTION

[0015] In order to realize the intelligent management and control of massive distributed devices in the distribution network and realize the concept of intelligent aggregate, it is necessary to design a corresponding software agent description model. On this basis, a multi-device intelligent management and control algorithm considering uncertainty is designed to achieve optimal energy management within the aggregate and maximize profits in the power market.

[0016] The steps of the present invention are: Step 1. Design a local power grid management framework system for intelligent aggregation of multiple controllable devices in active distribution networks. The system is divided into a distributed controllable device intelligent agent layer, a multi-agent aggregation layer (agent) and a multi-aggregation layer (agent) composed of a regional layer (grid) based on the geographical area and management level of the equipment.

[0017] Step 2: Design information description based on IEC61850: Drawing on the information protocol architecture of the international standard IEC6185ED.20 in the field of power system automation and combining the characteristics of multiple controllable devices in active distribution networks, a standardized IEC61850 XML extensible markup language XML model file is established.

[0018] Step 3: Design an agent language description model based on the IEC61850 information description model and MAS technology: The IEC61850 information description and the common language ACL of intelligent agents are integrated to describe the physical information mapping model, and to describe intelligent behaviors such as information sharing, dynamic characteristics and interactive control of intelligent agents.

[0019] Step 4: Based on the agent information model, design the underlying optimal energy management of the multi-agent intelligent aggregation system: The dual-layer optimization aggregation technology is adopted. The bottom layer adopts the optimal energy management based on predictive control to optimize the energy management of the multi-agent aggregation system.

[0020] Step 5: Based on the agent information model, design the maximum benefit regulation of the multi-agent cooperative game in the upper layer of the multi-agent intelligent aggregation system: A cooperative game algorithm is used to achieve the high efficiency of power resources in the active distribution network in the multi-aggregate optimal configuration and the fairness of the aggregate system in the distribution of power market benefits.

[0021] The main challenges and difficulties in the intelligent aggregation management of distributed multi-devices in the existing active distribution network are: 1. The underlying communication methods and protocols of multiple devices are inconsistent, and the power behavior characteristics of multiple controllable devices are inconsistent. Therefore, how to implement the physical-information intelligent body software description problem of multiple adjustable devices with different communication protocols and communication methods, especially the description problem of intelligent interactive behavior.

[0022] 2. The problem of intelligent aggregation of multiple devices. Single multiple controllable devices have low power and large number. Multiple intelligent entities of multiple devices need to be aggregated into a whole to be dispatchable. The intelligent aggregate of multiple devices needs to ensure the optimal energy management within the aggregate, especially the energy management under uncertainty. At the same time, the aggregate also needs to participate in the benefit distribution problem of the power market formed by the alliance of other intelligent entities in the active distribution network in an overall manner.

[0023] The present invention is further described in detail below in conjunction with the accompanying drawings: The process of the present invention: Step 1: Intelligent aggregation management architecture system of multiple controllable devices in active distribution network: The present invention designs a three-layer management framework system with an intelligent agent of local multi-device physical-information mapping, an aggregation layer (body) formed by the aggregation of multiple multi-device intelligent agents, and a local layer where multiple aggregates participate in active power distribution response. Figure 1 As shown. Multiple controllable physical devices are accessed through multi-protocol intelligent gateway devices through their own measurement information and characteristics, and form device-layer intelligent entities based on the IEC61850 communication protocol standard and the intelligent entity common language. Multiple intelligent entities are divided into aggregation layers by geographical regions. The aggregation layer realizes sub-area network energy management through each controllable device in the active multi-aggregate sub-area network. The multi-aggregate local layer uses the aggregation layer (body) as a single participating unit to optimize the energy aggregation and market behavior of multiple aggregates, ensure the safe operation of the distribution network and the distribution of benefits of each aggregate. The three-layer structure makes the intelligent management of multiple devices in the distribution network clear and the functions more clear, and improves the management and control capabilities of massive multiple controllable devices in the active distribution network.

[0024] The multi-protocol edge gateway is the physical device basis of the present invention, see Figure 2 As shown. The multi-protocol edge gateway is a communication gateway with edge computing capabilities independently developed by the multi-protocol edge gateway. It has multi-communication access, multi-protocol decoding and edge computing capabilities. It is the hardware carrier of the aggregation management system of the present invention.

[0025] The hardware composition of the smart gateway: the gateway is equipped with an eight-core 64-bit CPU, 4GB RAM, 35GB Flash storage and a 2TB mechanical hard disk, supports WiFi and 5G full Netcom connections, and has Bluetooth functions and dual wired network cards. In addition, it also has a built-in gigabit network switch and rich interface resources, including multiple USB, RJ-45 network ports, HDMI, WiFi and 5G antennas, as well as RS-485, RS-232 serial ports, etc. In terms of software configuration, the gateway runs the Ubuntu 16.04LTS operating system, adopts the aarch64 system architecture, the kernel version is 5.10.110, and supports the GCC 5.4.0 compiler. The configuration software version is Web 3.0, which can meet the needs of modern network applications. In terms of protocol support, the gateway supports a variety of transport layer and application layer protocols, including TCP, UDP, ModBus, IEC101, IEC103, IEC104 and MQTT, etc., ensuring compatibility with various devices and reliability of data transmission. The transport layer protocols currently supported by the device are: ALiYun_MQTT_M2M_V2_03, KX-TCPClient2CAN, MQTT, TCPClient, TCpClient_Backup, TCPclient DTu Mode, TCPclient HongDianDTU, TCPClient v2, TCpClient test, TCPServer, TCPServer_V2, UDP, WK MQTT M2M, ZLG-TCPClient2CAN. The application layer collection and forwarding protocols are: ABB, SPA_BUSACT, AIBUS, AISE_LORAWANTOMQTT, etc.

[0026] The software functions of the smart gateway: the gateway can collect and forward data, store data, provide alarm services, control logic processing, generate reports, and provide user interface design. It supports multiple database storage, such as MySQL, SQLServer and PostgreSQL, and can handle long-term storage of massive data. The alarm service can monitor the collected variables in real time, alarm according to preset conditions, and store the alarm information in the database. The control logic editing function allows users to start, cycle, time and control data changes according to different needs. The report function supports customized daily reports, monthly shifts and annual reports, while the UI interface design provides a flexible graphics system and background language tools. The graphics page uses JavaScript language; in addition to the classes and methods given by JavaScript language, BOM, and DOM, many third-party plug-ins are also introduced: jQuery, bootstrap, lobibox, highcharts, bootstrap-datetimepicker, socket.io, zTree, moment, amap. In addition, there are some custom functions for use: getAttr to obtain control attributes, setAttr to set control attributes, getPointValue to obtain real-time data, getDbInfo to obtain historical data, etc. Through the graphic system and background language tools, users can flexibly design UI interfaces to meet visual requirements such as curve drawing, light-emitting sign alarm, and loop animation, and can also realize more logical functions that cannot be realized by controls. In addition to the above functions, the device also supports common functions such as WeChat service, EZVIZ service, OpenVPN, and Peanut Shell; based on sufficient performance and a complete system, the device supports hard encryption functions, such as encrypted TF card, encrypted SIM card, etc., supports tun virtual network card, and can establish a secure virtual channel. In addition, the device supports the establishment of multiple users with different permissions, and supports multiple users online at the same time, and supports unified user management; it can perform project backup and restore, support project file encryption; support remote and online upgrade and maintenance; it works reliably, automatically starts when powered on, and automatically tries to reconnect when disconnected from the network.

[0027] Step 2: Standardized information description of multiple controllable devices based on IEC61850: The IEC61850ED.20 information protocol architecture is used to access different distributed multi-devices with a unified information model. Various distributed multi-controllable devices, such as distributed photovoltaics, distributed energy storage, charging piles, controllable loads, etc., have their own measurement, collection and information functions, but the communication methods and communication protocols between different devices are inconsistent. Each multi-device communicates with a multi-protocol edge gateway. The multi-protocol edge gateway performs protocol conversion to obtain the collection information of different multi-devices. Furthermore, in the present invention, based on IEC61850-9-2, the multi-controllable devices of different protocols and the obtained multi-device information are subjected to protocol standardization processing.

[0028] Distributed multi-controllable devices refer to IEC61850Ed2.0 std7-3 and 7-4 protocols, and model the information of multi-controllable devices according to the process of physical devices-logical devices (Logical Devices, LD)-logical nodes (Logical Nodes, LN)-data objects (Data Objects, DO)-common data classes (Common Data Classes, CDC)-common attributes (Common Attributes)-standard data types (Standard Data Types). This paper logically models typical renewable energy based on the logical nodes newly defined in the IEC61850 standard and constructs a unified information model to solve the interoperability problem between devices from different manufacturers in the intelligent body.

[0029] The IEC61850 information standardization model of the distributed photovoltaic intelligent body in the present invention is described as follows. The photovoltaic power generation system is first modeled using the IEC61850 standard to establish an information model for photovoltaic grid-connected communication. Its logical devices and nodes are as follows: Figure 3As shown in the figure, the method uses the logical nodes and logical devices defined in the IEC 61850-7-420 standard to model the key components in the photovoltaic system. Logical devices (LD) include photovoltaic modules DPVM, photovoltaic array controllers DPVC, and photovoltaic array characteristics DPVA tracking controllers DTRC. Taking the photovoltaic array controller DPVC as an example, the logical devices are further divided into logical nodes, which are logical units with specific functions. DPVC is a logical node that contains status information, setting information, and control information. The logical node consists of multiple data objects, which store the status and parameters of the node. The data object CtrlModSt of status information, the data object TrkDlWupTms of setting information, and the data object ArrModCtr of control information are data objects in the DPVC logical node, representing the array control mode status, photovoltaic startup delay, and array output power control mode, respectively. The structure and content of the data objects are defined by the common data class CDC. The CtrlModSt and TrkDlWupTms data objects use the INS common data class, while ArrModCtr uses the ENC common data class. These CDCs define the attributes and data types of data objects. CDC ensures the semantic consistency of data objects between different systems and devices. In IEC61850, common attributes define the metadata of data objects, such as CtrlModSt and TrkDlWupTms as INS type data objects, which may share common attributes. Standard data types are specific types of data values ​​in data objects, such as CtrlModSt contains an integer value to indicate the state of the control mode, and TrkDlWupTms contains an integer to indicate the number of seconds of the startup delay. ArrModCtr uses the ENC (EnumeratedControl) common data class, which means it contains an enumerated value to represent different control modes. Standard data types ensure data consistency and interchangeability. This approach not only improves the operating efficiency of photovoltaic power generation systems, but also enhances their adaptability and response speed to grid changes, providing strong technical support for the further development of smart grids.

[0030] The information model of the electric vehicle charging pile based on the IEC61850 standard in the present invention is described as follows. The method adopts the IEC61850 modeling technology and the UML analysis method, analyzes the monitoring function and information volume of the AC charging pile, and establishes the information model of the logical device, logical node and data of the AC charging pile in the bay layer communication management machine. Figure 4As shown, the charging pile monitoring function is logically divided into three LDs, which respectively complete the measurement, control and alarm functions (LD1), metering function (LD2), and protection function (LD3). The logical device is further divided into logical nodes, among which the logical nodes of the logical device LD1 are: LPHD (logical physical device), LLNO (logical node 0), GGIO (general input and output logical node), and MMXU (measurement data logical node). The logical nodes of the logical device LD2 are: LPHD (logical physical device), LLNO (logical node 0), and MMTR (metering data logical node). The logical nodes of the logical device LD3 are: LPHD (logical physical device), LLNO (logical node 0), PTOC (overcurrent protection logical node), and PTOV (overvoltage protection logical node). Taking the logical node MMXU as an example, it is responsible for measuring data. The data objects of MMXU include PhV (voltage) and A (current). The attribute types of PhV and A are relatively related measurement values ​​WYE, and WYE includes phsA, phsB, and phsC with attribute types of CMV. It is further determined that the data attributes contained in CMV include the measured value cVal, the quality attribute q and the timestamp t. The logical node GGIO is used for status monitoring, alarm and control. The data objects of GGIO include IntIn1 (status input), Alm1 (alarm) and DPCS01 (control output). The attribute type is INS, and its data attribute stVal specifically represents various corresponding status values. The attribute type of the alarm data Alm is SPS, which is represented by the data attribute stVal. The attribute type of the control data DPCSO is DPC, which supports conventional safety pre-operation selection control or conventional safety direct control. The control operation is completed through the coordination of the control mode ctlModel, SBO class sboclass, and timeout time sboTimeOut data attributes. The metering MMTR contains the data net active energy TotWh, the data class is BCR (energy metering), and the data attribute actVal of TotWh describes the output electric energy of the charging pile. The data of the protection function logical nodes PTOV and PTOC include start Str, action Op, start value StrVal and protection action time RltTmms Relative Time. The attribute type of Str is ACD, and the data attributes include total action and three-phase protection settings. The attribute type of Op is ACT, and the data attributes include total action and three-phase protection action. StrVal is the action start value. RltTmms describes the action delay time. These logical devices are standardized through common data classes to ensure consistency and interoperability throughout the system. This improves the interoperability and operation management efficiency of the charging station system.

[0031] The present invention provides a method for constructing an information model of a battery energy storage system based on the IEC61850 standard as follows. According to the IEC61850 standard, an object-oriented unified modeling technology is adopted. Logical devices and logical nodes are as follows Figure 5As shown in the figure, "battery system" and "battery monomer" are physical devices, which are abstracted as logical devices (LD) in the communication system, "ZBAT" is battery, "ZBTC" is battery charge controller, "ZRCT" is DC converter, "ZINV" represents inverter, etc. Logical devices are further decomposed into logical nodes, which represent the specific functions or characteristics of the devices. In the figure, "DRCT" is distributed energy controller, "DRCS" is distributed energy state, and "DRCC" is distributed energy control action. Logical nodes are the specific implementation of device functions, which contain data objects (DO). ZBAT's data objects include battery state, discharge curve, protection setting, operation value measurement, etc., among which CDC uses ING (integer value) to represent the number of batteries and ASG (analog value) to represent battery voltage or current. Public attributes include M (mandatory) and O (optional); ZBTC data objects include charging state, charging current, charging voltage, and use ASG (analog value) to represent charging current or voltage. In addition, the construction of the battery energy storage system information model also uses the measurement data logical node (MMXU) and the general input and output logical node (GGIO). MMXU contains data Phv, A, TotW, TotVar, HZ, TotPF and battery measurement information (vol1, vol2, tmp1, etc.). Phv describes the three-phase voltage of PCS, A describes the three-phase current of PCS, TotW describes the three-phase total active power, TotVar describes the three-phase total reactive power, HZ describes the system frequency, and TotPF describes the power factor. The attribute type of PhV and A is the relatively related measurement value WYE. WYE includes phsA, phsB, and phsC with the attribute type of CMV, and further determines that the data attributes contained in CMV include the measurement value cVal, the quality attribute q, and the timestamp t. The attribute type of TotW, TotVar, HZ, and TotPF is MV. The GGIO logical node data mainly includes control, alarm signals, and device operation status data. In GGIO, the extended definitions of Ind1, Ind2…, Ind5 describe the power-on / off status of the energy storage device, the equipment fault status, the local remote operation mode, the off-grid mode, and the communication status information. The attribute is INS, and its data attribute stVal represents the specific status value. The attribute type of the alarm data Alm1…Alm25 is SPS, which describes the battery information alarm and device fault alarm information, and is specifically represented by the data attribute stVal. The attribute type of the control data DPCS01 is SPC, and the control operation is completed through the data attributes ctlVal, ctlNum, and check. The standardized management of the energy storage monitoring system is realized, which effectively improves the safety management level of the energy storage system and the efficient operation of the power grid.

[0032] Step 3: Agent software description of multi-controllable devices based on FIPA-ACL: Distributed adjustable resource devices are mapped to a unified information model through the IEC61850 protocol. The information model based on IEC61850 is converted into the agent universal language ACL; a communication protocol between intelligent agents is established to ensure that they can communicate effectively using ACL, thereby realizing intelligent agent interaction.

[0033] An intelligent agent is a packaged software agent system that receives input related to the state of the environment through sensors and acts on the environment through effectors in a specific environment. The present invention converts the distributed adjustable resource information model constructed under the IEC 61850 standard into an agent common language (ACL).

[0034] ACL (Agent Communication Language) is a language used for communication between agents in a multi-agent system (MAS). ACL provides a standardized framework that enables different agents to exchange information and coordinate actions in a way that is commonly understood and executed. There are two types of ACLs that are currently used and discussed more: KQML and FIPA-ACL. The present invention adopts the universal language FIPA, and FIPA-ACL is currently the most widely used agent communication language. FIPA is a non-profit international organization dedicated to formulating standards that support interoperability between agents and agent-based applications to promote the use of industrial intelligent agents. FIPA's information transmission model ACC, ACC (Agent Communication Channel) is an entity that directly provides services for agents on the agent platform AP (Agent Platform), and can access information provided by other AP services to complete message transmission tasks. The standard MTP (Message Transport Service) interface of ACC is used to provide mutual message transmission between APs that comply with FIPA. To comply with FIPA, ACC must at least provide an interface that supports FIPA MTP. When the agent constructs information, each message consists of a message envelope and a message body, providing the destination address of the message and specific message transmission information; the ACC parses the message and determines the recipient of the message based on the message envelope. FIPA defines a message transmission service. If the recipient of the message is on the same platform, the ACC directly transmits the message to the recipient. Otherwise, the ACC decides to transmit the message to the ACC of the platform where the recipient is located or to the ACC of another platform based on the recipient's address, and the message is transmitted by other ACCs. During the message transmission process, the ACC is responsible for the encryption and decryption of the message to ensure the security of communication. In addition, the ACC implements a message confirmation and retransmission mechanism to improve the reliability of message transmission. Through this mechanism, the agent system can achieve efficient and reliable message transmission and support collaboration and task execution between agents.

[0035] IEC601850 and ACL fusion: An ACL description model usually includes senders, recipient lists, communication behaviors, content, content language, ontology, etc. The ontology elements include concepts, predicates, and agent actions. Figure 6As shown in the figure, the concept element is mapped with the fixed value information and status information of the LN of IEC 61850, and is used to nest to form the same tree structure as the LN; the assertion element is used to distinguish between the command issued by the superior and the information sent back by the subordinate; in the agent system, when the agent needs to send or receive information, it will use this mapping to construct or parse the ACL message, and the agent action element is mapped with the control information of the LN, indicating a series of actions performed by the agent. For example, if an agent needs to report its current power generation status, it will map the status information to the corresponding concept and use assertions to express this information. Agents use these mappings to communicate and coordinate actions. For example, an agent may send an ACL message containing control information to another agent, and the receiving agent will understand the message according to the mapping and perform the corresponding action.

[0036] Step 4: Aggregation layer multi-agent two-layer control system, lower layer daily optimal operating cost management: The present invention designs a double-layer intelligent optimization control system for local power grid based on the integration of intelligent information model and decentralized architecture system. Figure 7 As shown in the figure. The aggregation layer management adopts a predictive control algorithm based on model-data hybrid drive to achieve energy and cost management within each aggregate, that is, the underlying energy management. The local layer adopts the day-ahead optimal scheduling and solves the problems of safe operation and multi-aggregate stakeholder allocation in the distribution network based on the optimization algorithm.

[0037] Typical multi-resource energy aggregation layer models are as follows: Figure 8 As shown, it includes renewable energy (RES) power generation, gas turbine combined heat and power (CHP) devices, natural gas furnace (GF), electric boiler (EB), battery energy storage system (BESS) and thermal energy storage (TES). The mathematical model of the aggregation layer can be expressed as follows: in represents the electricity imported from the grid by the aggregator at time t, Indicates the RES power in maximum power point tracking mode; respectively represents the natural gas consumed by CHP and GF; and are the gas-to-electricity efficiency and gas-to-heat efficiency of CHP; Indicates the amount of electricity consumed by EB; and are the efficiencies of G and EB, respectively; and Represent the charging power and discharging power of BESS respectively; and are the charge and discharge amounts of TES respectively; and Indicates the value of electrical and thermal energy losses; and Represents movable electrical and thermal loads; L E,t and L H,t Represent fixed electrical load and thermal load respectively.

[0038] The objective function and constraints of energy management at the aggregation layer are described as follows: Each aggregator will work to minimize the operating costs within the aggregator, during this period, the operating costs can be divided into two parts: the cost of purchased electricity and the cost of purchased natural gas where μ e,t and μ g,t They respectively represent the electricity price and natural gas price for time period t given by the superior.

[0039] In addition to the power balance constraint (1), other constraints include: in and Indicates the maximum exchange power through the connecting line; and represents the installed capacity of the combined heat and power pump, GF and EB respectively; P CHP , P GF and P EB Indicates the lower limit of the corresponding power; and Indicates the maximum charge and discharge power of BESS; and Represent the maximum values ​​of TES charging and discharging, respectively.

[0040] Model predictive control can solve the uncertainty problems caused by model mismatch and interference, and help improve the actual control performance of the system. Therefore, it is widely used in multi-time scale optimization scheduling to alleviate the problems caused by prediction errors. This paper uses MPC to correct the deviation between the actual daily operation plan and the previous day's scheduling plan.

[0041] The state variables in the model are: x(k)=[P e P RES P CHP P EB P BESS H GF H EB H TES ] (9)

[0042] The control variables in the model are: u(k)=[ΔPe ΔP RES ΔP CHP ΔP EB ΔP BESS ΔH GF ΔH EB ΔH TES ] (10)

[0043] The disturbance variable in the model is: r(k)=[ΔP Load ΔH Load ΔP RES ] (11)

[0044] In order to minimize the operating costs within the aggregator, the following objective function is established:

[0045] The power control quantity of the aggregation layer is obtained after calculating the target (9) and sent to the upper regional controller. The regional controller will calculate the power balance of the distribution network system and correct the market price.

[0046] Step 5: Regional-level multi-aggregate optimal dispatch: The day-ahead optimal dispatch phase for participating in the demand-side power market uses the predicted renewable energy generation and load data for the current day and the predicted grid electricity price to iteratively solve the inter-week optimization problem, aiming to determine the local operating electricity price for the current day.

[0047] In each cycle, the goal of the zone controller is to minimize the total cost of the system in the remaining cycles while keeping the supply and demand balanced and satisfying the operation constraints. The optimization problem can be modeled as follows: Where n is the index of the aggregator; N is the total number of aggregators; is the input power of the main transformer; is the maximum switching power.

[0048] The relaxed Lagrangian dual problem based on (13) is:

[0049] At a given λ t Under this condition, the dual problem (15) can be decomposed into N+1 sub-problems, corresponding to N aggregators,

[0050] The transformer

[0051] If the Lagrange multiplier λ tIn Eqs. (16) and (17), it is interpreted as the price, the local price signal λ e,t It can be defined as, λ e,t =μ e,t +λ t . (18)

[0052] The problem in equation (13) can be solved in a two-layer framework. That is, in the upper-layer region, the controller adjusts the local price vector to achieve an overall balance between supply and demand. In the lower-layer region, each aggregator autonomously minimizes its cost under the price vector. In each iteration, each aggregator solves subproblem (16) with the provided price vector and submits the optimal power vector as a bid. At the same time, the transformer obtains the optimal power vector by solving subproblem (17). The regional controller calculates the power balance after receiving all bids and updates the price vector. These steps are repeated until a balance is reached.

[0053] Experimental simulation evaluates the designed local power grid two-layer management system. Reference system Fig. 9 , and conduct example verification. Fig. 9 In the paper, the regional power grid contains a regional power grid system model with three aggregation layers (bodies), each of which is composed of distributed photovoltaics, energy storage, electric vehicle charging piles, and gas turbines. Based on the scenario generation method and scenario reduction method, typical curves of photovoltaic power generation, wind power, electric load, and thermal load are obtained in each aggregation layer (body). Fig.10 is the renewable energy output curve of each aggregation layer (body), and Fig.11 and Fig.12 The electrical load curve and thermal load curve are described separately.

[0054] For comparison, this example will consider a scenario where the collaborative optimization method is not used. In this framework, the aggregation layer (body) does not consider the constraints of the dual-layer intelligent controller scheduling. The aggregation layer (body) acts autonomously and only considers the maximization of its own interests. The transformer power simulation results under the two modes are shown in Figure 2. Fig.13 As shown. In the non-cooperative mode, the main transformer was overloaded at 2:00-5:00 and 15:00. After the regional controller coordinated the power output of each aggregation layer (body) in the collaborative autonomous mode, the main transformer avoided overloading. At the same time, in order to increase the local consumption rate of local renewable energy and prevent the return of power from the distribution network, the collaborative optimization method sets the feed-in tariff very low, aiming to prevent the aggregator from over-generating.

[0055] Fig.14 The output / input power and real-time settlement price of each aggregation layer (body) under the two control modes are shown. Fig.13As shown in Figure 1, the collaborative autonomous model raises the real-time settlement price above the grid price during 2:00-6:00 to ease import congestion. As a result, aggregation layers 2 and 3 are discouraged from consuming power, while aggregation layer 1 increases its power output and makes a profit by selling power to other aggregators. By coordinating the management of multiple aggregation layers, the main transformer is effectively prevented from being overloaded.

[0056] The electrical power and thermal power results of all polymer layers (body) are as follows Fig.15 and Fig.16 As shown, the state of charge (SOC) of the stored energy is defined as the ratio of the stored energy to the total capacity. Fig.15 and Fig.16 In the bar graph, power generated to meet demand is positive, and power exported or consumed is negative. Fig.16 It is shown that both Aggregator 1 and Aggregator 3 determine their CHP output based on the heat load throughout the day. Therefore, there is excess renewable energy generation, and these aggregators cannot consume it internally and must sell electricity to other aggregators during this period. In contrast, Aggregator 2 lacks a CHP unit and only relies on boilers for heat supply and purchases electricity from the main grid to meet its electricity demand. Aggregator 2 has a relatively high electricity demand throughout the day.

[0057] During off-peak hours, the aggregate layer usually buys cheap electricity from the main grid and uses boilers to provide heat. In contrast, during peak hours, it is more advantageous to use CHP to generate electricity and provide heat. The aggregate layer tends to charge the BESS during off-peak hours to store electricity during peak hours, while the CHP unit is equipped with a thermal energy reservoir and tends to store heat during peak hours and release heat during off-peak hours. The coordination of BESS, TES and movable loads matches the net power demand of the aggregate layer for heat loads with the combined heat and power ratio of the CHP unit. On the one hand, it minimizes the use of boilers and reduces the heat cost during peak hours. On the other hand, in this case, both the discard of power and heat energy are minimized, improving the overall efficiency of the aggregate layer and the local power grid.

Claims

1. A distributed multi-device intelligent agent aggregation double-layer management and control system in a regional distribution network, characterized by: The steps are: S1. Design a local power grid management framework system for intelligent aggregation of multiple controllable devices in active distribution networks: Based on the geographical area and management level of the equipment, it is divided into a distributed controllable equipment agent layer, a multi-agent aggregation layer, and a multi-aggregation layer to form a regional layer; S2. Design based on IEC61850 information description: Based on the information protocol framework of the international standard IEC6185ED.20 in the field of power system automation and the characteristics of multiple controllable devices in active distribution networks, a standardized IEC61850 XML extensible markup language XML model file is established. S3. Design an agent language description model based on the IEC61850 information description model and MAS technology: It integrates IEC61850 information description and agent common language ACL to describe the physical information mapping model, and describes intelligent behaviors such as agent information sharing, dynamic characteristics and interactive control; S4. Design the underlying optimal energy management of the multi-agent intelligent aggregation system based on the agent information model: It uses a two-layer optimization aggregation technology, with the bottom layer using optimal energy management based on predictive control; S5. Based on the agent information model, design the maximum benefit regulation of the multi-agent cooperative game in the upper layer of the multi-agent intelligent aggregation system: A cooperative game algorithm is used to achieve the high efficiency of power resources in the active distribution network in the multi-aggregate optimal configuration and the fairness of the aggregate system in the distribution of power market benefits.

2. The distributed multi-device intelligent agent aggregation double-layer management and control system in the regional distribution network according to claim 1 is characterized by: Multi-agent intelligent aggregation system, that is, the aggregation layer multi-agent double-layer control system: the lower layer intraday optimal operation cost management; multi-resource energy aggregation layer model: in represents the electricity imported from the grid by the aggregator at time t, Indicates the RES power in maximum power point tracking mode; respectively represents the natural gas consumed by CHP and GF; and are the gas-to-electricity efficiency and gas-to-heat efficiency of CHP; Indicates the amount of electricity consumed by EB; and are the efficiencies of G and EB, respectively; and Represent the charging power and discharging power of BESS respectively; and are the charge and discharge amounts of TES respectively; and Indicates the value of electrical and thermal energy losses; and Represents movable electrical and thermal loads; L E,t and L H,t represent fixed electrical load and thermal load respectively; Aggregation layer energy management objective function and constraints: Operating costs are divided into two parts: electricity costs and natural gas costs where μ e,t and μ g,t They represent the electricity price and natural gas price for time period t given by the superior respectively; Other constraints include: in and Indicates the maximum exchange power through the connecting line; and Respectively represent the installed capacity of the combined heat and power pump, GF and EB; P CHP , P GF and P EB Indicates the lower limit of the corresponding power; and Indicates the maximum charge and discharge power of BESS; and Respectively represent the maximum value of TES charging and discharging; Use MPC to correct the deviation between the actual daily operation plan and the previous day's scheduling plan: The state variables in the model are: x(k)=[P e P RES P CHP P EB P BESS H GF H EB H TES ] (9) The control variables in the model are: u(k)=[ΔP e ΔP RES ΔP CHP ΔP EB ΔP BESS ΔH GF ΔH EB ΔH TES ] (10) The disturbance variable in the model is: r(k)=[ΔP Load DI Load ΔP RES ] (11) Create the following objective function: The power control quantity of the aggregation layer is obtained after calculating the target (9) and sent to the upper regional controller. The regional controller will calculate the power balance of the distribution network system and correct the market price.

3. The distributed multi-device intelligent agent aggregation double-layer management and control system in the regional distribution network according to claim 1 is characterized by: Optimal dispatch of multi-aggregates at the regional level: optimal dispatch of participating demand-side power markets on the day-ahead Optimization problem modeling: Where n is the index of the aggregator; N is the total number of aggregators; is the input power of the main transformer; is the maximum exchange power; The relaxed Lagrangian dual problem based on equation (13) is: Given λ t Formula (15) is decomposed into N+1 sub-problems, corresponding to N aggregators: Transformer for If the Lagrange multiplier λ t In Eqs. (16) and (17), it is interpreted as the price, the local price signal λ e,t Defined as: l e,t =μ e,t +λ t . (18) The optimization problem of (13) is solved in a two-layer framework, i.e., in the upper region, the controller adjusts the local price vector to achieve an overall balance between supply and demand, and in the lower region, each aggregator autonomously minimizes its cost under the price vector; In each iteration, each aggregator solves subproblem (16) with the provided price vector and submits the optimal power vector as a bid. At the same time, the transformer obtains the optimal power vector by solving subproblem (17). After receiving all bids, the regional controller calculates the power balance and updates the price vector. Repeat these steps until equilibrium is reached.

Images