Multi-modal communication edge aggregation gateway system for virtual power plant

Abstract

The invention provides a multi-modal communication edge aggregation gateway system for a virtual power plant, and the system comprises a smart energy unit which serves as edge gateway equipment and is used for carrying out the bidirectional communication with a user branch load, a distributed energy source and a flexible load; the measurement monitoring unit is arranged on a user branch load loop or a user side equipment end and is used for acquiring power utilization data of a branch line and an equipment operation condition; a multi-mode communication module; and the self-adaptive protocol switching unit is used for dynamically selecting communication protocols and realizing switching among the protocols according to the communication environment parameters, the equipment types and the data transmission requirements. The invention thoroughly solves the problem that the traditional communication system cannot adapt to different application scenes and equipment type communication requirements by adopting a single communication protocol, realizes the goal of dynamically adjusting the communication strategy according to the network environment change, and effectively avoids the phenomenon that the communication quality is reduced or even interrupted.

Description

Technical Field

[0001] The present invention relates to virtual power plant communication technology, and in particular to a multimodal communication edge aggregation gateway system for virtual power plants. Background Art

[0002] With the rapid development of renewable energy generation and distributed energy, virtual power plants (VPPs), as a key means of aggregating distributed energy resources, are playing an increasingly important role in power systems. VPPs require real-time data collection and coordinated control of a large number of distributed energy devices, placing extremely high demands on communication systems.

[0003] At present, the virtual power plant communication system mainly has the following problems: First, traditional communication systems typically use a single protocol, making it difficult to adapt to the communication needs of diverse application scenarios and device types. This makes it impossible to dynamically adjust communication strategies as network environments change, leading to decreased communication quality or even interruption.

[0004] Second, existing systems lack intelligent protocol selection mechanisms, unable to automatically select the optimal communication protocol based on real-time network conditions. Most systems still rely on manual configuration, resulting in slow response times and difficulty meeting the real-time requirements of virtual power plants. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides a multimodal communication edge aggregation gateway system for virtual power plants to solve the problem that existing communication systems usually adopt a single communication protocol, are difficult to adapt to the communication needs of different application scenarios and device types, and cannot automatically select the optimal communication protocol according to real-time network conditions.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a multimodal communication edge aggregation gateway system for virtual power plants, comprising: Smart energy unit, as an edge gateway device, is used for two-way communication with user branch loads, distributed energy resources, and flexible loads; At least one measurement and monitoring unit, installed in the user's branch load circuit or the user's side equipment end, used to collect branch line power consumption data and equipment operating conditions; a multimodal communication module, integrated into the smart energy unit, supporting multiple communication protocols, including at least two of VPDN, 5G, WiFi, LoRa, and Zigbee; Adaptive protocol switching unit, used to dynamically select communication protocols and switch between protocols based on communication environment parameters, device type and data transmission requirements; The measurement and monitoring unit communicates with the smart energy unit through the multimodal communication module, and the smart energy unit uploads the aggregated data to the virtual power plant platform or load management system through the selected communication protocol.

[0008] As a preferred solution of the multimodal communication edge aggregation gateway system for virtual power plants described in the present invention, the adaptive protocol switching unit includes: Communication quality detection module, used to monitor the delay, packet loss rate, signal strength, bandwidth utilization of each communication link in real time, and identify network topology and interference sources; A protocol evaluation module, configured to calculate a comprehensive performance score for each communication protocol based on the monitoring results of the communication quality detection module; A switching decision module, configured to select a communication protocol and perform a switching operation based on the comprehensive performance score, data priority, and service type; The protocol switching cache mechanism is used to temporarily store data to be transmitted during the protocol switching process to ensure the continuity of data transmission; The switching status synchronization module is used to report the current communication protocol status to the virtual power plant platform in real time.

[0009] As a preferred solution of the multimodal communication edge aggregation gateway system for virtual power plants described in the present invention, the protocol evaluation module adopts a weighted scoring algorithm, wherein: The delay weight coefficient is W1, the packet loss rate weight coefficient is W2, the signal strength weight coefficient is W3, and the bandwidth utilization weight coefficient is W4; The comprehensive performance score S = W1×(1-normalized delay value) + W2×(1-normalized packet loss rate value) + W3×normalized signal strength value + W4×(1-normalized bandwidth utilization value).

[0010] As a preferred solution of the multimodal communication edge aggregation gateway system for virtual power plants described in the present invention, the smart energy unit further includes: Edge computing processors for local data preprocessing, load forecasting, and decision analysis; A data compression module is used to dynamically adjust the data compression rate and reporting frequency according to the bandwidth characteristics of the selected communication protocol; Device identification module, used to identify newly connected user-side devices and perform parameter configuration; User interaction terminal, with touch screen display function, used to display equipment operating status, response invitation information and profit estimation; The decision-making assistance module provides users with response strategy recommendations based on historical response data and current load conditions.

[0011] As an optimal solution of the multimodal communication edge aggregation gateway system for virtual power plants described in the present invention, the measurement and monitoring unit supports cascade deployment to form a tree-like network topology, wherein the lower-level measurement and monitoring unit aggregates the collected data to the upper-level measurement and monitoring unit, and the upper-level measurement and monitoring unit performs preliminary processing on the aggregated data and transmits it to the smart energy unit through the multimodal communication module.

[0012] As a preferred solution of the multimodal communication edge aggregation gateway system for virtual power plants described in the present invention, the multimodal communication module further includes a security encryption submodule, which adopts corresponding encryption algorithms for different communication protocols for data transmission security; The adaptive protocol switching unit also includes a fault self-healing submodule, which automatically switches to a backup communication protocol when a primary communication link fails, and sends a fault report to a system administrator.

[0013] As a preferred solution of the multimodal communication edge aggregation gateway system for virtual power plants described in the present invention, the edge computing processor also includes a communication load prediction function, which uses a time series analysis algorithm to predict the load change trend of each communication protocol in the future period; The switching decision module executes protocol switching in advance before communication congestion occurs according to the prediction result of the communication load prediction function.

[0014] The beneficial effects of this preferred technical solution are: by predicting the load change trend of each communication protocol through the time series analysis algorithm, the switching decision module can execute protocol switching in advance before communication congestion occurs, avoiding communication interruption caused by passive switching and ensuring the continuity and stability of data transmission.

[0015] As a preferred solution of the multimodal communication edge aggregation gateway system for virtual power plants described in the present invention, wherein: the protocol evaluation module uses the Q-learning algorithm to continuously learn the communication protocol selection strategy under various communication scenarios; The Q-learning algorithm uses the current network status, service type, and historical switching effects as the state space, the communication protocol selection as the action space, and the degree of communication quality improvement as the reward function; The weight coefficients in the weighted scoring algorithm are dynamically optimized by the Q-learning algorithm. When it is predicted that a communication protocol is about to be congested, its corresponding weight coefficient is reduced for self-evolution of the communication protocol selection strategy.

[0016] The beneficial effects of this preferred technical solution are: by dynamically optimizing the weight coefficients in the weighted scoring algorithm, the self-evolution of the communication protocol selection strategy is achieved, thereby improving the intelligence level and adaptability of the protocol selection.

[0017] As a preferred solution of the multimodal communication edge aggregation gateway system for virtual power plants described in the present invention, the edge computing processor also includes an energy-aware scheduling function to monitor the real-time power consumption of each communication protocol and the edge computing processor; The data compression module adjusts the complexity of the compression algorithm according to the power consumption evaluation of the energy-aware scheduling function, and calculates the balance between power consumption and communication power consumption; The Q-learning algorithm takes energy efficiency as an additional reward factor.

[0018] The beneficial effects of this preferred technical solution are: the energy-aware scheduling function of the edge computing processor monitors the power consumption of each communication protocol and the edge computing processor in real time, and the data compression module adjusts the complexity of the compression algorithm accordingly to achieve the optimal balance between computing power consumption and communication power consumption, thereby reducing the overall energy consumption of the system.

[0019] As a preferred solution of the multimodal communication edge aggregation gateway system for virtual power plants described in the present invention, the switching decision module adopts a Byzantine fault-tolerant algorithm to make distributed decisions on the selection of communication protocols in the cascade network. When multiple measurement and monitoring units detect a protocol switching requirement, the communication protocol is determined through a distributed consensus mechanism. The communication quality detection module also has a network topology adaptive reconstruction function. Based on the network topology discovery and fault detection results, it dynamically reconstructs the network topology structure and automatically establishes an alternative communication path that bypasses the interference source when an interference source is detected; The protocol switching cache mechanism sets a distributed cache at each node of the cascade network to achieve seamless switching coordinated across the entire network.

[0020] The beneficial effects of this preferred technical solution are: the switching decision module adopts the Byzantine fault-tolerant algorithm for distributed decision-making, and cooperates with the network topology adaptive reconstruction function of the communication quality detection module and the distributed cache of the protocol switching cache mechanism to achieve high-reliability distributed collaborative communication and seamless switching coordinated across the entire network.

[0021] Compared with the prior art, the present invention has the following beneficial effects: By integrating a multimodal communication module, it supports multiple communication protocols such as VPDN, 5G, WiFi, LoRa and Zigbee. Combined with the communication quality detection module of the adaptive protocol switching unit, it monitors the delay, packet loss rate, signal strength and bandwidth utilization of each communication link in real time. The protocol evaluation module uses a weighted scoring algorithm to calculate the comprehensive performance score of each communication protocol. The switching decision module dynamically selects the communication protocol and performs the switching operation based on the comprehensive performance score, data priority and service type. This completely solves the problem that traditional communication systems using a single communication protocol cannot adapt to the communication requirements of different application scenarios and device types, and achieves the goal of dynamically adjusting communication strategies according to changes in the network environment, effectively avoiding the phenomenon of communication quality degradation or even interruption.

[0022] The communication load prediction function of the edge computing processor uses a time series analysis algorithm to predict the load change trend of each communication protocol in the future period. The protocol evaluation module uses the Q-learning algorithm to continuously learn the communication protocol selection strategy under various communication scenarios and dynamically optimize the weight coefficient in the weighted scoring algorithm. The switching decision module executes protocol switching in advance before communication congestion occurs based on the prediction results, building a fully automated intelligent protocol selection mechanism, completely getting rid of the dependence on manual configuration, improving the response speed, and fully meeting the virtual power plant's strict real-time requirements.

[0023] The present invention supports cascade deployment through measurement and monitoring units to form a tree-like network topology. The fault self-healing submodule automatically switches to the backup communication protocol when a failure occurs in the main communication link. The communication quality detection module has the network topology adaptive reconstruction function to dynamically reconstruct the network topology and establish a backup communication path that bypasses the interference source. The protocol switching cache mechanism sets a distributed cache at each node of the cascade network to achieve seamless switching coordinated by the entire network, build a highly reliable distributed communication network, ensure the stable operation of the virtual power plant in a complex network environment, and provide a solid communication guarantee for the real-time collection and coordinated control of large-scale distributed energy equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 This is a schematic diagram of the overall structure of a multimodal communication edge aggregation gateway system for a virtual power plant according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0027] Example 1, reference Figure 1 , as one embodiment of the present invention, provides a multimodal communication edge aggregation gateway system for virtual power plants, including: Smart energy unit, as an edge gateway device, is used for two-way communication with user branch loads, distributed energy resources, and flexible loads; At least one measurement and monitoring unit, installed in the user's branch load circuit or the user's side equipment end, used to collect branch line power consumption data and equipment operating conditions; a multimodal communication module, integrated into the smart energy unit, supporting multiple communication protocols, including at least two of VPDN, 5G, WiFi, LoRa, and Zigbee; Adaptive protocol switching unit, used to dynamically select communication protocols and switch between protocols based on communication environment parameters, device type and data transmission requirements; The measurement and monitoring unit communicates with the smart energy unit through the multimodal communication module, and the smart energy unit uploads the aggregated data to the virtual power plant platform or load management system through the selected communication protocol.

[0028] Specifically, the smart energy unit uses an embedded ARM architecture processor, has a Linux operating system, and supports the TCP / IP protocol stack. The smart energy unit is equipped with standard Ethernet, RS485, and wireless communication interfaces, and can connect to multiple measurement and monitoring units simultaneously.

[0029] The measurement and monitoring unit utilizes a high-precision energy metering chip, supports both three-phase four-wire and three-phase three-wire wiring, and achieves measurement accuracy of 0.2S. It also includes built-in temperature and humidity sensors to monitor the operating environment. Each unit has a unique device identification code and supports plug-and-play functionality.

[0030] The multimodal communication module integrates 5G, WiFi, LoRa, and Zigbee communication modules. The 5G module supports both SA and NSA dual-mode networking, with a peak downlink rate of up to 2Gbps. The WiFi module supports the 802.11n / ac standards, operating in the 2.4GHz and 5GHz frequency bands. The LoRa module operates in the 470-510MHz frequency band and has a communication range of up to 10 kilometers. The Zigbee module complies with the IEEE802.15.4 standard and supports mesh network topology.

[0031] The adaptive protocol switching unit includes a communication quality detection module, a protocol evaluation module, a switching decision module, a protocol switching cache mechanism, and a switching state synchronization module. The communication quality detection module collects parameters such as delay, packet loss rate, signal strength, and bandwidth utilization of each communication link once per second. The protocol evaluation module uses a weighted scoring algorithm, with a delay weight coefficient W1 = 0.3, a packet loss rate weight coefficient W2 = 0.3, a signal strength weight coefficient W3 = 0.2, and a bandwidth utilization weight coefficient W4 = 0.2.

[0032] In actual operation, the measurement and monitoring unit collects branch line voltage, current, power, energy, and other power consumption data on a per-second basis, as well as operating condition data such as device on / off status and fault information. The collected data undergoes a CRC check to ensure data integrity and is then transmitted to the smart energy unit via the multimodal communication module.

[0033] After receiving the data, the smart energy unit first verifies its validity and timestamp, then encapsulates it according to the data format requirements of the virtual power plant platform. The adaptive protocol switching unit monitors the current network status in real time and automatically switches to a communication protocol with better performance when it detects a degradation in network quality.

[0034] In this embodiment, when the system is first started, the 5G protocol is selected by default for communication. At a certain moment, the communication quality detection module detects that the 5G network signal strength drops to -85dBm and the packet loss rate rises to 2%. The calculation formula is: comprehensive performance score S = W1×(1-delay normalization value)+W2×(1-packet loss rate normalization value)+W3×signal strength normalization value+W4×(1-bandwidth utilization normalization value). The protocol evaluation module calculates that the comprehensive performance score of the 5G protocol is 0.65, while the comprehensive performance score of the WiFi protocol is 0.82. The switching decision module automatically switches the communication protocol to WiFi. The entire switching process is completed within 500 milliseconds to ensure uninterrupted data transmission.

[0035] After receiving the data uploaded by the smart energy unit, the virtual power plant platform or load management system performs load analysis and control strategy calculation, and sends control instructions to the smart energy unit to achieve coordinated control of distributed energy equipment.

[0036] This embodiment can effectively solve the problem of unstable communication of traditional single communication protocol systems in complex network environments, and improve the reliability and adaptability of the virtual power plant communication system.

[0037] In an embodiment of the present application, the smart energy unit further includes: Edge computing processors for local data preprocessing, load forecasting, and decision analysis; A data compression module is used to dynamically adjust the data compression rate and reporting frequency according to the bandwidth characteristics of the selected communication protocol; Device identification module, used to identify newly connected user-side devices and perform parameter configuration; User interaction terminal, with touch screen display function, used to display equipment operating status, response invitation information and profit estimation; The decision-making assistance module provides users with response strategy recommendations based on historical response data and current load conditions.

[0038] Specifically, the edge computing processor uses the ARM Cortex-A78 architecture, has a main frequency of 2.4GHz, and is equipped with 8GB of DDR4 memory and 128GB of eMMC storage space. The edge computing processor has a built-in neural network processing unit (NPU), a computing power of 4TOPS, and supports the TensorFlow and PyTorch deep learning frameworks. The processor filters, denoises, and detects outliers on the raw data collected by the measurement and monitoring unit, and the data preprocessing time is no more than 100 milliseconds. The load forecasting function uses the LSTM long short-term memory network algorithm to predict the load change trend for the next 24 hours based on 15 days of historical load data, with a prediction accuracy of over 95%.

[0039] The data compression module supports multiple compression algorithms, including LZ77, Huffman, and LZW. When using the 5G protocol, due to ample bandwidth, the data compression rate is set to 30%, and the reporting frequency is once every 5 seconds. When switching to the LoRa protocol, due to bandwidth limitations, the data compression rate is automatically adjusted to 70%, and the reporting frequency is reduced to once every 30 seconds. The data compression module performs graded compression based on data type and importance, using lossless compression for power data and lossy compression for environmental monitoring data to ensure the integrity of critical data.

[0040] The device identification module has an automatic scanning function, scanning for new devices on the network every 10 seconds. When a new device is detected, the module first reads the device's MAC address and model information, then queries the built-in device parameter database for a match. The database pre-stores the technical parameters of over 500 common user-side devices, including rated power, communication protocols, data formats, and other information. For devices not already in the database, the module supports manual addition of device information and parameter configuration, and the newly added device information is automatically synchronized to the cloud device database.

[0041] The user interaction terminal adopts a 10.1-inch capacitive touch screen with a resolution of 1920×1200 pixels and supports 10-point touch. The terminal interface adopts a graphical design, which displays key data such as the operating status, power consumption, and power generation of each device in real time. When the virtual power plant platform sends a demand response invitation, the user interaction terminal displays the invitation details in a pop-up window, including response time, response capacity, compensation price and other information. The profit estimation function calculates the expected profit of users participating in demand response in real time based on historical response data and current electricity prices, and the profit calculation accuracy reaches two decimal places.

[0042] The decision-making support module includes a built-in expert system that stores over 1,000 response strategy rules. This module analyzes the user's response history over the past 30 days, including metrics such as response success rate, response capacity, and user satisfaction. It then recommends the optimal response strategy based on current load conditions and device adjustability. For example, if the system predicts a 2kW peak load reduction response will be required between 2:00 PM and 4:00 PM tomorrow, the decision-making support module will recommend that the user shut down some non-essential loads, such as electric water heaters and washing machines, during this period, and will estimate compensation of 48 yuan.

[0043] In actual application, when a user installs a new 5kW distributed photovoltaic power generation system, the device identification module automatically detects the new device, identifies the model as "SolarMax-5000," and automatically matches the corresponding parameters from the device library and completes the configuration. The edge computing processor then begins real-time analysis of the photovoltaic system's power generation data, predicting power generation trends for the coming week. The user interface adds a monitoring window for the device to the main interface, displaying real-time power generation and cumulative power generation. When the virtual power plant platform sends a request for photovoltaic output adjustment, the decision-making support module recommends that the user adjust the inverter output power appropriately based on current sunlight conditions and user electricity demand, allowing the user to participate in grid regulation while ensuring their own electricity needs.

[0044] This embodiment significantly improves the intelligence level and user experience of the smart energy unit by integrating edge computing, intelligent compression, automatic identification, human-computer interaction, and decision-making assistance functions, providing strong technical support for the efficient operation of the virtual power plant.

[0045] The measurement and monitoring units support cascade deployment to form a tree-like network topology, wherein the lower-level measurement and monitoring units aggregate the collected data to the upper-level measurement and monitoring units, which perform preliminary processing on the aggregated data and then transmit it to the smart energy unit through the multimodal communication module.

[0046] Specifically, the tree-like network topology supports up to three cascade levels, with each upper-level measurement and monitoring unit connecting to up to 16 lower-level measurement and monitoring units. The first level is the root node measurement and monitoring unit, which communicates directly with the smart energy unit; the second level is the intermediate node measurement and monitoring unit, responsible for aggregating lower-level data and forwarding it to the upper level; the third level is the leaf node measurement and monitoring unit, responsible only for data collection and reporting. Each measurement and monitoring unit has a unique 16-bit device address, where the upper 8 bits represent the hierarchical and grouping information, and the lower 8 bits represent the device serial number.

[0047] The lower-level measurement and monitoring unit connects to the upper-level measurement and monitoring unit via an RS485 communication interface, with a communication rate of 9600bps and Modbus RTU protocol for data transmission. The lower-level measurement and monitoring unit collects local device data every 5 seconds, including parameters such as voltage, current, power, energy, temperature, and humidity, with data accuracy of 16-bit integers. After collection, the lower-level measurement and monitoring unit encapsulates the data into a standard Modbus data frame. The frame header contains information such as the device address, function code, and data length, and a CRC16 checksum is added to the frame footer to ensure data integrity.

[0048] The upper-level measurement and monitoring unit performs data aggregation and preliminary processing. It features a built-in ARM Cortex-M4 processor with a main frequency of 168MHz, 512KB of Flash storage, and 128KB of RAM. It collects data from lower-level devices using a polling method with a 500-millisecond polling cycle, ensuring timely access to the latest data from all lower-level devices. Upon receiving lower-level data, the upper-level measurement and monitoring unit first performs a CRC check to eliminate erroneous data, then timestamps and converts the valid data.

[0049] The preliminary data processing includes the following steps: first, data validity verification is performed to check whether the data is within a reasonable range, for example, the voltage value should be within the range of 380V±10%, and data outside the range is marked as abnormal; second, data deduplication is performed. When the data collected twice in a row are exactly the same, only one copy of the data is retained to reduce the transmission volume; then data compression is performed, using a differential encoding algorithm to only transmit the difference with the previous data, with a compression rate of up to 60%; finally, data packaging is performed to encapsulate all valid data at this level and the lower level into data packets in a unified format.

[0050] In a cascaded network, data transmission utilizes a hierarchical aggregation mechanism. Third-level leaf nodes report data to second-level intermediate nodes every 5 seconds. Second-level intermediate nodes aggregate data to first-level root nodes every 10 seconds. The first-level root nodes then upload aggregated data to the smart energy units every 15 seconds via a multimodal communication module. This hierarchical transmission mechanism effectively reduces network congestion and improves data transmission efficiency.

[0051] The measurement and monitoring unit also features automatic address allocation. When a new device joins the network, the system automatically assigns it an appropriate device address and hierarchical position. The new device first sends a network access request. Upon receiving the request, the higher-level device assigns an available address and responds with a confirmation message. Upon receiving the confirmation, the new device officially joins the network and begins operations.

[0052] To ensure network reliability, the cascaded network employs a redundant backup mechanism. If an intermediate node fails, downstream devices can aggregate data to other intermediate nodes via pre-defined backup paths. Network topology information is stored in non-volatile memory on each device, automatically restoring network connectivity after a power outage or restart.

[0053] In a real-world application scenario, an industrial park deployed 45 measurement and monitoring units using a three-level cascade structure: three root nodes at the first level, 12 intermediate nodes at the second level, and 30 leaf nodes at the third level. Leaf nodes are installed in the power distribution cabinets of each production equipment to monitor power usage in real time; intermediate nodes are installed in the power distribution rooms of each workshop to aggregate data from all equipment within the workshop; and the root node, installed in the park's main power distribution room, aggregates park-wide data and uploads it to the smart energy unit.

[0054] In an embodiment of the present application, the multimodal communication module further includes a security encryption submodule that uses corresponding encryption algorithms for different communication protocols for data transmission security; Specifically, the security encryption submodule uses corresponding encryption algorithms to ensure data transmission security based on the characteristics and security requirements of different communication protocols. For the 5G communication protocol, the AES-256-GCM encryption algorithm is used, with a key length of 256 bits, supporting integrity verification and identity authentication, and an encryption processing time of no more than 10 milliseconds. For the WiFi communication protocol, the WPA3-SAE security protocol is used, combined with the AES-128-CCM encryption algorithm to provide forward security. For the LoRa communication protocol, due to its low power consumption characteristics, the lightweight ChaCha20-Poly1305 encryption algorithm is used to ensure security while reducing power consumption. For the Zigbee communication protocol, the AES-128-CTR encryption mode is used, combined with the IEEE802.15.4 security framework, supporting dual encryption at the network layer and application layer.

[0055] The security encryption submodule incorporates a built-in hardware security module (HSM) and utilizes the nationally recognized SM series of algorithms as an alternative encryption scheme, including the SM2 elliptic curve public key algorithm, the SM3 cryptographic hash algorithm, and the SM4 block cipher algorithm, meeting domestic security requirements. Key management utilizes a hierarchical key system, comprising a three-tiered structure of root key, master key, and session key. The root key is stored in a hardware security chip with a physically tamper-resistant design; the master key is derived from the root key and used for device authentication; and the session key is dynamically generated with a 24-hour key update cycle, ensuring that even a single communication breach will not compromise overall security.

[0056] The adaptive protocol switching unit also includes a fault self-healing submodule, which automatically switches to a backup communication protocol when a primary communication link fails, and sends a fault report to a system administrator.

[0057] The fault self-healing submodule uses a multi-layer fault detection mechanism, including physical layer detection, network layer detection, and application layer detection. The physical layer detection monitors physical parameters such as signal strength and carrier-to-noise ratio. When the signal strength is less than -90dBm or the carrier-to-noise ratio is less than 10dB, it is determined to be a physical layer fault. The network layer detection monitors network connectivity by sending heartbeat packets at an interval of 10 seconds. If three consecutive heartbeat packets are lost, it is determined to be a network layer fault. The application layer detection monitors the data transmission success rate. When the data transmission success rate is less than 80% for one minute, it is determined to be an application layer fault.

[0058] The fault self-healing process is as follows: First, the fault self-healing submodule continuously monitors the working status of the current main communication protocol. When a fault at any level is detected, the fault self-healing process is immediately initiated; second, based on the preset protocol priority table and the current list of available protocols, the optimal backup communication protocol is selected. The protocol priority order is 5G>WiFi>LoRa>Zigbee; then, the protocol switching operation is executed. The entire switching process is completed within 500 milliseconds. During this period, the protocol switching cache mechanism ensures that data is not lost; finally, a fault report is sent to the virtual power plant platform and system administrator. The report content includes information such as fault time, fault type, fault protocol, post-switching protocol, and the number of affected devices.

[0059] In the embodiment of the present application, the protocol evaluation module uses the Q-learning algorithm to continuously learn the communication protocol selection strategy in each communication scenario; The Q-learning algorithm uses the current network status, service type, and historical switching effects as the state space, the communication protocol selection as the action space, and the degree of communication quality improvement as the reward function; The weight coefficients in the weighted scoring algorithm are dynamically optimized by the Q-learning algorithm. When it is predicted that a communication protocol is about to be congested, its corresponding weight coefficient is reduced for self-evolution of the communication protocol selection strategy.

[0060] Specifically, the state space Z of the Q-learning algorithm is defined as a triplet Z = (network state, service type, historical handover performance). Network state includes four dimensions: latency (low < 50ms, medium 50-200ms, high > 200ms), packet loss (low < 1%, medium 1-5%, high > 5%), signal strength (strong > -70dBm, medium -70 to -85dBm, weak < -85dBm), and bandwidth utilization (low < 30%, medium 30-70%, high > 70%). Service types include three categories: real-time control data, periodic reporting data, and historical query data. Historical handover performance includes three levels: excellent (performance improvement > 20% after handover), good (performance improvement 5-20% after handover), and fair (performance improvement < 5% after handover). Therefore, the total number of states in the state space is 4 × 4 × 4 × 4 × 3 × 3 = 2304.

[0061] The action space A is defined as a set of optional communication protocols , a total of 4 actions. The Q-learning algorithm maintains a Q-value table Q(s,a), which represents the expected reward of taking action a in state s. The Q-value table is initialized to a zero matrix of size 2304 × 4.

[0062] The reward function The design of is as follows: Assume that the comprehensive performance score before switching is , the comprehensive performance score after switching is , then the reward value The reward is positive when performance improves, and negative when performance degrades. Furthermore, considering the switching cost, a fixed reward of 5 is deducted for each protocol switch, encouraging the algorithm to switch only when necessary. If the system runs stably for more than 10 minutes after the switch, an additional 10 points are awarded to encourage long-term stable protocol selection.

[0063] The update formula of the Q-learning algorithm is: , where the learning rate α = 0.1, the discount factor γ = 0.9, the ε-greedy strategy is used for action selection, the exploration rate ε starts from 1.0, decays by 0.01 every 1000 iterations, and the minimum value is 0.05.

[0064] The dynamic optimization mechanism of the weight coefficients is as follows: the weight coefficients W1, W2, W3, and W4 in the traditional weighted scoring algorithm are dynamically adjusted through the Q-learning algorithm. , the Q-learning algorithm outputs the weight adjustment vector , the final weight coefficient , and normalized by the softmax function to ensure that the weight sum is 1. The weight adjustment range is limited to ±0.15 to prevent a certain weight from deviating too much from the reasonable value.

[0065] The congestion prediction mechanism uses a sliding window time series analysis with a 30-minute window length. The system calculates the load trend of each protocol every 5 minutes. If a protocol's bandwidth utilization has shown a linear increase with a slope greater than 1.5% / minute over the past 30 minutes, congestion is predicted for that protocol within the next 15 minutes. At this point, the Q-learning algorithm reduces the weight coefficient W4 corresponding to that protocol by 50% while increasing the weight coefficients of other protocols, guiding the system to switch to other protocols in advance.

[0066] In the algorithm implementation, the Q-learning module performs a policy update every 30 seconds. The specific process is as follows: Step 1: State perception: The protocol evaluation module obtains the current network parameters from the communication quality detection module and determines the current state s based on the current service type and historical switching records.

[0067] Step 2: Action selection, according to the ε-greedy strategy, randomly select an action with probability ε, and select the action with the largest Q value with probability 1-ε .

[0068] Step 3: Execute the action. If the selected protocol is different from the current protocol, perform the protocol switching operation.

[0069] Step 4: State transfer, observe the new state after switching , calculate the reward function .

[0070] Step 5: Update the Q value and adjust the Q(s,a) value according to the update formula.

[0071] Step 6: Weight adjustment: Calculate the weight adjustment vector based on the Q-value distribution and update the weight coefficient of the weighted scoring algorithm.

[0072] In a real-world application scenario, after six months of operation, the Q-learning algorithm in a virtual power plant system at an industrial park had acquired extensive strategic knowledge. During the weekday morning rush hour (8:00-9:00 AM), when a large number of devices simultaneously reporting data caused 5G network congestion, the algorithm automatically adjusted the weight of W4 from 0.2 to 0.1, reducing its focus on bandwidth utilization. It also adjusted the weight of W3 from 0.2 to 0.25, prioritizing Wi-Fi protocols with stable signal strength. This dynamic adjustment successfully avoided communication interruptions caused by network congestion, increasing the data transmission success rate from 85% to 96%.

[0073] After 12 months of continuous learning, the Q-learning algorithm identified 24 typical network scenarios, including weekday morning and evening peaks, weekend lows, and special holiday periods. The algorithm's strategy selection accuracy reached 93%, and the protocol switching frequency was reduced from an initial 15 times per hour to 3 times per hour, significantly improving system stability.

[0074] In an embodiment of the present application, the edge computing processor further includes an energy-aware scheduling function to monitor the real-time power consumption of each communication protocol and the edge computing processor; The data compression module adjusts the complexity of the compression algorithm according to the power consumption evaluation of the energy-aware scheduling function, and calculates the balance between power consumption and communication power consumption; The Q-learning algorithm takes energy efficiency as an additional reward factor.

[0075] Specifically, the energy-aware scheduling function integrates a high-precision power consumption monitoring circuit, using the INA3221 three-channel current / voltage monitoring chip, with a monitoring accuracy of ±0.1% and a sampling frequency of 1kHz. This function monitors the CPU power consumption, memory power consumption, storage power consumption, and power consumption of each communication protocol module of the edge computing processor 11 in real time. Among them, the 5G module has a transmit power consumption of 2.5W and a receive power consumption of 1.8W; the WiFi module has a transmit power consumption of 1.2W and a receive power consumption of 0.8W; the LoRa module has a transmit power consumption of 0.1W and a receive power consumption of 0.05W; the Zigbee module has a transmit power consumption of 0.03W and a receive power consumption of 0.02W. The edge computing processor consumes 8W when running at full load and 2W in standby mode.

[0076] The energy-aware scheduling function establishes a power consumption prediction model ,in The basic power consumption is 2W, is the CPU power consumption, which is related to the processor frequency f and utilization u, is the communication power consumption, which is related to the selected protocol p and data transmission volume b.

[0077] The data compression module supports four compression algorithms with different complexities: the simple compression algorithm uses RLE run-length encoding, with a CPU occupancy of 10%, a compression ratio of 2:1, and a processing delay of 5ms; the standard compression algorithm uses the LZ77 algorithm, with a CPU occupancy of 25%, a compression ratio of 4:1, and a processing delay of 15ms; the efficient compression algorithm uses the LZ4 algorithm, with a CPU occupancy of 40%, a compression ratio of 6:1, and a processing delay of 30ms; and the optimal compression algorithm uses the LZMA algorithm, with a CPU occupancy of 70%, a compression ratio of 10:1, and a processing delay of 80ms.

[0078] The data compression module dynamically selects the optimal compression algorithm based on the power consumption evaluation results of the energy-aware scheduling function. The power consumption balance decision formula is: ,in, To calculate energy consumption, is the communication energy consumption. The system goal is to minimize the total energy consumption , find the optimal balance between computing energy consumption and communication energy consumption.

[0079] The specific algorithm selection strategy is as follows: When the available power budget is greater than 6W and the communication bandwidth is less than 1Mbps, the optimal compression algorithm is selected to reduce the data transmission volume and communication energy consumption through a high compression ratio; When the available power budget is 3-6W and the communication bandwidth is 1-10Mbps, choose an efficient compression algorithm to strike a balance between compression effect and computational overhead; When the available power budget is between 1.5-3W and the communication bandwidth is greater than 10Mbps, choose a standard compression algorithm to avoid excessive computational overhead. When the available power budget is less than 1.5W or the system enters low power mode, a simple compression algorithm is selected to prioritize stable system operation.

[0080] The reward function of the Q-learning algorithm is expanded to: ,in, Reward for existing communication performance, Energy efficiency reward, λ is the energy consumption weight factor, and its value is 0.3. The calculation formula is: ,in, is the baseline energy consumption (energy consumption when using a fixed 5G protocol and standard compression algorithm), is the actual energy consumption. When the actual energy consumption is lower than the benchmark energy consumption, a positive reward is obtained; otherwise, a negative reward is obtained.

[0081] To balance performance and energy consumption, the Q-learning algorithm introduces the Energy Efficiency Ratio (EER), defined as EER = communication performance score / energy consumption per unit of data transmission. This shifts the algorithm's optimization goal from simply maximizing communication performance to maximizing the energy efficiency ratio, achieving a comprehensive optimization of both performance and energy consumption.

[0082] A new energy consumption state dimension has been added to the state space, including the current power consumption level (low <3W, medium 3-6W, high >6W) and the remaining battery level (full >80%, medium 50-80%, insufficient <50%). The expanded state space now has 2304 × 3 × 3 = 20736 states.

[0083] In actual application, a distributed photovoltaic power station in a remote area uses solar-powered smart energy units with a 100Ah battery capacity. On sunny days, the solar panels provide ample power, and the system selects a high-efficiency compression algorithm and 5G protocol to ensure data transmission quality. On cloudy days, when solar power is insufficient, the energy-aware scheduling function detects that the power budget has dropped to 2W, and the system automatically switches to a simpler compression algorithm and the LoRa protocol, extending device operation time. At night, the system enters a low-power mode, maintaining only the LoRa protocol for basic data transmission, reducing standby power consumption to 0.5W.

[0084] In the embodiment of the present application, the switching decision module adopts the Byzantine fault-tolerant algorithm to make distributed decisions on the selection of the communication protocol in the cascade network. When multiple measurement and monitoring units detect the need for protocol switching, the communication protocol is determined through a distributed consensus mechanism; The communication quality detection module also has a network topology adaptive reconstruction function. Based on the network topology discovery and fault detection results, it dynamically reconstructs the network topology structure and automatically establishes an alternative communication path that bypasses the interference source when an interference source is detected; The protocol switching cache mechanism sets a distributed cache at each node of the cascade network to achieve seamless switching coordinated across the entire network.

[0085] Specifically, the Byzantine Fault Tolerance algorithm uses the PBFT (Practical Byzantine Fault Tolerance) consensus mechanism, supporting consistent decisions even with a maximum of (n-1) / 3 faulty nodes in the network, where n is the total number of measurement and monitoring units involved in the decision-making process. Each measurement and monitoring unit serves as a decision-making node, fulfilling both the primary and backup roles. The primary node is responsible for initiating protocol switch proposals, while the backup nodes are responsible for verification and voting.

[0086] The PBFT algorithm's execution process consists of three phases: pre-prepare, prepare, and commit. In the pre-prepare phase, when a measurement and monitoring unit detects a protocol switch requirement, it broadcasts a switch proposal to other nodes in the network. The proposal includes the current timestamp, the proposed target protocol, the reason for the switch, and performance evaluation data. In the prepare phase, each node verifies the validity of the proposal, including timestamp validity, protocol availability, and switch rationality. Once verified, the prepare message is broadcast to the entire network. In the commit phase, when a node receives more than 2f+1 valid prepare messages (f is the maximum number of fault-tolerant nodes), it broadcasts a commit message to the entire network. When more than 2f+1 commit messages are received, the protocol switch decision officially takes effect.

[0087] The distributed consensus mechanism sets a decision timeout of 30 seconds. If consensus cannot be reached within the timeout, the view change process is triggered, and a new master node is elected to re-initiate the decision. To avoid frequent protocol switching, the minimum switching interval is set to 2 minutes. That is, proposals that are less than 2 minutes after the last switch will be automatically rejected.

[0088] The network topology adaptive reconstruction function uses a distributed topology discovery algorithm. Each measurement and monitoring unit periodically sends topology probe packets containing information such as the source node address, destination node address, timestamp, and hop counter. The probe packets are sent every 60 seconds, with a TTL of 10 hops. By collecting responses to the probe packets, each node constructs a local view of the network topology, including neighbor lists, path quality information, link latency, and other parameters.

[0089] The network topology reconstruction algorithm is as follows: First, each node periodically exchanges topology information to construct a global network topology graph; second, the Dijkstra algorithm is used to calculate the shortest path, while considering the link quality weight. In this calculation formula, the weight function is ,in ; Then, identify the key nodes and bottleneck links in the network. Key nodes are defined as nodes with a connectivity greater than or equal to 3 and carrying traffic exceeding 50% of the network average; finally, when a key node failure or bottleneck link congestion is detected, the routing path is automatically recalculated to achieve dynamic reconstruction of the network topology.

[0090] The interference source detection mechanism uses a combination of signal quality monitoring and spectrum analysis. When the signal strength of a communication link drops by more than 10dB within 5 minutes, or the bit error rate exceeds When a source of interference is detected, the system automatically initiates a spectrum scan to identify the interfering frequency band and type, then calculates an alternate communication path that bypasses the source of interference. The alternate path is prioritized for the path with the fewest hops and lowest total latency, ensuring that the overlap between the alternate path and the primary path does not exceed 50%, preventing the same interference source from affecting both the primary and backup paths.

[0091] The distributed cache uses a consistent hashing algorithm for data distribution, distributing data to different measurement and monitoring units based on hash values. Each node maintains a local cache and a replica cache, with the number of replicas set to three to ensure data is not lost even if two nodes fail simultaneously. The cached data includes service data to be transmitted, protocol switching status information, network topology data, and more.

[0092] The data synchronization mechanism of a distributed cache is as follows: When a node receives new data, it first stores it in its local cache. It then uses a consistent hashing algorithm to determine a replica storage node and sends a data synchronization request to the replica node. Upon receiving the synchronization request, the replica node verifies the data integrity, stores the replica data, and then returns a confirmation message to the source node. After receiving confirmation messages from all replica nodes, the source node marks the data as synchronized. The entire synchronization process is asynchronous, without blocking the normal transmission of business data.

[0093] The implementation mechanism of seamless switching is as follows: before the protocol switching begins, each node temporarily stores the data currently being transmitted in the distributed cache, including information such as data content, transmission progress, and target address; during the switching process, after the new protocol is started, each node reads the unfinished transmission tasks from the distributed cache and continues to transmit using the new protocol; to ensure the timeliness of the data, the data in the cache is sorted by timestamp, and data with an earlier transmission time is transmitted first; after the switching is completed, each node clears the cached data that has been transmitted and releases storage space.

[0094] In a real-world application scenario, a large industrial park deployed a three-level cascaded network consisting of 45 measurement and monitoring units. During a thunderstorm, multiple LoRa communication modules located outdoors were subjected to strong electromagnetic interference, causing a sharp decline in communication quality. Twelve nodes in the network simultaneously detected an anomaly in LoRa protocol performance, triggering a distributed decision-making process.

[0095] The decision-making process is as follows: Node A, acting as the master node, initiates a proposal to switch from LoRa to WiFi. The other 44 nodes in the network complete the proposal verification within 10 seconds, of which 42 nodes vote in favor, and 2 nodes do not respond due to communication interruption. Since the number of votes in favor (42) is greater than 2×14+1=29 (the network can tolerate a maximum of 14 faulty nodes), the decision is passed. The entire network completes the protocol switch within 15 seconds, during which the distributed caching mechanism ensures the seamless transmission of 1,847 business data items.

[0096] After the switchover, the network's adaptive topology reconstruction function detected that the original LoRa path was unavailable, automatically calculated three backup WiFi-based paths, and updated the routing table in the topology database. When the thunderstorm ended and the LoRa signal returned to normal, the system switched back to the LoRa protocol through distributed decision-making, achieving autonomous network adaptation and optimization.

[0097] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A multimodal communication edge aggregation gateway system for virtual power plants, characterized in that: include: Smart energy unit, as an edge gateway device, is used for two-way communication with user branch loads, distributed energy resources, and flexible loads; At least one measurement and monitoring unit, installed in the user's branch load circuit or the user's side equipment end, used to collect branch line power consumption data and equipment operating conditions; a multimodal communication module, integrated into the smart energy unit, supporting multiple communication protocols, including at least two of VPDN, 5G, WiFi, LoRa, and Zigbee; Adaptive protocol switching unit, used to dynamically select communication protocols and implement switching between protocols based on communication environment parameters, device type and data transmission requirements; The measurement and monitoring unit communicates with the smart energy unit through the multimodal communication module, and the smart energy unit uploads the aggregated data to the virtual power plant platform or load management system through the selected communication protocol.

2. The multimodal communication edge aggregation gateway system for virtual power plants according to claim 1, characterized in that: The adaptive protocol switching unit includes: Communication quality detection module, used to monitor the delay, packet loss rate, signal strength, bandwidth utilization of each communication link in real time, and identify network topology and interference sources; A protocol evaluation module, configured to calculate a comprehensive performance score for each communication protocol based on the monitoring results of the communication quality detection module; A switching decision module, configured to select a communication protocol and perform a switching operation based on the comprehensive performance score, data priority, and service type; The protocol switching cache mechanism is used to temporarily store data to be transmitted during the protocol switching process to ensure the continuity of data transmission; The switching status synchronization module is used to report the current communication protocol status to the virtual power plant platform in real time.

3. The multimodal communication edge aggregation gateway system for virtual power plants according to claim 2, characterized in that: The protocol evaluation module adopts a weighted scoring algorithm, where: The delay weight coefficient is W1, the packet loss rate weight coefficient is W2, the signal strength weight coefficient is W3, and the bandwidth utilization weight coefficient is W4; The comprehensive performance score S = W1×(1-normalized delay value)+W2×(1-normalized packet loss rate value)+W3×normalized signal strength value+W4×(1-normalized bandwidth utilization value).

4. The multimodal communication edge aggregation gateway system for virtual power plants according to claim 3, characterized in that: The smart energy unit also includes: Edge computing processors for local data preprocessing, load forecasting, and decision analysis; A data compression module is used to dynamically adjust the data compression rate and reporting frequency according to the bandwidth characteristics of the selected communication protocol; Device identification module, used to identify newly connected user-side devices and perform parameter configuration; User interaction terminal, with touch screen display function, used to display equipment operating status, response invitation information and profit estimation; The decision-making assistance module provides users with response strategy recommendations based on historical response data and current load conditions.

5. The multimodal communication edge aggregation gateway system for virtual power plants according to claim 4, characterized in that: The measurement and monitoring units support cascade deployment to form a tree-like network topology, wherein the lower-level measurement and monitoring units aggregate the collected data to the upper-level measurement and monitoring units, which perform preliminary processing on the aggregated data and then transmit it to the smart energy unit through the multimodal communication module.

6. The multimodal communication edge aggregation gateway system for virtual power plants according to claim 5, characterized in that: The multimodal communication module also includes a security encryption submodule that uses corresponding encryption algorithms for different communication protocols to ensure data transmission security; The adaptive protocol switching unit also includes a fault self-healing submodule, which automatically switches to a backup communication protocol when a primary communication link fails, and sends a fault report to a system administrator.

7. The multimodal communication edge aggregation gateway system for virtual power plants according to claim 6, characterized in that: The edge computing processor also includes a communication load prediction function, which uses a time series analysis algorithm to predict the load change trend of each communication protocol in the future period; The switching decision module executes protocol switching in advance before communication congestion occurs according to the prediction result of the communication load prediction function.

8. The multimodal communication edge aggregation gateway system for virtual power plants according to claim 7, characterized in that: The protocol evaluation module uses the Q-learning algorithm to continuously learn the communication protocol selection strategy under various communication scenarios; The Q-learning algorithm uses the current network status, service type, and historical switching effects as the state space, the communication protocol selection as the action space, and the degree of communication quality improvement as the reward function; The weight coefficients in the weighted scoring algorithm are dynamically optimized by the Q-learning algorithm. When it is predicted that a communication protocol is about to be congested, its corresponding weight coefficient is reduced for self-evolution of the communication protocol selection strategy.

9. The multimodal communication edge aggregation gateway system for virtual power plants according to claim 8, characterized in that: The edge computing processor also includes an energy-aware scheduling function to monitor the real-time power consumption of each communication protocol and the edge computing processor; The data compression module adjusts the complexity of the compression algorithm according to the power consumption evaluation of the energy-aware scheduling function, and calculates the balance between power consumption and communication power consumption; The Q-learning algorithm takes energy efficiency as an additional reward factor.

10. The multimodal communication edge aggregation gateway system for virtual power plants according to claim 9, characterized in that: The switching decision module uses a Byzantine fault-tolerant algorithm in the cascade network to make distributed decisions on the selection of the communication protocol. When multiple measurement and monitoring units detect the need for protocol switching, the communication protocol is determined through a distributed consensus mechanism. The communication quality detection module also has a network topology adaptive reconstruction function. Based on the network topology discovery and fault detection results, it dynamically reconstructs the network topology structure and automatically establishes an alternative communication path that bypasses the interference source when an interference source is detected; The protocol switching cache mechanism sets a distributed cache at each node of the cascade network to achieve seamless switching coordinated across the entire network.

Images