Protocol conversion and communication adaptation system for optical storage and charging grid-connected device

Abstract

The invention relates to the technical field of power system communication, in particular to a protocol conversion and communication adaptation system for an optical storage and charging grid-connected device. Comprising a protocol conversion unit used for dynamically analyzing heterogeneous protocols of a photovoltaic inverter, an energy storage converter, a charging pile and power grid side equipment; a communication adaptation unit; an energy collaboration unit; a power grid interface unit; and a man-machine interaction unit. According to the method, the protocol mapping rule base is dynamically updated through the plug-in type architecture module and the machine learning algorithm, self-adaptive analysis of multiple heterogeneous protocols and variant protocols is achieved, the protocol analysis accuracy and generalization ability are improved, and the problem that a traditional static rule base is insufficient in adaptability to non-standard protocols is solved; according to the invention, the distributed edge computing architecture is adopted to construct the bidirectional communication link, so that the real-time performance and security of the communication link are improved, and the defects of high data delay and incomplete security mechanism in the traditional communication architecture are improved.

Description

Technical Field

[0001] The present invention relates to the field of power system communication technology, and in particular to a protocol conversion and communication adaptation system for a photovoltaic storage and charging grid-connected device. Background Art

[0002] With the large-scale grid integration of photovoltaic (PV), energy storage, and charging facilities, the protocol heterogeneity of PV inverters, energy storage converters, charging stations, and other devices in PV-storage-charging systems has become increasingly prominent. Industrial protocols (such as Modbus and IEC60870-5-103) used by devices from different manufacturers have different syntax, data formats, and communication rules, resulting in poor data exchange between devices. Efficient protocol conversion and communication adaptation solutions are urgently needed to achieve coordinated system operation. In current PV-storage-charging grid-connected scenarios, multi-device protocol parsing relies on static mapping rules, making it difficult to adapt to protocol version iterations and dynamic changes in non-standard protocols. Furthermore, traditional centralized communication architectures face bottlenecks in real-time data collection and reliable command interaction, making them unable to meet the grid's requirements for efficient energy dispatch.

[0003] Existing technologies have the following technical deficiencies in protocol conversion and communication adaptation: First, traditional solutions rely on a static rule base to parse protocols. When faced with non-standard protocols such as custom check bits and field order variations, their lack of intelligent learning capabilities leads to insufficient parsing adaptability and makes it difficult to cope with iterative protocol updates. Second, the communication architecture has limited data processing efficiency, and there is no standardized interaction process between edge devices and the central control module. Real-time control data is prone to delays or packet loss during transmission. Furthermore, the security encryption mechanism of the communication link is imperfect, which cannot meet the high security requirements of power grid dispatch. Third, the lack of a joint optimization mechanism for photovoltaic output, energy storage status, and charging load in energy coordinated regulation makes it difficult to achieve a dynamic balance between grid command response and energy utilization efficiency. In view of this, we propose a protocol conversion and communication adaptation system for grid-connected photovoltaic storage and charging devices. Summary of the Invention

[0004] The purpose of the present invention is to provide a protocol conversion and communication adaptation system for a photovoltaic storage and charging grid-connected device, so as to solve the problems that the traditional solutions proposed in the above-mentioned background technology lack intelligent learning capabilities, resulting in insufficient analytical adaptability, difficulty in coping with protocol iteration updates, limited data processing efficiency of the communication architecture and imperfect security encryption mechanism of the communication link, and lack of a joint optimization mechanism for photovoltaic output, energy storage status, and charging load in energy coordinated regulation, making it difficult to achieve a dynamic balance between grid command response and energy utilization efficiency.

[0005] To solve the above technical problems, the present invention provides a protocol conversion and communication adaptation system for a photovoltaic storage and charging grid-connected device, comprising:

[0006] The protocol conversion unit is used to dynamically analyze the heterogeneous protocols of photovoltaic inverters, energy storage converters, charging piles, and grid-side equipment. It realizes plug-and-play of multiple protocols through plug-in architecture modules and automatically updates the protocol mapping rule base based on machine learning algorithms.

[0007] The communication adapter unit uses a distributed edge computing architecture to achieve real-time interaction of multi-device data. The communication adapter unit includes an edge computing node module and a central control module. The edge computing node module and the central control module establish a bidirectional connection through an industrial-grade communication link to achieve real-time multi-device data collection, protocol conversion, command interaction and closed-loop control.

[0008] The energy coordination unit is used to perform multi-objective optimization of photovoltaic output, energy storage SOC, charging load, and grid instructions. It integrates the LSTM time series prediction model and mixed integer programming algorithm, controls the energy storage charging and discharging power through the CAN bus, and uses virtual meter technology to dynamically adjust the charging power.

[0009] The grid interface unit implements secure communication with the dispatching master station based on the national encryption SM2 and SM4 algorithms, and establishes an end-to-end encrypted channel through the vertical encryption module;

[0010] The human-computer interaction unit is used to provide visual interaction and remote control functions, use digital twin technology to build a three-dimensional device status model, and use blockchain technology to realize the trusted storage of charging data.

[0011] As a further improvement of this technical solution, the plug-in architecture module includes a protocol parsing submodule, a data mapping submodule and a plug-in management submodule, wherein:

[0012] The protocol parsing submodule is internally integrated with a state machine parsing engine, which supports syntax analysis of binary coding protocols and function code protocols in the field of industrial automation;

[0013] The data mapping submodule builds a protocol data model mapping library based on knowledge graph technology, dynamically mapping device parameters (such as voltage, current, SOC) of different protocols to a unified data format;

[0014] The plug-in management submodule implements plug-and-play of industrial protocol plug-ins through a hot-swap interface, and supports plug-in operation status monitoring and fault self-healing (such as automatic restart of abnormal plug-ins).

[0015] As a further improvement of the present technical solution, the protocol conversion unit automatically updates the protocol mapping rule base based on a machine learning algorithm, comprising the following steps:

[0016] S120.1. Protocol message feature extraction:

[0017] Use Transformer-XL network to input protocol message sequence , encode, and process long sequence context dependencies through the relative position encoding mechanism. The formula is:

[0018] ;

[0019] in, 、 The query weight matrix and key-value weight matrix are used for feature mapping; 、 A learnable position embedding vector that captures position dependencies; is the offset between the current position and the context position, representing the relative position relationship; For location With context position The protocol feature mapping correlation degree;

[0020] Generate the protocol feature vector through the above mechanism , solves the problem of traditional Transformer losing location information for long messages (such as Modbus RTU long frames) and enhances the differentiation of different protocol features.

[0021] S120.2. Dynamic generation of mutation protocol rules:

[0022] A deep reinforcement learning framework is constructed, with the protocol parser as the agent and the device as the environment for interactive training. The mapping rule generation strategy is dynamically adjusted through the reward function, which is defined as:

[0023] ;

[0024] in, for Time parsing accuracy, a measure of parsing correctness; for Time data throughput, a measure of parsing efficiency; for The number of parsing errors at any moment, measuring parsing stability; is the weight coefficient and =1, used to balance different optimization objectives; Immediate rewards for reinforcement learning of protocol parsers;

[0025] Based on this reward function, the intelligent agent generates adaptation mapping rules for variant protocols (such as Modbus ASCII with custom check bits and IEC60870-5-103 with field order variations), improving the parsing ability of non-standard protocols.

[0026] S120.3. Iterative optimization of the global rule base:

[0027] Edge nodes regularly update their local rule base Synchronize to the central control module and use the federated learning algorithm to aggregate multi-site rule data. The aggregation formula is:

[0028] ;

[0029] in, is the total number of edge nodes, representing the network scale; For the The amount of training data for each node reflects the data contribution of the node; is the total amount of global data, used for weighted average calculation; A global protocol rule base aggregated across regions.

[0030] This mechanism can achieve collaborative optimization of the global rule base while protecting the privacy of original data on the device side (without uploading message content), adapt to multi-site heterogeneous protocol environments, and improve the generalization capability of the rule base.

[0031] As a further improvement of this technical solution, the edge computing node module is installed at the local device site, connected to the local device through an industrial-grade communication interface, and supports real-time conversion of multiple heterogeneous industrial protocols and Ethernet protocols; the edge computing node module includes a local protocol processing submodule and an edge communication interface submodule, wherein:

[0032] The local protocol processing submodule is used to store the protocol mapping rules of local devices, realize the real-time collection and format standardization of device data (such as converting device-specific protocol data into a unified system format):

[0033] When receiving the original protocol data from the local device, it is converted into the system unified data format through the hardware acceleration parsing engine;

[0034] When receiving the system unified format control instructions issued by the central control module, it converts them into protocol instructions executable by the local device through the protocol inverse mapping logic;

[0035] The edge communication interface submodule supports two-way data interaction with the central control module, including real-time data upload and control command reception.

[0036] Real-time data upload of the device: Transmits data in the system unified format converted by the local protocol processing submodule;

[0037] Control instruction reception: forwarding the system unified format control instructions issued by the central control module to the local protocol processing submodule;

[0038] The central control module is installed on the system management platform side and establishes a connection with multiple edge computing node modules through an industrial-grade communication link to support cross-regional centralized management of the local device areas covered by each edge computing node module; the central control module includes a global rule management submodule and a data aggregation processing submodule, wherein:

[0039] The global rule management submodule is used to store, update, and distribute protocol mapping rules to cross-region edge computing node modules, and supports rule version control (such as marking rule application scenarios and entry into force conditions) and conflict detection (identifying logical contradictions between rules);

[0040] The data aggregation processing submodule is used to receive device data uploaded by edge computing node modules across regions and perform protocol-independent standardized aggregation (such as integrating data by device type and regional dimensions).

[0041] As a further improvement of this technical solution, the edge computing node module and the central control module perform real-time interaction and control of multi-device data, including the following steps:

[0042] S230.1. The edge computing node module connects to various local devices (energy production / storage / consumption devices) through industrial-grade communication interfaces and collects device operating data (such as power, voltage, and operating status) at preset intervals. It also performs local preprocessing on the raw data, including removing outliers, standardizing parameter units (such as normalizing voltage to V), and adding device identification and timestamp information to generate preliminary, usable device status data.

[0043] S230.2. The edge computing node module converts pre-processed device data from its original protocol format (e.g., a device-specific industrial protocol) into a standardized data format (e.g., JSON) based on a built-in protocol mapping rule base. The converted data is then encapsulated into prioritized communication data frames, distinguishing between real-time control data (high priority) and routine monitoring data (low priority), providing a priority scheduling basis for data transmission.

[0044] S230.3. The edge computing node module uploads the encapsulated standardized data frames to the central control module via an industrial-grade communication link. The central control module then performs global integration and processing on the data uploaded by multiple edge nodes, including storage by region or device type, cross-node time synchronization and calibration (ensuring data timestamp consistency), and generation of comprehensive device operation indicators (such as the total power and average operating efficiency of the local device cluster).

[0045] S230.4. Based on the integrated global data and system control strategy, the central control module generates device control instructions (such as energy storage system charge and discharge power regulation and charging station start and stop control) or rule update instructions. Based on the instruction type and target device characteristics, the module selects unicast or multicast transmission mode (e.g., unicast for control instructions targeting specific devices and multicast for global rule updates). It also adds transmission priority tags to time-sensitive instructions to ensure the real-time transmission of critical instructions.

[0046] S230.5. After receiving instructions from the central control module, the edge computing node module converts the standardized instructions into the original protocol format recognizable by the target device based on the local protocol mapping rule base. The converted instructions are sent to the corresponding local device and the local device execution results (such as instruction reception status, execution completion status, and exception codes) are collected in real time to form instruction execution feedback data frames and temporarily stored locally on the edge node.

[0047] S230.6. The edge computing node module uploads the instruction execution feedback data frame to the central control module; the central control module verifies the effectiveness of the instruction execution based on the feedback results (such as whether the device status changes as expected by the instruction), triggers an adaptive retransmission mechanism for instructions that fail to execute (such as increasing the number of retransmissions and improving the transmission priority), generates alarm information for abnormal situations with multiple failures and records it in the system log, forming a closed-loop management of device control.

[0048] As a further improvement of this technical solution, the energy coordination unit includes a prediction modeling module, a multi-objective optimization module and a coordinated control module, wherein:

[0049] The predictive modeling module integrates a long-short-term memory network time series prediction model to predict photovoltaic power generation, charging load demand, and remaining energy storage capacity in future periods based on historical data of photovoltaic output, charging load, and energy storage SOC;

[0050] The multi-objective optimization module uses photovoltaic absorption, energy storage life, and grid response as optimization objectives. It combines energy storage SOC constraints, charge and discharge power constraints, and grid interaction power constraints to build a multi-objective optimization model and uses a mixed integer programming algorithm to solve it, generating a distribution plan for energy storage charge and discharge power and charging power.

[0051] The collaborative control module issues energy storage charging and discharging control instructions through the CAN bus and uses virtual meter technology to dynamically allocate charging pile power. It collects the controlled data (actual energy storage power, charging pile load) and feeds it back to the multi-objective optimization module to achieve rolling optimization control.

[0052] As a further improvement of the present technical solution, the multi-objective optimization module performs multi-objective optimization including the following steps:

[0053] S320.1. Collect real-time PV output, energy storage SOC, charging load, and grid command data from the communication adapter unit, eliminate outliers using the 3σ criterion, and use linear interpolation to repair missing data.

[0054] S320.2. Call the long short-term memory network time series prediction model of the prediction modeling module to learn the preprocessed historical data sequence and output the predicted values ​​of photovoltaic power generation power, charging load, and remaining energy storage capacity in the future period;

[0055] S320.3. Set the energy storage SOC operating range, charge and discharge power limits, and grid interaction power threshold as constraints, and build a multi-objective optimization model with maximizing PV absorption, extending energy storage life, and improving grid response capabilities as the objective functions.

[0056] S320.4. Use a mixed integer programming algorithm to solve the multi-objective optimization model, generate executable instructions for energy storage charging and discharging power and charging pile power allocation, and mark the instruction priority (e.g., power grid emergency instructions are set to the highest priority);

[0057] S320.5. If the real-time control data fed back by the collaborative control module deviates from the optimization instructions, the parameter update of the multi-objective optimization model is triggered, and steps S320.3-S320.4 are re-executed to achieve periodic rolling optimization.

[0058] As a further improvement of this technical solution, the vertical encryption module includes an encryption tunnel establishment submodule, a two-way identity authentication submodule and a hardware acceleration submodule, wherein:

[0059] The encrypted tunnel establishment submodule builds an end-to-end encrypted channel based on IPsec-VPN technology, supporting the coordinated application of the national encryption SM2 and SM4 algorithms (SM2 for key exchange, SM4 for data encryption). It uses tunnel mode to encapsulate network layer data and encrypts both the header and payload of the IP data packet to ensure the confidentiality of the communication link.

[0060] The two-way identity authentication submodule has a built-in device digital certificate issued based on the SM2 algorithm, which performs two-way identity legitimacy verification with the dispatching master station (the sender's certificate contains the public key, and the receiver verifies the signature through the SM2 algorithm); it also supports certificate validity checks (such as certificate expiration time and CA signature verification) to prevent illegal devices from accessing the communication network;

[0061] The hardware acceleration submodule integrates a dedicated encryption chip to perform hardware acceleration on the SM2 asymmetric encryption and SM4 symmetric encryption algorithms. It supports pipeline processing of national secret algorithms, realizes parallel processing of encryption and decryption operations and data transmission, and improves communication efficiency.

[0062] As a further improvement of this technical solution, the human-computer interaction unit includes a visual interaction module and a data evidence storage module, wherein:

[0063] The visualization interaction module uses digital twin technology to build a three-dimensional device status model, mapping the operating parameters (such as voltage, current, and SOC) of photovoltaic equipment, energy storage devices, and charging piles in real time. It supports dynamic rendering of device status (such as green for normal status and flashing red for fault status) and provides an interactive operation interface (such as clicking on the device model to retrieve real-time data and issue control instructions).

[0064] The data evidence module implements credible evidence storage of charging data based on blockchain technology, and hashes the charging order information (time, power, and cost) and device interaction logs on the chain (e.g., using the SHA-256 algorithm to generate data fingerprints); supports automatic execution of smart contracts (e.g., triggering a fee settlement contract after charging is completed), and provides a blockchain data query interface.

[0065] As a further improvement of this technical solution, the visualization interaction module constructs a three-dimensional device status model including the following steps:

[0066] S510.1. Obtain real-time device operation data from the communication adapter unit, parse it into a unified format (such as JSON) through the protocol conversion interface, remove abnormal data, and add a timestamp;

[0067] S510.2. Build a 3D geometric model of the device based on CAD drawings or point cloud data, and use a digital twin mapping algorithm to bind pre-processed operating data to model parameters (e.g., mapping the energy storage SOC to the color gradient properties of the battery model).

[0068] S510.3. Use WebGL technology to generate a browser-side 3D visualization interface that supports PC and mobile access; set up interactive event response mechanisms (such as displaying detailed device parameters when hovering the mouse, and triggering a remote control pop-up window when double-clicking).

[0069] Furthermore, the data storage module implements charging data storage including the following steps:

[0070] S520.1. Combine key data such as charging start / end time, device ID, and transaction amount, and generate a unique data hash value using the SHA-256 algorithm.

[0071] S520.2. Connect to the blockchain network node through the RPC interface and package the hash value and transaction information into a data block (including the previous block hash, timestamp, and Merkle root);

[0072] S520.3. Call the preset smart contract to verify the legitimacy of the transaction (such as charging station permissions and account balance). After verification, the block data is broadcast to the blockchain network to complete the data on-chain storage.

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] 1. This invention dynamically updates the protocol mapping rule base through a plug-in architecture module and a machine learning algorithm, enabling adaptive parsing of heterogeneous protocols and variant protocols such as Modbus and IEC60870-5-103. This improves the accuracy and generalization of protocol parsing, and addresses the problem of insufficient adaptability of traditional static rule bases to non-standard protocols.

[0075] 2. This invention uses a distributed edge computing architecture to build a bidirectional communication link. It optimizes real-time data transmission efficiency through a priority scheduling mechanism and closed-loop control process. It combines the national secret SM2 algorithm and the national secret SM4 algorithm to establish an end-to-end encrypted channel, improving the real-time performance and security of the communication link, and overcoming the defects of high data latency and imperfect security mechanisms in traditional communication architectures.

[0076] 3. This invention integrates the LSTM time series prediction model with a mixed integer programming algorithm for multi-objective optimization, and combines CAN bus control with virtual meter technology to dynamically allocate energy storage charging and discharging power and charging load, thereby achieving a coordinated balance between photovoltaic consumption, energy storage life, and grid response, and solving the problem of the lack of an energy joint optimization mechanism in traditional solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is a system framework diagram of the present invention;

[0078] The meaning of each number in the figure is:

[0079] 100, protocol conversion unit; 110, plug-in architecture module; 111, protocol parsing submodule; 112, data mapping submodule; 113, plug-in management submodule;

[0080] 200, communication adapter unit; 210, edge computing node module; 211, local protocol processing submodule; 212, edge communication interface submodule; 220, central control module; 221, global rule management submodule; 222, data aggregation processing submodule;

[0081] 300, Energy Synergy Unit; 310, Prediction Modeling Module; 320, Multi-Objective Optimization Module; 330, Synergistic Control Module;

[0082] 400, grid interface unit; 410, vertical encryption module; 411, encryption tunnel establishment submodule; 412, two-way identity authentication submodule; 413, hardware acceleration submodule;

[0083] 500, human-computer interaction unit; 510, visual interaction module; 520, data storage module. DETAILED DESCRIPTION

[0084] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0085] like Figure 1 As shown, this embodiment provides a protocol conversion and communication adaptation system for a photovoltaic storage and charging grid-connected device, including:

[0086] The protocol conversion unit 100 is used to dynamically analyze the heterogeneous protocols of photovoltaic inverters, energy storage converters, charging piles, and grid-side equipment, realize multi-protocol plug-and-play through the plug-in architecture module 110, and automatically update the protocol mapping rule base based on machine learning algorithms;

[0087] In this embodiment, the plug-in architecture module 110 includes a protocol parsing submodule 111, a data mapping submodule 112, and a plug-in management submodule 113, wherein:

[0088] The protocol parsing submodule 111 is internally integrated with a state machine parsing engine, which supports syntax analysis of binary coding protocols and function code protocols in the field of industrial automation;

[0089] The data mapping submodule 112 builds a protocol data model mapping library based on knowledge graph technology, dynamically mapping device parameters (such as voltage, current, SOC) of different protocols to a unified data format;

[0090] The plug-in management submodule 113 implements plug-and-play of industrial protocol plug-ins through a hot-swap interface, and supports plug-in operation status monitoring and fault self-healing (such as automatic restart of abnormal plug-ins).

[0091] In this embodiment, the protocol conversion unit 100 automatically updates the protocol mapping rule base based on a machine learning algorithm, including the following steps:

[0092] S120.1. Protocol message feature extraction:

[0093] Use Transformer-XL network to input protocol message sequence , encode, and process long sequence context dependencies through the relative position encoding mechanism. The formula is:

[0094] ;

[0095] in, 、 The query weight matrix and key-value weight matrix are used for feature mapping; 、 A learnable position embedding vector that captures position dependencies; is the offset between the current position and the context position, representing the relative position relationship; For location With context position The protocol feature mapping correlation degree;

[0096] Generate the protocol feature vector through the above mechanism , solves the problem of traditional Transformer losing location information for long messages (such as Modbus RTU long frames) and enhances the differentiation of different protocol features.

[0097] S120.2. Dynamic generation of mutation protocol rules:

[0098] A deep reinforcement learning framework is constructed, with the protocol parser as the agent and the device as the environment for interactive training. The mapping rule generation strategy is dynamically adjusted through the reward function, which is defined as:

[0099] ;

[0100] in, for Time parsing accuracy, a measure of parsing correctness; for Time data throughput, a measure of parsing efficiency; for The number of parsing errors at any moment, measuring parsing stability; is the weight coefficient and =1, used to balance different optimization objectives; Immediate rewards for reinforcement learning of protocol parsers;

[0101] Based on this reward function, the intelligent agent generates adaptation mapping rules for variant protocols (such as Modbus ASCII with custom check bits and IEC60870-5-103 with field order variations), improving the parsing ability of non-standard protocols.

[0102] S120.3. Iterative optimization of the global rule base:

[0103] Edge nodes regularly update their local rule base Synchronize to the central control module 220 and use the federated learning algorithm to aggregate multi-site rule data. The aggregation formula is:

[0104] ;

[0105] in, is the total number of edge nodes, representing the network scale; For the The amount of training data for each node reflects the data contribution of the node; is the total amount of global data, used for weighted average calculation; A global protocol rule base aggregated across regions.

[0106] This mechanism can achieve collaborative optimization of the global rule base while protecting the privacy of original data on the device side (without uploading message content), adapt to multi-site heterogeneous protocol environments, and improve the generalization capability of the rule base.

[0107] As a further illustration of this embodiment, the protocol parsing submodule 111 in this embodiment adopts a layered state machine design. The transport layer state machine processes Modbus RTU frame checksum, IEC60870-5-103 frame synchronization, and other underlying protocol specifications. The application layer state machine parses function code logic based on the state transition table. Taking the Modbus ASCII protocol as an example, the state machine covers core states such as start character detection, address resolution, function code identification, checksum processing, and frame end judgment. Bit operation parsing of binary protocols is implemented through a hardware accelerator.

[0108] Furthermore, the data mapping submodule 112 constructs a three-layer knowledge graph based on the Neo4j graph database, including a device entity layer (defining device physical properties), a protocol field layer (mapping register addresses to a unified data model), and a mapping rule layer (establishing a bidirectional mapping through relational nodes), and supports dynamic updating of mapping rules using the SPARQL language.

[0109] Furthermore, the plug-in management submodule 113 adopts the OCP3.0 standard hot-swap interface and monitors the power status of the plug-in through the I2C bus. When a plug-in abnormality is detected, a hardware reset or a backup plug-in switching mechanism is triggered to ensure the stable operation of the industrial protocol plug-in.

[0110] As a further illustration of this embodiment, the protocol conversion unit 100 extracts protocol message features through the Transformer-XL network. Its relative position encoding mechanism solves the problem of position information loss in long messages and enhances the distinguishability of different protocol features. The deep reinforcement learning framework uses the protocol parser as the intelligent agent and dynamically generates mutated protocol adaptation rules through a reward function that includes parsing accuracy, throughput, and number of errors. The edge nodes and the central control module 220 use a federated learning algorithm to aggregate multi-site rule bases to achieve collaborative optimization of global rules while protecting the privacy of original data.

[0111] As you can understand, the protocol conversion unit 100 provides external device access interfaces such as RS-485 and Ethernet, is compatible with industrial protocols such as Modbus and IEC 60870-5, and interacts with the master station via a REST API. The hardware design complies with the GB / T 17626 series of electromagnetic compatibility standards and is suitable for industrial environments such as substations and charging stations. Its power consumption control and heat dissipation solutions ensure long-term stable operation.

[0112] The communication adapter unit 200 uses a distributed edge computing architecture to achieve real-time interaction of multi-device data. The communication adapter unit 200 includes an edge computing node module 210 and a central control module 220. The edge computing node module 210 and the central control module 220 establish a bidirectional connection via an industrial-grade communication link to achieve real-time multi-device data collection, protocol conversion, command interaction, and closed-loop control.

[0113] In this embodiment, the edge computing node module 210 is installed at the local device site and connected to the local device through an industrial-grade communication interface, supporting real-time conversion between multiple heterogeneous industrial protocols and the Ethernet protocol. The edge computing node module 210 includes a local protocol processing submodule 211 and an edge communication interface submodule 212, wherein:

[0114] The local protocol processing submodule 211 is used to store the protocol mapping rules of the local device, and realize the real-time collection and format standardization of device data (such as converting the device-specific protocol data into a unified system format):

[0115] When receiving the original protocol data from the local device, it is converted into the system unified data format through the hardware acceleration parsing engine;

[0116] When receiving the system unified format control instructions issued by the central control module 220, it converts them into protocol instructions executable by the local device through the protocol inverse mapping logic;

[0117] The edge communication interface submodule 212 supports two-way data interaction with the central control module 220, including real-time data upload and control command reception.

[0118] Real-time data upload of the device: transmitting the system unified format data converted by the local protocol processing submodule 211;

[0119] Control instruction reception: forwarding the system unified format control instructions issued by the central control module 220 to the local protocol processing submodule;

[0120] Furthermore, in the grid-connected solar-storage-charging system of this embodiment, a local device refers to a single terminal device directly connected to the edge computing node module 210 via an industrial-grade communication interface, including charging piles, energy storage converters, smart meters, and the like. Its communication protocols cover industrial standards such as Modbus, GB / T27930, and CANopen. The system uses a unified, standardized data format for cross-module data exchange. This format is based on the JSON architecture and includes a device type identifier, timestamp, and parameter set.

[0121] Furthermore, the edge computing node module 210 is installed at the local device site and connects to local devices such as charging piles and energy storage converters via an industrial-grade communication interface, supporting real-time conversion between heterogeneous industrial protocols such as Modbus and CANopen and Ethernet protocols. This edge computing node module 210 includes a local protocol processing submodule 211 and an edge communication interface submodule 212. The local protocol processing submodule 211 uses an FPGA hardware acceleration engine to achieve high-speed parsing of protocol messages, breaking down raw protocols such as Modbus frames and CANopen data segments into basic data units. Combined with a JSON-formatted protocol mapping rule library, it dynamically maps device-specific protocol fields to the system's standardized data format, while also supporting remote updates of mapping rules from the central control module. The edge communication interface submodule 212 establishes a bidirectional communication link with the central control module 220 based on the MQTT protocol. During uplink transmission, it encapsulates standardized data into MQTT messages based on device type and node ID. During downlink transmission, it receives central commands and triggers the reverse conversion process in the local protocol processing submodule 211. TLS encryption is used at the link layer to ensure data transmission security, enabling bidirectional interaction between real-time device data upload and control command reception.

[0122] The central control module 220 is installed on the system management platform side and establishes a connection with multiple edge computing node modules 210 through an industrial-grade communication link to support cross-regional centralized management of the local device areas covered by each edge computing node module 210. The central control module 220 includes a global rule management submodule 221 and a data aggregation processing submodule 222, wherein:

[0123] The global rule management submodule 221 is used to store, update, and distribute protocol mapping rules to the cross-region edge computing node modules 210, and supports rule version control (such as marking rule application scenarios and entry into force conditions) and conflict detection (identifying logical contradictions between rules);

[0124] The data aggregation processing submodule 222 is used to receive device data uploaded by the cross-region edge computing node module 210 and perform protocol-independent standardized aggregation (such as integrating data by device type and regional dimensions).

[0125] In this embodiment, the edge computing node module 210 and the central control module 220 perform real-time interaction and control of multi-device data, including the following steps:

[0126] S230.1. Edge computing node module 210 connects to various local devices (energy production / storage / consumption devices) via industrial-grade communication interfaces and collects device operating data (e.g., power, voltage, and operating status) at preset intervals. It also performs local preprocessing on the raw data, including removing outliers, standardizing parameter units (e.g., normalizing voltage to V), and adding device identifiers and timestamps to generate preliminary, usable device status data.

[0127] S230.2. The edge computing node module 210 converts the pre-processed device data from the original protocol format (e.g., a device-specific industrial protocol) into a system-standardized data format (e.g., JSON) based on the built-in protocol mapping rule library. The converted data is then encapsulated into priority communication data frames, distinguishing between real-time control data (high priority) and routine monitoring data (low priority), providing a priority scheduling basis for data transmission.

[0128] S230.3. Edge computing node module 210 uploads the encapsulated standardized data frames to central control module 220 via an industrial-grade communication link (e.g., wireless private network, industrial Ethernet). Central control module 220 then performs global integration and processing on the data uploaded by multiple edge nodes, including storage by region or device type, cross-node time synchronization and calibration (ensuring data timestamp consistency), and generation of comprehensive device operation indicators (e.g., total power and average operating efficiency of the local device cluster).

[0129] S230.4. Based on the integrated global data and system control strategy, the central control module 220 generates device control instructions (e.g., energy storage system charge and discharge power regulation, charging station start and stop control) or rule update instructions. Based on the instruction type and target device characteristics, the central control module 220 selects unicast or multicast transmission mode (e.g., unicast for control instructions for specific devices, multicast for global rule updates), and adds transmission priority tags to time-sensitive instructions to ensure real-time transmission of critical instructions.

[0130] S230.5. After receiving the instruction from the central control module 220, the edge computing node module 210 converts the standardized instruction into the original protocol format recognizable by the target device based on the local protocol mapping rule base. The converted instruction is sent to the corresponding local device and the execution results of the local device (such as instruction reception status, execution completion status, and exception code) are collected in real time to form an instruction execution feedback data frame and temporarily stored locally on the edge node.

[0131] S230.6. The edge computing node module 210 uploads the instruction execution feedback data frame to the central control module 220; the central control module 220 verifies the validity of the instruction execution based on the feedback result (such as whether the device status changes as expected by the instruction), triggers an adaptive retransmission mechanism for instructions that fail to execute (such as increasing the number of retransmissions and improving the transmission priority), generates alarm information for abnormal situations with multiple failures and records it in the system log, forming a closed-loop management of device control.

[0132] As a further illustration of this embodiment, the edge computing node module 210 in this embodiment connects to devices such as photovoltaic inverters and energy storage converters via industrial-grade communication interfaces (such as RS-485 and CANopen), collecting operational data such as power, voltage, and switch status on a second-by-second basis. A sliding window filtering mechanism is used in the preprocessing phase to eliminate sudden interference data, and a unified timestamp is added to cross-node data via a clock synchronization protocol (IEEE1588v2) to ensure time consistency of data from different devices. For non-standard units in protocol fields (such as kV to V), unit standardization is automatically completed via a knowledge graph mapping library. Furthermore, the hardware of the edge computing node module 210 in this embodiment complies with the GB / T17626.4 industrial immunity standard, passing the Level 4 test for electrical fast transient bursts. A surge protection device (SPD) is configured at the power input to ensure stable operation in strong electromagnetic interference environments.

[0133] As a further illustration of this embodiment, the local protocol processing submodule 211 in this embodiment converts ModbusRTU, IEC60870-5-103 and other protocol data into a unified data model in JSON format based on a state machine parsing engine. Data frames are prioritized by service type:

[0134] High priority includes energy storage charging and discharging instructions and charging pile emergency stop signals;

[0135] Medium priority is for equipment status monitoring data, and low priority is for historical data query requests;

[0136] Priority scheduling uses a weighted circular queue (WRR) algorithm to allocate double transmission bandwidth for high-priority data, and implements parallel processing of data of different priorities through hardware FIFO queues to avoid delays in real-time control instructions.

[0137] As a further illustration of this embodiment, the data aggregation processing submodule 222 of the central control module 220 in this embodiment receives data uploaded by multiple edge nodes, categorizes and stores it by region (e.g., substation number) and device type (photovoltaic / energy storage / charging), and generates comprehensive metrics such as regional total power and device operating efficiency through a distributed computing framework. The command generation phase, based on a model predictive control (MPC) algorithm, combines grid dispatch instructions (e.g., peak and off-peak electricity price signals) with device status prediction data to generate energy storage power adjustment instructions (e.g., a 20% reduction in charging power) or charging station start / stop control instructions. The command format complies with the DL / T634.5104 standard. Furthermore, the central control module 220 is deployed in the substation communications room and utilizes a dual-server hot standby architecture. The primary and backup servers synchronize their status in real time through a heartbeat detection mechanism, ensuring continuous system operation.

[0138] It should be noted that the communication adapter unit 200 in this embodiment adopts a dual-link redundant architecture to ensure the reliability of command transmission: the wired link relies on the PRP protocol of the optical fiber ring network to build the main channel, carrying the stable transmission of high-priority commands; the wireless link uses the 5.8 GHz industrial private network as a backup link to cope with wiring-constrained scenarios;

[0139] Furthermore, the central control module 220 dynamically selects a transmission strategy based on the characteristics of the instructions: for time-sensitive instructions such as fault tripping and emergency power adjustment, the target edge node is directly addressed via unicast to ensure accurate delivery of the instructions; for batch instructions such as global protocol rule updates, multicast is used to simultaneously push them to all edge nodes to improve the efficiency of rule deployment;

[0140] Furthermore, after receiving the instruction, the edge computing node initiates the protocol conversion and execution process. The specific process is as follows: calling the local protocol mapping rule library, converting the standardized instruction into a device-specific protocol message (such as the Modbus write register frame format), and sending it to the target device through the industrial-grade communication interface; at the same time, the execution status feedback of the device (including instruction reception confirmation, execution progress code, exception code, etc.) is collected in real time and temporarily stored in the local cache;

[0141] In addition, the central control module 220 verifies the validity of the instruction execution through the following dual mechanism:

[0142] First, parse the device feedback message uploaded by the edge node, extract the Modbus exception code, CANopen status bit and other identifiers, and determine whether the command parsing is successful;

[0143] Secondly, compare the device sampling data before and after the command is issued (such as the power value of the energy storage converter and the start and stop status of the charging pile) to verify whether the physical status meets expectations;

[0144] If the expected state change is not detected within 100ms, the system automatically triggers the adaptive retransmission process (the first retransmission interval is 100ms, and the subsequent intervals increase exponentially). If the cumulative retransmission fails three times, an alarm event is generated and recorded in the system log, and the manual intervention process is triggered simultaneously to ensure the integrity of the control closed loop.

[0145] It is understandable that the collaborative design of the edge computing node module 210 and the central control module 220 in this embodiment achieves the following technical advantages:

[0146] Protocol compatibility: Through standardized data formats, it is compatible with more than ten industrial protocols such as Modbus and CANopen. When new devices are connected, only the local protocol mapping rules need to be updated without modifying the core system architecture.

[0147] Real-time performance guarantee: Hardware-accelerated parsing and edge-side pre-processing mechanisms reduce data transmission volume and central computing pressure, and control instruction closed-loop delay within industrial-grade real-time standards;

[0148] System scalability: The global rule management function of the central control module 220 supports centralized configuration and rule synchronization of cross-regional edge nodes, which is suitable for large-scale deployment of solar-storage-charging systems.

[0149] Energy coordination unit 300, used for multi-objective optimization of photovoltaic output, energy storage SOC, charging load, and grid instructions, integrates LSTM time series prediction model and mixed integer programming algorithm, controls energy storage charging and discharging power via CAN bus, and uses virtual meter technology to dynamically adjust charging power;

[0150] In this embodiment, the energy coordination unit 300 includes a prediction modeling module 310, a multi-objective optimization module 320, and a coordinated control module 330, wherein:

[0151] The prediction modeling module 310 integrates a long short-term memory network time series prediction model to predict the photovoltaic power generation power, charging load demand and energy storage remaining capacity in the future period based on the historical data of photovoltaic output, charging load and energy storage SOC;

[0152] As a further illustration of this embodiment, the prediction modeling module 310 implements trend deduction of multi-dimensional data based on the long short-term memory network (LSTM) time series prediction model. The specific process is as follows:

[0153] First, the real-time photovoltaic power, energy storage SOC, charging load and meteorological data of the communication adapter unit 200 are collected, and outliers are eliminated and missing data are repaired through statistical methods to form a standardized time series;

[0154] Subsequently, the LSTM model is used to learn from historical data sequences. This model processes long-sequence dependencies through a gating mechanism, capturing the temporal characteristics and periodic patterns of energy data. It then outputs forecasts of photovoltaic power generation, charging load demand, and remaining energy storage capacity for future periods, providing data support for optimization decisions.

[0155] If the prediction accuracy does not meet the requirements, it will automatically switch to alternative prediction schemes such as exponential smoothing to ensure data continuity.

[0156] The multi-objective optimization module 320 uses photovoltaic absorption, energy storage life, and grid response as optimization objectives. It combines energy storage SOC constraints, charge and discharge power constraints, and grid interaction power constraints to build a multi-objective optimization model and uses a mixed integer programming algorithm to solve it, generating an energy storage charge and discharge power and charging power allocation plan.

[0157] The collaborative control module 330 issues energy storage charging and discharging control instructions via the CAN bus and dynamically allocates charging pile power using virtual meter technology. It collects the regulated data (actual energy storage power, charging pile load) and feeds it back to the multi-objective optimization module 320 to achieve rolling optimization control.

[0158] As a further illustration of this embodiment, the collaborative control module 330 in this embodiment transmits charging and discharging control instructions to the energy storage inverter via the CAN bus, dynamically allocates charging pile power using virtual meter technology, and converts optimization decisions into physical controls that can be executed by the equipment; at the same time, by collecting data such as actual energy storage power, charging pile load, and the latest SOC, it transmits the data back to the multi-objective optimization module 320, forming a closed-loop control link of "prediction-optimization-control-feedback"; if the instruction execution fails or the deviation exceeds the limit, the adaptive retransmission mechanism or alarm process is triggered to ensure the continuity of system control.

[0159] In this embodiment, the multi-objective optimization module 320 performs the multi-objective optimization including the following steps:

[0160] S320.1. Collect real-time PV output, energy storage SOC, charging load, and grid command data from the communication adapter unit 200, eliminate outliers using the 3σ criterion, and use linear interpolation to repair missing data.

[0161] As a further explanation of this step, this embodiment uses the 3σ criterion to eliminate outliers and uses linear interpolation to repair missing data, including the following steps:

[0162] First, a one-minute real-time data stream is obtained from the communication adapter unit 200, including photovoltaic output, energy storage SOC, charging load and grid command data;

[0163] Then, statistical analysis methods are used to identify outliers: statistical characteristics of the data series (such as mean and standard deviation) are calculated, and data points that deviate from the normal range (such as a sudden drop of 50% in photovoltaic power) are marked;

[0164] Then, the outliers are replaced by the median within the window to avoid the interference of outliers on subsequent analysis;

[0165] Finally, for consecutive missing data points (such as no data at three consecutive sampling points), they are repaired by linear interpolation of adjacent valid data points to ensure the integrity and continuity of the data.

[0166] S320.2. Call the long short-term memory network time series prediction model of the prediction modeling module 310 to learn the preprocessed historical data sequence and output the predicted values ​​of photovoltaic power generation power, charging load, and remaining energy storage capacity in the future time period;

[0167] As a further explanation of this step, this embodiment uses a long short-term memory network time series prediction model to perform trend deduction. The specific process is as follows: First, the LSTM time series prediction algorithm of the prediction modeling module 310 is called: This algorithm is a well-known deep learning model that captures long-term dependencies in time series through a gating mechanism (input gate, forget gate, output gate);

[0168] Then, the pre-processed historical data series (such as power and meteorological data from the previous 72 hours) is fed into the model, which automatically learns the periodic features in the data (such as the daily periodicity of photovoltaic output and the peak and valley characteristics of charging load).

[0169] Next, the algorithm outputs forecast results for the next four hours and 15 minutes, including trend forecasts for photovoltaic power generation, charging load demand, and remaining energy storage capacity. Simultaneously, the algorithm generates quantitative indicators of forecast uncertainty (such as confidence intervals) to assess the reliability of the forecast results.

[0170] Finally, the system monitors the prediction accuracy (such as the deviation between the predicted value and the actual value) in real time. If the accuracy does not meet the standard (such as the prediction deviation is >15% for three consecutive times), it will automatically switch to alternative prediction schemes such as exponential smoothing. This design ensures the robustness of the prediction process.

[0171] S320.3. Set the energy storage SOC operating range, charge and discharge power limits, and grid interaction power threshold as constraints, and build a multi-objective optimization model with maximizing PV absorption, extending energy storage life, and improving grid response capabilities as the objective functions.

[0172] As a further explanation of this step, the specific process of setting multi-objective constraints and building a multi-objective optimization model in this embodiment is as follows:

[0173] First, build a physical constraint system as shown below:

[0174] Energy storage operation constraints: Set a SOC safety range (e.g., 30%-90%) to prevent overcharging and over-discharging that affect the energy storage lifespan; limit the upper limit of charge and discharge power (e.g., 80% of rated power) to match the actual capacity of the energy storage device;

[0175] Grid interaction constraints: Set a grid interaction power threshold (e.g., ±100kW) based on the distribution network capacity to ensure that system operation does not exceed the grid's carrying capacity.

[0176] Device status constraints: Dynamically adjust constraint boundaries based on the real-time status of devices (such as the number of charging piles in use and the cleanliness of photovoltaic panels) (e.g., tightening the charging pile power limit during peak charging periods);

[0177] Then, define the following multi-objective optimization system:

[0178] Photovoltaic power consumption goals: maximize the local consumption of photovoltaic power, reduce abandoned power, and improve the utilization rate of renewable energy;

[0179] Energy storage lifespan goal: Extend the lifespan of energy storage equipment and reduce system lifecycle costs by optimizing charging and discharging strategies (e.g., avoiding deep charging and discharging, and balancing SOC fluctuations);

[0180] Grid response goal: Quickly track grid instructions (such as peak shaving and frequency modulation signals), achieve friendly interaction between the system and the grid, and improve power system stability;

[0181] Finally, a dynamic weight allocation mechanism is used to balance the priorities of multiple objectives: based on the type of grid dispatch instructions (such as emergency peak regulation and conventional regulation) and the current operating conditions of the system (such as the period of peak photovoltaic power generation and the low SOC state of energy storage), the weight coefficient of each objective is adjusted in real time (for example, when the grid issues an emergency instruction, the grid response target weight is increased to 0.6, and the photovoltaic absorption weight is reduced to 0.3), ensuring that the optimization results meet current business needs.

[0182] S320.4. Use a mixed integer programming algorithm to solve the multi-objective optimization model, generate executable instructions for energy storage charging and discharging power and charging pile power allocation, and mark the instruction priority (e.g., power grid emergency instructions are set to the highest priority);

[0183] As a further illustration of this embodiment, the specific process of using the mixed integer programming algorithm to solve the multi-objective optimization model and generate executable instructions is as follows:

[0184] First, the mixed integer programming algorithm can handle complex optimization problems involving a mixture of discrete variables (such as the start and stop status of charging piles) and continuous variables (such as energy storage power).

[0185] Then, set a solution timeout threshold (e.g., 30 seconds) to prioritize real-time instruction generation (if the solution is not completed within the timeout, the optimization result of the previous cycle is used as a temporary solution);

[0186] Next, based on the constraints and objective function, the mixed integer programming algorithm outputs energy storage charging and discharging power commands (with an accuracy of 1kW) and a charging pile power allocation plan (dynamically dividing the power cap by charging pile group). The system also prioritizes the command type: grid emergency commands (such as frequency support signals) are assigned the highest priority, requiring them to be issued within 1 second; energy storage lifespan protection commands (such as adjustments when the SOC approaches the safety margin) are assigned a medium priority; and conventional power allocation commands (such as photovoltaic surplus power absorption) are assigned a low priority.

[0187] Finally, the optimization results are converted into a communication format recognizable by the device: energy storage instructions are encapsulated as CAN bus messages (compliant with the energy storage inverter communication protocol), and charging pile instructions are converted into OCPP protocol control frames to ensure that the instructions can be correctly parsed and executed by the hardware device.

[0188] S320.5. Based on the real-time control data fed back by the collaborative control module 330, if there is a deviation from the optimization instruction, the parameter update of the multi-objective optimization model is triggered, and steps S320.3-S320.4 are re-executed to achieve periodic rolling optimization.

[0189] As a further explanation of this embodiment, the process of implementing periodic rolling optimization in this embodiment is as follows:

[0190] First, receive real-time device operation data (such as actual energy storage power, actual charging pile load, and latest SOC) fed back by the collaborative control module 330, with a sampling frequency of 1 second.

[0191] Then, the instruction execution effect is evaluated using a deviation analysis algorithm: the absolute deviation (e.g., the difference between the actual energy storage power and the instruction power) and the relative deviation (e.g., the deviation percentage) between the actual value and the instruction value are calculated. If the deviation exceeds a preset threshold (e.g., 15%) or lasts for more than three sampling periods (i.e., three seconds), it is determined to be an execution anomaly.

[0192] Next, the model parameter update mechanism is triggered: the latest data (including anomaly feedback) is recollected, constraints in the optimization model are adjusted (e.g., tightening energy storage power limits) or target weights (e.g., increasing the grid response target weight), and the "constraint construction - optimization solution - instruction generation" process is re-executed. Simultaneously, the system records anomalies (e.g., anomaly type, occurrence time, and adjustment strategy) to form an operation log to support subsequent analysis and optimization.

[0193] Finally, set a rolling optimization cycle (e.g., 5 minutes): regardless of whether there is a deviation, the system automatically repeats the optimization process every 5 minutes, combining the latest forecast data with the equipment status to dynamically adjust the control strategy to ensure that the system always operates in the optimal state.

[0194] It should be noted that the multi-objective optimization module 320 in this embodiment is deployed on an industrial-grade edge server. It relies on a multi-core processor and a real-time operating system to meet the computing power requirements of complex algorithms. It obtains device data from the communication adapter unit 200 in real time through an Ethernet interface and issues control commands to the energy storage converter via a CAN bus expansion card, establishing a low-latency interactive channel for "data input - command output."

[0195] At the same time, the following hierarchical fault tolerance mechanism is designed to deal with operational failures:

[0196] When communication is interrupted, the system uses locally cached data (retaining valid sequences from the previous 10 minutes) to maintain optimized calculations. If a solution anomaly occurs, the system automatically switches to regularized strategies such as "fixed PV power-charging load ratio allocation." Under extreme operating conditions (such as when the energy storage SOC approaches the safety limit), the system triggers a safety mode (limiting energy storage charge and discharge power to 30% of the rated value and forcing charging piles into standby mode) to ensure stable system operation.

[0197] In addition, this embodiment can also adopt the following "offline simulation + online testing" dual-track verification:

[0198] During the offline phase, based on historical data (such as annual PV output and load curves), scenarios such as continuous rainy weather and rapid grid peak regulation were simulated to verify the effectiveness of strategies to increase PV absorption rates and extend energy storage lifespan.

[0199] During the online phase, the system monitors indicators such as grid command tracking delay (target ≤ 2 seconds) and abnormal response time (target ≤ 1 minute) in real time. The algorithm operation status is intuitively presented through the system data dashboard, and the model parameters and constraints are continuously iterated and optimized to ensure the effectiveness of the strategy in actual working conditions.

[0200] The grid interface unit 400 implements secure communication with the dispatching master station based on the national encryption SM2 and SM4 algorithms, and establishes an end-to-end encrypted channel through the vertical encryption module 410;

[0201] In this embodiment, the vertical encryption module 410 includes an encryption tunnel establishment submodule 411, a two-way identity authentication submodule 412, and a hardware acceleration submodule 413, wherein:

[0202] The encrypted tunnel establishment submodule 411 builds an end-to-end encrypted channel based on IPsec-VPN technology, supporting the coordinated application of the national encryption SM2 and SM4 algorithms (SM2 is used for key exchange, and SM4 is used for data encryption). It uses tunnel mode to encapsulate network layer data and encrypts both the header and payload of the IP data packet to ensure the confidentiality of the communication link.

[0203] As a further explanation of this embodiment, this embodiment uses the IKEv2 protocol + the national secret SM2 algorithm to implement key exchange:

[0204] Phase 1 (Security Association Negotiation): The device and the scheduling master exchange SM2 public key certificates, generate session keys using the elliptic curve Diffie-Hellman algorithm, negotiate encryption parameters (SM4-CBC mode, SM3 authentication algorithm), and complete the security policy handshake within 10 seconds.

[0205] Phase 2 (IPsec tunnel establishment): Based on the session key, ESP tunnel mode is used to encapsulate IP packets, encrypting both the packet header and payload (to prevent traffic analysis). Dual-link redundancy is supported, with primary fiber and backup 4G private network.

[0206] In addition, the above data encapsulation mechanism includes:

[0207] For grid interaction data (such as power control instructions and power quality monitoring data), the ESP protocol adds an SM3 authentication header (integrity verification) and an SM4 encryption payload (confidentiality protection) to achieve "one-time transmission, double protection".

[0208] The two-way identity authentication submodule 412 has a built-in device digital certificate issued based on the SM2 algorithm, which performs two-way identity legitimacy verification with the dispatching master station (the sender's certificate contains the public key, and the receiver verifies the signature through the SM2 algorithm); and supports certificate validity checks (such as certificate expiration time and CA signature verification) to prevent illegal devices from accessing the communication network;

[0209] As a further illustration of this embodiment, the bidirectional identity authentication submodule 412 of this embodiment uses the national secret digital certificate as the core to construct the following bidirectional trust mechanism between the device and the dispatching master station:

[0210] Certificate security carrier: The device has a built-in national secret certificate issued by the power industry CA (including device ID, SM2 public key, validity period, etc.). The certificate and the corresponding private key are stored in the hardware encryption module (HSM). The private key only participates in the signature calculation within the HSM hardware to ensure that the private key is not leaked;

[0211] Authentication process: The dispatching master generates a 64-bit random challenge value and sends it. The device calls the HSM's SM2 private key to sign the "challenge value + timestamp (accurate to the moment) + monotonic serial number" and returns the signature result along with the certificate.

[0212] The master station first verifies the validity of the certificate (checking the CA signature, validity period, and certificate revocation list (CRL)), and then verifies the signature using the device's public key. After completion, it initiates a reverse challenge, and the device executes the same verification logic.

[0213] If authentication fails three times in a row, the device triggers a local lock (manual unlocking is required) to block illegal access attempts;

[0214] Full-cycle certificate management: 30 days before a certificate expires, the module automatically submits a renewal request to the CA. If renewal fails, a pre-configured backup certificate is used to maintain communication. This design ensures authentication continuity.

[0215] The hardware acceleration submodule 413 integrates a dedicated encryption chip to perform hardware acceleration on the SM2 asymmetric encryption and SM4 symmetric encryption algorithms; it supports pipeline processing of national secret algorithms, realizes parallel processing of encryption and decryption operations and data transmission, and improves communication efficiency.

[0216] As a further illustration of this embodiment, the hardware acceleration submodule 413 in this embodiment is based on a dedicated national encryption chip to implement efficient operation of the national encryption algorithm in the vertical encryption process, as follows:

[0217] When the encrypted tunnel establishment submodule 411 initiates IKE protocol negotiation, the chip's built-in coprocessor independently undertakes SM2 asymmetric operations (including key exchange and signature verification), freeing up main CPU resources to ensure that the security policy handshake is completed within 10 seconds. After entering the business data interaction phase, the chip uses DMA direct memory access technology to execute SM4 symmetric encryption / decryption operations in parallel with network data transmission and reception. During the ESP tunnel encapsulation process, data encryption and protocol encapsulation are completed simultaneously, reducing transmission delays.

[0218] At the same time, the chip has a built-in temperature sensor and watchdog circuit: when the operating temperature exceeds the limit (such as 85°C) or the operation logic is abnormal, it automatically triggers a hardware reset and resumes normal operation; and supports Gigabit Ethernet full bandwidth encryption to meet the real-time transmission requirements of the power "four remote" data (telemetry, telesignaling, remote control, and remote adjustment).

[0219] In addition, the hardware acceleration submodule 413 and the encryption tunnel establishment submodule 411 work in deep collaboration: during the IKE negotiation phase, the SM2 key is synchronously derived to support fast handshake of the policy; during the data transmission phase, SM4 encryption and ESP tunnel encapsulation are processed in parallel to improve the overall communication efficiency of the vertical encryption module 410.

[0220] The human-computer interaction unit 500 is used to provide visual interaction and remote control functions, use digital twin technology to build a three-dimensional device status model, and realize the trusted storage of charging data through blockchain technology.

[0221] In this embodiment, the human-computer interaction unit 500 includes a visual interaction module 510 and a data evidence storage module 520, wherein:

[0222] The visualization interaction module 510 uses digital twin technology to construct a three-dimensional device status model, mapping the operating parameters (such as voltage, current, and SOC) of photovoltaic equipment, energy storage devices, and charging piles in real time. It supports dynamic rendering of device status (such as green for normal status and flashing red for fault status) and provides an interactive operation interface (such as clicking on the device model to retrieve real-time data and issue control instructions).

[0223] As a further illustration of this embodiment, the visualization interaction module 510 in this embodiment subscribes to real-time data from the communication adapter unit 200 via the MQTT protocol. This data is then parsed into JSON format, filtered using a sliding window to remove outliers, and supplemented with millisecond timestamps based on the NTP clock. Subsequently, device CAD drawings or point cloud data are imported to construct geometric models (e.g., photovoltaic panels and energy storage batteries). A mapping relationship between operating parameters and model properties is established: the energy storage SOC is mapped to the battery model's power visualization (0% → gray, 100% → green, with gradient support); the photovoltaic power drives dynamic brightness adjustment of the panel model; and device fault codes trigger model status alerts (flashing red for a fault, solid yellow for a warning). A browser-based 3D interface was developed using the Three.js WebGL framework, supporting adaptive access on both PC and mobile devices. Interactions include: hovering the mouse to display parameter pop-ups (displaying voltage, current, etc.); clicking the model to display a 24-hour data curve; and double-clicking the model to invoke a remote control pop-up window (connected to the energy collaboration unit 300 API, with commands forwarded to the device via the communication adapter unit).

[0224] The data evidence module 520 implements credible evidence storage of charging data based on blockchain technology, and hashes the charging order information (time, power, and cost) and device interaction logs onto the chain (e.g., using the SHA-256 algorithm to generate data fingerprints); supports automatic execution of smart contracts (e.g., triggering a fee settlement contract after charging is completed), and provides a blockchain data query interface.

[0225] In this embodiment, the visualization interaction module 510 constructs a three-dimensional device state model including the following steps:

[0226] S510.1. Obtain real-time device operation data from communication adapter unit 200, parse it into a unified format (e.g., JSON) through a protocol conversion interface, remove abnormal data, and add a timestamp.

[0227] S510.2. Build a 3D geometric model of the device based on CAD drawings or point cloud data, and use a digital twin mapping algorithm to bind pre-processed operating data to model parameters (e.g., mapping the energy storage SOC to the color gradient properties of the battery model).

[0228] S510.3. Use WebGL technology to generate a browser-side 3D visualization interface that supports PC and mobile access; set up interactive event response mechanisms (such as displaying detailed device parameters when hovering the mouse, and triggering a remote control pop-up window when double-clicking).

[0229] Furthermore, the data storage module 520 implements charging data storage including the following steps:

[0230] S520.1. Combine key data such as charging start and end time, device ID, and transaction amount, and generate a unique data hash value using the SHA-256 algorithm.

[0231] S520.2. Connect to the blockchain network node through the RPC interface and package the hash value and transaction information into a data block (including the previous block hash, timestamp, and Merkle root);

[0232] S520.3. Call the preset smart contract to verify the legitimacy of the transaction (such as charging station permissions and account balance). After verification, the block data is broadcast to the blockchain network to complete the data on-chain storage.

[0233] To further illustrate this embodiment, the data evidence module 520 in this embodiment extracts key charging order data (start and end time, device ID, power, and cost), concatenates it in the format of "device ID + timestamp + parameters," and generates a unique hash value using the SHA-256 algorithm. It connects to a consortium blockchain node (such as Hyperledger Fabric) via an HTTP-RPC interface, packages the hash value and transaction information into a block (the block header contains the previous block hash, evidence timestamp, and Merkle root; the block body encapsulates charging data and device interaction logs). A pre-configured smart contract is invoked to automatically verify the legitimacy of the transaction (e.g., the charging station identity certificate and user account balance). Once verified, it is confirmed using the PBFT consensus algorithm and broadcast to the consortium blockchain nodes for on-chain evidence storage. After on-chain storage, the evidence hash and block height are returned, supporting data traceability (the hash traces the entire lifecycle and makes tampering detectable) and smart settlement (charging completion triggers a fee settlement contract, which then links with a third-party payment platform to transfer the electricity fee).

[0234] In summary, the multiple units in this embodiment implement data interaction and collaboration through a layered architecture, specifically including:

[0235] The communication adapter unit 200, serving as the underlying hub, collects real-time operating data (such as voltage, current, and SOC) from photovoltaics, energy storage, and charging piles through industrial protocols such as Modbus and CAN. After protocol conversion and exception handling, this data is synchronously transmitted to the energy coordination unit 300 and the human-machine interaction unit 500 via a fiber-optic ring network or wireless private network. Based on this data and grid dispatch instructions (such as peak-shaving signals) forwarded by the grid interface unit 400, the energy coordination unit 300 generates energy storage charging and discharging and charging pile power allocation plans through LSTM prediction and mixed integer programming. These plans are then distributed to the physical devices through the communication adapter unit 200, and the control results are fed back to the grid interface unit 400 (for master station interaction) and the human-machine interaction unit 500 (for status display). The grid interface unit 400 interacts bidirectionally with the dispatch master station through a national secret encrypted channel, uploading system operating data (such as photovoltaic absorption and grid interaction power) and receiving instructions. This forms a closed-loop data link of "device data collection - central optimization decision-making - grid security interconnection - human-machine visual management and control." This design enables real-time linkage and information sharing among multiple units.

[0236] Those skilled in the art will appreciate that the process of implementing all or part of the steps of the above embodiments may be accomplished by hardware, or by instructing related hardware through a program.

[0237] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A protocol conversion and communication adaptation system for a photovoltaic storage and charging grid-connected device, characterized in that: include: A protocol conversion unit (100) is used to dynamically analyze heterogeneous protocols of photovoltaic inverters, energy storage converters, charging piles, and grid-side equipment, realize plug-and-play of multiple protocols through a plug-in architecture module (110), and automatically update a protocol mapping rule base based on a machine learning algorithm; A communication adapter unit (200) adopts a distributed edge computing architecture to realize real-time interaction of multi-device data; the communication adapter unit (200) includes an edge computing node module (210) and a central control module (220), and a bidirectional connection is established between the edge computing node module (210) and the central control module (220) via an industrial-grade communication link to realize real-time collection of multi-device data, protocol conversion, instruction interaction and closed-loop control; An energy coordination unit (300) is used to perform multi-objective optimization on photovoltaic output, energy storage SOC, charging load and grid instructions, integrates an LSTM time series prediction model and a mixed integer programming algorithm, controls energy storage charging and discharging power via a CAN bus, and dynamically adjusts charging power using virtual meter technology; The grid interface unit (400) realizes secure communication with the dispatching master station based on the national encryption SM2 and national encryption SM4 algorithms, and establishes an end-to-end encryption channel through the vertical encryption module (410); The human-computer interaction unit (500) is used to provide visual interaction and remote control functions, use digital twin technology to build a three-dimensional device status model, and realize the trusted storage of charging data through blockchain technology.

2. The protocol conversion and communication adaptation system for the photovoltaic storage and charging grid-connected device according to claim 1 is characterized in that: The plug-in architecture module (110) includes a protocol parsing submodule (111), a data mapping submodule (112) and a plug-in management submodule (113), wherein: The protocol parsing submodule (111) is internally integrated with a state machine parsing engine, which supports syntax analysis of binary coding protocols and function code protocols in the field of industrial automation; The data mapping submodule (112) constructs a protocol data model mapping library based on knowledge graph technology, and dynamically maps device parameters of different protocols to a unified data format; The plug-in management submodule (113) realizes plug-and-play of industrial protocol plug-ins through a hot-swap interface, and supports plug-in operation status monitoring and fault self-healing.

3. The protocol conversion and communication adaptation system for the photovoltaic storage and charging grid-connected device according to claim 1 is characterized in that: The protocol conversion unit (100) automatically updates the protocol mapping rule base based on a machine learning algorithm, comprising the following steps: S120.

1. Protocol message feature extraction: Use Transformer-XL network to input protocol message sequence , encode, and process long sequence context dependencies through the relative position encoding mechanism. The formula is: ； in, 、 The query weight matrix and key-value weight matrix are used for feature mapping; 、 A learnable position embedding vector that captures position dependencies; is the offset between the current position and the context position, representing the relative position relationship; For location With context position The protocol feature mapping correlation degree; S120.

2. Dynamic generation of mutation protocol rules: A deep reinforcement learning framework is constructed, with the protocol parser as the agent and the device as the environment for interactive training. The mapping rule generation strategy is dynamically adjusted through the reward function, which is defined as: ； in, for Time parsing accuracy, a measure of parsing correctness; for Time data throughput, a measure of parsing efficiency; for The number of parsing errors at any moment, measuring parsing stability; is the weight coefficient and =1, used to balance different optimization objectives; Immediate rewards for reinforcement learning of protocol parsers; S120.

3. Iterative optimization of the global rule base: Edge nodes regularly update their local rule base Synchronize to the central control module (220), and use the federated learning algorithm to aggregate multi-site rule data. The aggregation formula is: ； in, is the total number of edge nodes, representing the network scale; For the The amount of training data for each node reflects the data contribution of the node; is the total amount of global data, used for weighted average calculation; A global protocol rule base aggregated across regions.

4. The protocol conversion and communication adaptation system for the photovoltaic storage and charging grid-connected device according to claim 1 is characterized in that: The edge computing node module (210) is installed at a local device site, connected to the local device via an industrial-grade communication interface, and supports real-time conversion between multiple heterogeneous industrial protocols and Ethernet protocols; the edge computing node module (210) includes a local protocol processing submodule (211) and an edge communication interface submodule (212), wherein: The local protocol processing submodule (211) is used to store the protocol mapping rules of the local device, and realize the real-time collection and format standardization of device data: When receiving the original protocol data from the local device, it is converted into the system unified data format through the hardware acceleration parsing engine; When receiving a system unified format control instruction issued by the central control module (220), converting it into a protocol instruction executable by the local device through protocol inverse mapping logic; The edge communication interface submodule (212) supports two-way data interaction with the central control module (220), including real-time data upload of devices and control instruction reception: Real-time data upload of the device: transmitting the system unified format data converted by the local protocol processing submodule (211); Control instruction reception: forwarding the system unified format control instructions issued by the central control module (220) to the local protocol processing submodule; The central control module (220) is installed on the system management platform side, and is connected to multiple edge computing node modules (210) via industrial-grade communication links to support cross-regional centralized management of local device areas covered by each edge computing node module (210); the central control module (220) includes a global rule management submodule (221) and a data aggregation processing submodule (222), wherein: The global rule management submodule (221) is used to store, update and distribute protocol mapping rules to cross-region edge computing node modules (210), and supports version control and conflict detection of rules; The data aggregation processing submodule (222) is used to receive device data uploaded by the cross-region edge computing node modules (210) and perform protocol-independent standardized aggregation.

5. The protocol conversion and communication adaptation system for the photovoltaic storage and charging grid-connected device according to claim 4 is characterized in that: The edge computing node module (210) and the central control module (220) perform real-time interaction and control of multi-device data, including the following steps: S230.1, the edge computing node module (210) is connected to local multiple types of equipment through an industrial-grade communication interface, and collects equipment operation data according to a preset period; and performs local pre-processing on the original data; S230.2, the edge computing node module (210) converts the pre-processed device data from the original protocol format into a standardized data format unified by the system based on the built-in protocol mapping rule library; and encapsulates the converted data into a communication data frame with priority, distinguishes real-time control data from regular monitoring data, and provides a priority scheduling basis for data transmission; S230.3, the edge computing node module (210) uploads the encapsulated standardized data frame to the central control module (220) via the industrial-grade communication link; and the central control module (220) performs global integration processing on the data uploaded by multiple edge nodes; S230.4, the central control module (220) generates a device control instruction or a rule update instruction based on the integrated global data and the system control strategy; selects a unicast or multicast transmission mode according to the instruction type and the characteristics of the target device, and adds a transmission priority tag to high-timeliness instructions to ensure the real-time transmission of key instructions; S230.5, after receiving the instruction issued by the central control module (220), the edge computing node module (210) converts the standardized instruction into the original protocol format recognizable by the target device based on the local protocol mapping rule base; sends the converted instruction to the corresponding local device, and collects the execution result of the local device in real time, forming an instruction execution feedback data frame and temporarily storing it locally at the edge node; S230.6, the edge computing node module (210) uploads the instruction execution feedback data frame to the central control module (220); the central control module (220) verifies the effectiveness of the instruction execution based on the feedback result, triggers an adaptive retransmission mechanism for instructions that fail to execute, generates alarm information for abnormal situations with multiple failures and records it in the system log, thereby forming a closed-loop management of device control.

6. The protocol conversion and communication adaptation system for the photovoltaic storage and charging grid-connected device according to claim 1 is characterized in that: The energy coordination unit (300) includes a prediction modeling module (310), a multi-objective optimization module (320) and a coordinated control module (330), wherein: The prediction modeling module (310) integrates a long short-term memory network time series prediction model to predict the photovoltaic power generation, charging load demand and energy storage remaining capacity in the future period based on the historical data of photovoltaic output, charging load and energy storage SOC; The multi-objective optimization module (320) takes photovoltaic absorption, energy storage life, and grid response as optimization objectives, combines energy storage SOC constraints, charge and discharge power constraints, and grid interaction power constraints to construct a multi-objective optimization model, and adopts a mixed integer programming algorithm to solve it, thereby generating a distribution plan for energy storage charge and discharge power and charging power; The coordinated control module (330) issues energy storage charging and discharging control instructions via the CAN bus and dynamically allocates charging pile power using virtual meter technology; collects the controlled data and feeds it back to the multi-objective optimization module (320) to achieve rolling optimization control.

7. The protocol conversion and communication adaptation system for the photovoltaic storage and charging grid-connected device according to claim 6 is characterized in that: The multi-objective optimization module (320) performs multi-objective optimization including the following steps: S320.1, collecting real-time photovoltaic output, energy storage SOC, charging load and grid instruction data of the communication adapter unit (200), eliminating abnormal values ​​by using the 3σ criterion, and repairing missing data by using linear interpolation; S320.2, calling the long short-term memory network time series prediction model of the prediction modeling module (310), learning the pre-processed historical data sequence, and outputting the predicted values ​​of photovoltaic power generation power, charging load and energy storage remaining capacity in the future period; S320.

3. Set the energy storage SOC operating range, charge and discharge power limits, and grid interaction power threshold as constraints, and build a multi-objective optimization model with maximizing PV absorption, extending energy storage life, and improving grid response capabilities as the objective functions. S320.

4. Use a mixed integer programming algorithm to solve the multi-objective optimization model, generate executable instructions for energy storage charging and discharging power and charging pile power allocation, and mark the instruction priorities; S320.

5. If the real-time control data fed back by the collaborative control module (330) deviates from the optimization instruction, the parameter update of the multi-objective optimization model is triggered, and steps S320.3-S320.4 are re-executed to achieve periodic rolling optimization.

8. The protocol conversion and communication adaptation system for the photovoltaic storage and charging grid-connected device according to claim 1 is characterized in that: The vertical encryption module (410) includes an encryption tunnel establishment submodule (411), a two-way identity authentication submodule (412), and a hardware acceleration submodule (413), wherein: The encrypted tunnel establishment submodule (411) builds an end-to-end encrypted channel based on IPsec-VPN technology, supports the coordinated application of the national encryption SM2 and national encryption SM4 algorithms; uses the tunnel mode to encapsulate network layer data, encrypts the header and payload of the IP data packet at the same time, and ensures the confidentiality of the communication link; The bidirectional identity authentication submodule (412) has a built-in device digital certificate issued based on the SM2 algorithm, and performs bidirectional identity legitimacy verification with the dispatching master station; and supports certificate validity checking to prevent illegal devices from accessing the communication network; The hardware acceleration submodule (413) integrates a dedicated encryption chip to perform hardware acceleration on the SM2 asymmetric encryption and SM4 symmetric encryption algorithms; it supports pipeline processing of national secret algorithms, realizes parallel processing of encryption and decryption operations and data transmission, and improves communication efficiency.

9. The protocol conversion and communication adaptation system for the photovoltaic storage and charging grid-connected device according to claim 1 is characterized in that: The human-computer interaction unit (500) includes a visual interaction module (510) and a data evidence storage module (520), wherein: The visualization interaction module (510) uses digital twin technology to construct a three-dimensional equipment status model, and maps the operating parameters of photovoltaic equipment, energy storage devices and charging piles in real time; supports dynamic rendering of equipment status, and provides an interactive operation interface; The data evidence storage module (520) realizes credible evidence storage of charging data based on blockchain technology, and hashes charging order information and device interaction logs onto the chain; supports automatic execution of smart contracts, and provides a blockchain data query interface.

10. The protocol conversion and communication adaptation system for the photovoltaic storage and charging grid-connected device according to claim 9, characterized in that: The visualization interaction module (510) constructs a three-dimensional device status model comprising the following steps: S510.

1. Obtaining real-time operation data of the device from the communication adapter unit (200), parsing it into a unified format through a protocol conversion interface, removing abnormal data and adding a timestamp; S510.

2. Build a 3D geometric model of the equipment based on CAD drawings or point cloud data, and bind the preprocessed operating data to the model parameters using a digital twin mapping algorithm; S510.

3. Use WebGL technology to generate a browser-side 3D visualization interface that supports PC and mobile access; set up an interactive event response mechanism.

Images