Charging pile ad hoc network method and system

Abstract

The invention provides a charging pile ad hoc network method and system in the technical field of charging pile networking, and the method comprises the steps: S1, enabling a first charging pile to serve as a main node to be connected with a server, enabling a subsequent charging pile to serve as a child node, carrying out the authentication with the main node, and enabling the subsequent charging pile to be connected with a wireless Mesh network managed by the main node; s2, scanning a wireless environment by each charging pile to construct a signal attenuation thermodynamic diagram, and re-electing a main node based on the signal attenuation thermodynamic diagram and the node state data; s3, when the main node and the sub-nodes are disconnected, each sub-node re-elects the main node based on the signal attenuation thermodynamic diagram and the node state data; and S4, when the main node is disconnected with the server, dynamically selecting the sub-node to communicate with the server through a multi-hop relay mode, or uploading a network disconnection fault through the mobile terminal. The method has the advantages that the problem of collaborative optimization of reliability, real-time performance and safety is solved while low-cost deployment and high expansibility are guaranteed, and large-scale landing of a distributed charging scene is supported.

Description

Technical Field

[0001] This invention relates to the field of charging pile networking technology, and in particular to a method and system for self-organizing charging pile networks. Background Technology

[0002] Charging piles (also known as electric vehicle power supply equipment, EVSE) are the core infrastructure for electric vehicle energy replenishment, and their large-scale deployment and efficient operation and maintenance have become key supports for industry development. With the surge in electric vehicle penetration, charging pile deployment is rapidly expanding from centralized charging stations to distributed scenarios such as underground parking garages, shopping mall parking lots, and community streets. To reduce networking costs and improve operational efficiency, self-organizing networking technology for charging pile clusters has emerged—by constructing a local area network (LAN), with the master node uniformly connecting to the external network, it achieves data aggregation and interaction with the cloud platform. However, existing networking methods have systemic defects, mainly reflected in the following dimensions:

[0003] 1. The cabling project is complex and has poor scalability:

[0004] Early solutions relied on wired Ethernet (RJ45 / fiber optic) or industrial buses (RS485 / CAN), requiring independent power and communication lines to be laid for each device. For example, a 10-channel centralized charging station required more than 20 cables, leading to high construction costs and repeated excavation work during expansion. Especially in underground scenarios, the shielding effect of metal structures on wireless signals (attenuation of 30–50dB) forced some manufacturers to adopt repeater stacking solutions, which actually increased the risk of faulty nodes.

[0005] 2. Network blind spots cause service interruptions:

[0006] In remote parking lots or underground levels without 4G signal, over 60% of charging piles lose their remote start / stop, payment, and fault diagnosis capabilities due to the inability to connect to the cloud platform, forcing users to operate on-site, which deviates from the trend of "unmanned operation and maintenance." Even more serious is the fact that a single point of failure in the main node (such as a power outage or damage to the communication module) can lead to data interruption across the entire local area network, and there is a lack of a master-slave switching mechanism (such as a hot backup node) to achieve self-healing from faults.

[0007] 3. Weak security mechanisms increase the risk of attacks:

[0008] Static password authentication is easily cracked. Charging stations often use default factory keys and do not introduce device fingerprinting or adaptive authentication mechanisms. Attackers can even inject malicious nodes in the local area network to steal user payment information or launch DDoS attacks.

[0009] To address the above issues, the industry has proposed the following improvement strategies, but significant bottlenecks still exist:

[0010] 1. Wireless Mesh networking (such as LoRa / WiFi Mesh): Although coverage is extended through multi-hop relays, the power consumption of long-distance nodes increases dramatically due to the lack of optimized path selection algorithms; and the node deployment ignores the signal heat map (such as failing to avoid reinforced concrete load-bearing walls), so the actual throughput only reaches 30% of the theoretical value.

[0011] 2. Hybrid communication solution (ZigBee + 4G DTU): Although it solves the network problem in underground scenarios, the unit price of DTU equipment is high and the monthly fee for SIM cards needs to be paid continuously, which greatly increases the cost of large-scale deployment; when all data is uploaded through a single DTU, the single-point bandwidth bottleneck restricts the cluster size (maximum of about 20 units).

[0012] 3. Basic heartbeat detection mechanism: Traditional heartbeat packets are only used for survival detection and do not integrate device fingerprint information, so they cannot identify "zombie node" forgery.

[0013] This demonstrates that traditional methods present irreconcilable contradictions in three dimensions: network reliability, dynamic resource scheduling, and security protection.

[0014] 1. Reliability vs. Cost: Primary / backup switching requires redundant hardware, which increases equipment costs;

[0015] 2. Real-time performance vs. coverage: Low latency requirements limit single-hop distance, sacrificing edge coverage;

[0016] 3. Security vs. User Experience: Strong authentication increases the number of steps required and reduces user satisfaction.

[0017] Therefore, how to provide a self-organizing network method and system for charging piles, while ensuring low-cost deployment and high scalability, and solving the problem of coordinated optimization of reliability, real-time performance and security, so as to support the large-scale implementation of distributed charging scenarios, has become an urgent technical problem to be solved. Summary of the Invention

[0018] The technical problem to be solved by this invention is to provide a method and system for self-organizing charging piles, which can solve the problem of collaborative optimization of reliability, real-time performance and security while ensuring low-cost deployment and high scalability, so as to support the large-scale implementation of distributed charging scenarios.

[0019] In a first aspect, the present invention provides a method for self-organizing a charging pile network, comprising the following steps:

[0020] Step S1: Each charging pile is pre-set with a unique device fingerprint and a pre-trained channel allocation model.

[0021] Step S2: After the first charging pile is powered on and started, it performs initialization operations and establishes a connection with the server as the master node.

[0022] Step S3: After the charging pile is powered on and started, it performs an initialization operation and, as a child node, performs an authentication operation with the main node through the device fingerprint before connecting to the wireless Mesh network managed by the main node.

[0023] Step S4: Each charging pile in the wireless local area network periodically scans the wireless environment to construct a signal attenuation heat map, and re-elects a master node based on the signal attenuation heat map and node status data.

[0024] Step S5: The master node and each child node maintain a connection through multipath heartbeat packets. When the master node loses connection with each child node, each child node re-elects a master node based on the signal attenuation heatmap and node status data.

[0025] Step S6: When the master node disconnects from the server, the child node is dynamically selected to communicate with the server through a multi-hop relay mode, or the network disconnection fault is uploaded through an encrypted tunnel via a mobile terminal connected to the master node.

[0026] Step S7: The master node monitors the real-time operating data of each sub-node in real time, inputs the real-time operating data into the channel allocation model to obtain the real-time channel allocation strategy, and performs load balancing on the network of each sub-node based on the real-time channel allocation strategy.

[0027] Furthermore, in step S1, the device fingerprint generation process specifically involves: normalizing and concatenating the device serial number, MAC address, and manufacturing date to obtain concatenated data; performing hash calculation on the concatenated data using SHA-256 to obtain a first hash value; and extracting bits 2-129 from the first hash value as the device fingerprint.

[0028] The channel allocation model is used to output a channel allocation strategy based on the running data, and is constructed based on a multi-source feature extraction layer, a dynamic decision layer, and a strategy optimization layer.

[0029] The multi-source feature extraction layer is constructed based on a wireless environment feature extraction module, a node state feature extraction module, and a user behavior feature extraction module. The wireless environment feature extraction module is used to extract spatial interference features from wireless environment data through a convolutional neural network combined with a multi-head attention mechanism. The node state feature extraction module is used to extract node load features from node state data through a gated recurrent unit. The user behavior feature extraction module is used to extract user behavior pattern features from user behavior data through a clustering algorithm and a fully connected layer.

[0030] The dynamic decision-making layer is constructed based on a feature fusion module and a policy generation module. The feature fusion module is used to fuse and concatenate spatial interference features, node load features, and user behavior pattern features through a graph neural network to obtain a comprehensive feature vector. The policy generation module is used to reason about the comprehensive feature vector through a deep reinforcement learning agent and output a triplet policy including a channel allocation plan, a load balancing instruction, and a network optimization rule.

[0031] The strategy optimization layer is constructed based on a conflict detection module and an adaptive optimization module; the conflict detection module is used to verify the channel allocation plan based on interference-free constraints; the adaptive optimization module is used to optimize the triplet strategy through an online learner and output a channel allocation strategy carrying a channel allocation plan, load balancing instructions and network optimization rules.

[0032] The operational data includes at least wireless environment data, node status data, and user behavior data.

[0033] The channel allocation strategy includes at least a channel allocation plan, load balancing instructions, and network optimization rules.

[0034] Furthermore, step S2 specifically includes:

[0035] After the first charging pile is powered on and started, it automatically performs initialization operations including at least hardware self-test and status confirmation, software system initialization, communication module initialization, and charging interface pre-test. It obtains the current timestamp as the request time, obtains the device serial number of the device, encrypts the device fingerprint, device serial number and request time into ciphertext information using the SM9 algorithm, calculates the second hash value of the ciphertext information using the SM3 algorithm, generates a connection request based on the ciphertext information and the second hash value, and uploads the connection request to the server through the 5G communication unit of the wireless communication module as the master node.

[0036] The server parses the received connection request to obtain encrypted information and a second hash value. After verifying the integrity of the encrypted information using the second hash value, it decrypts the encrypted information using the SM9 algorithm to obtain the device fingerprint, device serial number, and request time. After verifying the timeliness of the request, it matches the corresponding MAC address and manufacturing date from a preset device management table using the device serial number. After verifying the legality of the device fingerprint based on the device serial number, MAC address, and manufacturing date, it establishes a connection with the charging pile.

[0037] Furthermore, step S3 specifically includes:

[0038] After the charging pile is powered on and started, it automatically performs initialization operations and, as a child node, creates a pair of public and private keys based on the RSA algorithm, generates a first random number, obtains the current timestamp as the first authentication time, and uses the device fingerprint of the local machine as the sub-fingerprint.

[0039] The child node encrypts the sub-fingerprint, the first random number, and the first authentication time using a private key to obtain ciphertext data A1. It then encrypts the ciphertext data A1 and the public key using the SM9 algorithm to obtain ciphertext data A2. Finally, it shifts each character of the ciphertext data A2 cyclically to the right by 3 bits to obtain ciphertext data A3. Based on the ciphertext data A3, it generates the first authentication request and sends it to the master node.

[0040] The master node parses the received first authentication request to obtain ciphertext data A3, shifts each character of ciphertext data A3 to the left by 3 bits to obtain ciphertext data A2, decrypts ciphertext data A2 into ciphertext data A1 and public key using the SM9 algorithm, decrypts ciphertext data A1 with the public key to obtain sub-fingerprint, first random number and first authentication time, performs timeliness verification using the first authentication time, adds 3 to the value of the first random number to obtain a second random number, obtains the current timestamp as the second authentication time, and uses the device fingerprint of the local machine as the master fingerprint;

[0041] The master node encrypts the master fingerprint, sub-fingerprint, second random number, and second authentication time using the public key to obtain ciphertext data B1. It then encrypts the ciphertext data B1 into ciphertext data B2 using the SM9 algorithm. Finally, it shifts each character of the ciphertext data B2 cyclically to the right by 5 bits to obtain ciphertext data B3. Based on the ciphertext data B3, it generates a second authentication request and sends it to the sub-node.

[0042] The child node parses the received second authentication request to obtain ciphertext data B3, shifts each character of ciphertext data B3 cyclically 5 bits to the left to obtain ciphertext data B2, decrypts ciphertext data B2 into ciphertext data B1 using the SM9 algorithm, decrypts ciphertext data B1 using the private key to obtain the master fingerprint, sub-fingerprint, second random number, and second authentication time, performs timeliness verification using the second authentication time, verifies the second random number using the first random number, verifies the decrypted sub-fingerprint using the sub-fingerprint of the local machine, adds 5 to the value of the second random number to obtain the third random number, and obtains the current timestamp as the third authentication time;

[0043] The child node encrypts the master fingerprint, the third random number, and the third authentication time using a private key to obtain ciphertext data A4. It then encrypts the ciphertext data A4 into ciphertext data A5 using the SM9 algorithm. Finally, it shifts each character of the ciphertext data A5 cyclically to the right by 3 bits to obtain ciphertext data A6. Based on the ciphertext data A6, it generates a third authentication request and sends it to the master node.

[0044] The master node parses the received third authentication request to obtain ciphertext data A6. It then shifts each character of ciphertext data A6 cyclically 3 bits to the left to obtain ciphertext data A5. Using the SM9 algorithm, it decrypts ciphertext data A5 into ciphertext data A4. The master node then decrypts ciphertext data A4 using the public key to obtain the master fingerprint, the third random number, and the third authentication time. After verifying the timeliness of the third authentication time, it verifies the third random number using the second random number. Finally, it verifies the decrypted master fingerprint using the master fingerprint of the local machine. If the verification passes, the bidirectional authentication operation is completed, and the node accesses the wireless Mesh network managed by the master node through the Wi-SUN communication unit of the wireless communication module.

[0045] Furthermore, step S4 specifically includes:

[0046] Each charging station within the wireless local area network periodically scans the target signal strength and interference signal strength in the wireless environment, broadcasts the target signal strength, interference signal strength, and the charging station's location within the wireless local area network, and constructs a signal attenuation heatmap based on the target signal strength, interference signal strength, and charging station location of each charging station; the target signal is a 5G signal.

[0047] Each charging station will broadcast its node status data, including at least network load indicators, device operating status, and location and topology data, within the wireless local area network;

[0048] Based on the signal attenuation heatmap and node status data, each charging pile re-elects a master node using a weighted scoring algorithm;

[0049] Step S5 specifically involves:

[0050] The master node and each child node maintain a connection through a multi-path heartbeat packet, which embeds a device fingerprint and a time-series encrypted hash value for real-time node authentication.

[0051] When the disconnection between the master node and each child node exceeds a preset time threshold, each child node re-elects a master node based on the signal attenuation heatmap and node status data.

[0052] Step S6 specifically involves:

[0053] When the master node disconnects from the server, based on the signal attenuation heatmap and node status data, it dynamically selects a child node to communicate with the server through a multi-hop relay mode, or establishes an end-to-end encrypted tunnel to upload the network outage fault to the server through a mobile terminal connected to the master node via Bluetooth.

[0054] Secondly, the present invention provides a self-organizing network system for charging piles, comprising the following modules:

[0055] The initial setup module is used to pre-configure a unique device fingerprint and a pre-trained channel allocation model for each charging pile.

[0056] The master node connection module is used to perform initialization operations after the first charging pile is powered on and started, and to establish a connection with the server as the master node.

[0057] The wireless mesh networking module is used to perform initialization operations after the charging pile is powered on and started, and as a child node, it performs authentication operations with the main node through the device fingerprint and then accesses the wireless mesh network managed by the main node.

[0058] The first master node election module is used to periodically scan the wireless environment of each charging pile in the wireless local area network to construct a signal attenuation heat map, and re-elect the master node based on the signal attenuation heat map and node status data.

[0059] The second master node election module is used to maintain the connection between the master node and each child node through multi-path heartbeat packets. When the master node loses connection with each child node, each child node re-elects the master node based on the signal attenuation heatmap and node status data.

[0060] The network outage fault upload module is used to dynamically select a child node to communicate with the server through a multi-hop relay mode when the master node loses connection with the server, or to upload the network outage fault through a mobile terminal connected to the master node in an encrypted tunnel.

[0061] The load balancing module is used by the master node to monitor the real-time operating data of each sub-node, input the real-time operating data into the channel allocation model to obtain the real-time channel allocation strategy, and perform load balancing on the network of each sub-node based on the real-time channel allocation strategy.

[0062] Furthermore, in the initial setup module, the device fingerprint generation process is as follows: the device serial number, MAC address, and manufacturing date are normalized and concatenated to obtain concatenated data; the concatenated data is hashed using SHA-256 to obtain a first hash value; and bits 2-129 are extracted from the first hash value as the device fingerprint.

[0063] The channel allocation model is used to output a channel allocation strategy based on the running data, and is constructed based on a multi-source feature extraction layer, a dynamic decision layer, and a strategy optimization layer.

[0064] The multi-source feature extraction layer is constructed based on a wireless environment feature extraction module, a node state feature extraction module, and a user behavior feature extraction module. The wireless environment feature extraction module is used to extract spatial interference features from wireless environment data through a convolutional neural network combined with a multi-head attention mechanism. The node state feature extraction module is used to extract node load features from node state data through a gated recurrent unit. The user behavior feature extraction module is used to extract user behavior pattern features from user behavior data through a clustering algorithm and a fully connected layer.

[0065] The dynamic decision-making layer is constructed based on a feature fusion module and a policy generation module. The feature fusion module is used to fuse and concatenate spatial interference features, node load features, and user behavior pattern features through a graph neural network to obtain a comprehensive feature vector. The policy generation module is used to reason about the comprehensive feature vector through a deep reinforcement learning agent and output a triplet policy including a channel allocation plan, a load balancing instruction, and a network optimization rule.

[0066] The strategy optimization layer is constructed based on a conflict detection module and an adaptive optimization module; the conflict detection module is used to verify the channel allocation plan based on interference-free constraints; the adaptive optimization module is used to optimize the triplet strategy through an online learner and output a channel allocation strategy carrying a channel allocation plan, load balancing instructions and network optimization rules.

[0067] The operational data includes at least wireless environment data, node status data, and user behavior data.

[0068] The channel allocation strategy includes at least a channel allocation plan, load balancing instructions, and network optimization rules.

[0069] Furthermore, the master node connection module is specifically used for:

[0070] After the first charging pile is powered on and started, it automatically performs initialization operations including at least hardware self-test and status confirmation, software system initialization, communication module initialization, and charging interface pre-test. It obtains the current timestamp as the request time, obtains the device serial number of the device, encrypts the device fingerprint, device serial number and request time into ciphertext information using the SM9 algorithm, calculates the second hash value of the ciphertext information using the SM3 algorithm, generates a connection request based on the ciphertext information and the second hash value, and uploads the connection request to the server through the 5G communication unit of the wireless communication module as the master node.

[0071] The server parses the received connection request to obtain encrypted information and a second hash value. After verifying the integrity of the encrypted information using the second hash value, it decrypts the encrypted information using the SM9 algorithm to obtain the device fingerprint, device serial number, and request time. After verifying the timeliness of the request, it matches the corresponding MAC address and manufacturing date from a preset device management table using the device serial number. After verifying the legality of the device fingerprint based on the device serial number, MAC address, and manufacturing date, it establishes a connection with the charging pile.

[0072] Furthermore, the wireless mesh networking module is specifically used for:

[0073] After the charging pile is powered on and started, it automatically performs initialization operations and, as a child node, creates a pair of public and private keys based on the RSA algorithm, generates a first random number, obtains the current timestamp as the first authentication time, and uses the device fingerprint of the local machine as the sub-fingerprint.

[0074] The child node encrypts the sub-fingerprint, the first random number, and the first authentication time using a private key to obtain ciphertext data A1. It then encrypts the ciphertext data A1 and the public key using the SM9 algorithm to obtain ciphertext data A2. Finally, it shifts each character of the ciphertext data A2 cyclically to the right by 3 bits to obtain ciphertext data A3. Based on the ciphertext data A3, it generates the first authentication request and sends it to the master node.

[0075] The master node parses the received first authentication request to obtain ciphertext data A3, shifts each character of ciphertext data A3 to the left by 3 bits to obtain ciphertext data A2, decrypts ciphertext data A2 into ciphertext data A1 and public key using the SM9 algorithm, decrypts ciphertext data A1 with the public key to obtain sub-fingerprint, first random number and first authentication time, performs timeliness verification using the first authentication time, adds 3 to the value of the first random number to obtain a second random number, obtains the current timestamp as the second authentication time, and uses the device fingerprint of the local machine as the master fingerprint;

[0076] The master node encrypts the master fingerprint, sub-fingerprint, second random number, and second authentication time using the public key to obtain ciphertext data B1. It then encrypts the ciphertext data B1 into ciphertext data B2 using the SM9 algorithm. Finally, it shifts each character of the ciphertext data B2 cyclically to the right by 5 bits to obtain ciphertext data B3. Based on the ciphertext data B3, it generates a second authentication request and sends it to the sub-node.

[0077] The child node parses the received second authentication request to obtain ciphertext data B3, shifts each character of ciphertext data B3 cyclically 5 bits to the left to obtain ciphertext data B2, decrypts ciphertext data B2 into ciphertext data B1 using the SM9 algorithm, decrypts ciphertext data B1 using the private key to obtain the master fingerprint, sub-fingerprint, second random number, and second authentication time, performs timeliness verification using the second authentication time, verifies the second random number using the first random number, verifies the decrypted sub-fingerprint using the sub-fingerprint of the local machine, adds 5 to the value of the second random number to obtain the third random number, and obtains the current timestamp as the third authentication time;

[0078] The child node encrypts the master fingerprint, the third random number, and the third authentication time using a private key to obtain ciphertext data A4. It then encrypts the ciphertext data A4 into ciphertext data A5 using the SM9 algorithm. Finally, it shifts each character of the ciphertext data A5 cyclically to the right by 3 bits to obtain ciphertext data A6. Based on the ciphertext data A6, it generates a third authentication request and sends it to the master node.

[0079] The master node parses the received third authentication request to obtain ciphertext data A6. It then shifts each character of ciphertext data A6 cyclically 3 bits to the left to obtain ciphertext data A5. Using the SM9 algorithm, it decrypts ciphertext data A5 into ciphertext data A4. The master node then decrypts ciphertext data A4 using the public key to obtain the master fingerprint, the third random number, and the third authentication time. After verifying the timeliness of the third authentication time, it verifies the third random number using the second random number. Finally, it verifies the decrypted master fingerprint using the master fingerprint of the local machine. If the verification passes, the bidirectional authentication operation is completed, and the node accesses the wireless Mesh network managed by the master node through the Wi-SUN communication unit of the wireless communication module.

[0080] Furthermore, the first master node election module is specifically used for:

[0081] Each charging station within the wireless local area network periodically scans the target signal strength and interference signal strength in the wireless environment, broadcasts the target signal strength, interference signal strength, and the charging station's location within the wireless local area network, and constructs a signal attenuation heatmap based on the target signal strength, interference signal strength, and charging station location of each charging station; the target signal is a 5G signal.

[0082] Each charging station will broadcast its node status data, including at least network load indicators, device operating status, and location and topology data, within the wireless local area network;

[0083] Based on the signal attenuation heatmap and node status data, each charging pile re-elects a master node using a weighted scoring algorithm;

[0084] The second master node election module is specifically used for:

[0085] The master node and each child node maintain a connection through a multi-path heartbeat packet, which embeds a device fingerprint and a time-series encrypted hash value for real-time node authentication.

[0086] When the disconnection between the master node and each child node exceeds a preset time threshold, each child node re-elects a master node based on the signal attenuation heatmap and node status data.

[0087] The network outage fault upload module is specifically used for:

[0088] When the master node disconnects from the server, based on the signal attenuation heatmap and node status data, it dynamically selects a child node to communicate with the server through a multi-hop relay mode, or establishes an end-to-end encrypted tunnel to upload the network outage fault to the server through a mobile terminal connected to the master node via Bluetooth.

[0089] The advantages of this invention are:

[0090] 1. Each charging pile is pre-configured with device fingerprints and a channel allocation model. Upon power-on, the first charging pile performs initialization and establishes a connection with the server as the master node. Subsequent charging piles perform initialization upon power-on and, as child nodes, authenticate with the master node using their device fingerprints before accessing the wireless Mesh network managed by the master node. Each charging pile in the wireless LAN periodically scans the wireless environment to construct a signal attenuation heatmap, and re-elects a master node based on the heatmap and node status data. The master node maintains connection with each child node via multi-path heartbeat packets. When the master node loses connection with any child node, each child node re-elects a master node based on the heatmap and node status data. When the master node loses connection with the server, it dynamically selects a child node to communicate with the server using a multi-hop relay mode, or uploads the network failure information via an encrypted tunnel through a mobile terminal connected to the master node. The master node monitors the real-time operating data of each child node. The system inputs real-time operational data into the channel allocation model to obtain a real-time channel allocation strategy. Based on this strategy, load balancing is performed on the network of each sub-node. Specifically, it replaces traditional wired cabling with a wireless mesh network architecture, combining device fingerprint authentication and a dynamic master node election mechanism to solve the collaborative optimization challenge while ensuring low-cost deployment and high scalability. This is achieved by eliminating cable laying costs through a wireless self-organizing network, supporting plug-and-play expansion of charging piles; optimizing the network topology in real time by periodically constructing signal attenuation heatmaps, and achieving fault self-healing based on multi-path heartbeat packets and dynamic master node switching (sub-nodes autonomously elect a new master node when disconnected), ensuring high reliability without blind spots; blocking unauthorized access through device fingerprint authentication, and dynamically isolating risks through encrypted tunnel transmission and a pre-trained channel allocation model. Ultimately, this system solves the collaborative optimization challenge of reliability, real-time performance, and security while ensuring low-cost deployment and high scalability, supporting the large-scale deployment of distributed charging scenarios.

[0091] 2. Device fingerprints are generated using hardware information such as device serial numbers and MAC addresses through normalization and SHA-256 hashing, ensuring uniqueness and immutability. When sub-nodes connect, a triple two-way authentication process (RSA asymmetric encryption + SM9 national cryptographic algorithm + random number shift transformation) is adopted. Through timestamp verification, random number progressive verification, and two-way verification of device fingerprints, replay attacks and man-in-the-middle attacks are effectively resisted, significantly improving network security. Random numbers and timestamps are dynamically generated for each authentication, and the complexity of the ciphertext is increased through shift transformation (such as character cyclic shift), further reducing the risk of key leakage, which is superior to traditional static key authentication methods.

[0092] 3. The system elects a master node based on a weighted average of signal attenuation heatmap and node status data, enabling real-time optimization of network topology. For example, when the master node's signal attenuates or its load is too high, the system automatically switches to a better node to ensure network stability. The heartbeat packet embeds device fingerprints and time-series encrypted hash values, which maintains the connection between nodes and simultaneously completes real-time identity verification to prevent unauthorized nodes from accessing the network.

[0093] 4. When the master node loses connection with the server, the system dynamically selects child nodes to communicate with the server through multi-hop relay based on the signal heatmap, solving the industry pain point of being unable to connect to the network in areas with no signal, such as basements; it supports establishing an end-to-end encrypted tunnel to upload faults via Bluetooth connection to mobile terminals, providing dual protection for fault reporting capabilities in network outage scenarios and improving system robustness.

[0094] 5. The channel allocation model integrates wireless environment features (CNN + multi-head attention to extract spatial interference), node state features (GRU to extract load), and user behavior features (clustering to extract behavior patterns) to achieve comprehensive analysis of high-dimensional features. It generates triple strategies (channel allocation, load balancing, and network optimization) through deep reinforcement learning. After being verified by conflict detection and interference-free constraints, the online learner adaptively adjusts the strategy, which significantly improves spectrum utilization and network throughput.

[0095] 6. When the first node is powered on, server security authentication is completed through SM3 / SM9 encryption and timeliness verification to avoid manual configuration errors and shorten deployment time; subsequent nodes automatically connect to the Mesh network through a standardized authentication process, supporting the rapid expansion of large-scale charging pile clusters and reducing operation and maintenance costs.

[0096] 7. Multi-source feature extraction layer (spatial interference, load, behavioral patterns) → dynamic decision layer (graph neural network feature fusion + reinforcement learning policy generation) → policy optimization layer (conflict detection + online optimization) to achieve closed-loop optimization from data to policy, reducing the need for manual parameter tuning; GRU is used to replace traditional RNN to process time-series load data, reducing computational overhead and adapting to resource-constrained charging pile edge devices.

[0097] 8. By implementing a wireless mesh network architecture (Wi-SUN), self-organizing communication between devices is achieved, solving the networking problem in signal-deficient areas such as basements. Dynamic master node election (based on signal attenuation heatmaps and node status data) and a multi-hop relay mechanism ensure network robustness and significantly reduce cabling costs. Simultaneously, the use of national cryptographic algorithms (SM3 / SM9) and a triple bidirectional authentication process (including device fingerprint, dynamic random number, and timestamp verification), combined with real-time identity verification via heartbeat packets, constructs a financial-grade security protection system, effectively resisting replay attacks and man-in-the-middle threats. Its intelligent channel allocation model integrates CNN spatial interference features, GRU node load features, and user behavior clustering features, dynamically optimizing channel strategies through deep reinforcement learning to improve spectrum utilization and achieve load balancing. This results in significantly better coverage flexibility, network outage resistance, security, and maintenance costs compared to traditional centralized solutions. Attached Figure Description

[0098] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0099] Figure 1 This is a flowchart of a self-organizing network method for charging piles according to the present invention.

[0100] Figure 2 This is a schematic diagram of the structure of a self-organizing network system for charging piles according to the present invention. Detailed Implementation

[0101] The overall approach of the technical solution in this application is as follows: A wireless mesh network architecture replaces traditional wired cabling. Combined with device fingerprint authentication and a dynamic master node election mechanism, it solves the challenge of collaborative optimization while ensuring low-cost deployment and high scalability. This is achieved by eliminating cable laying costs through a wireless self-organizing network, supporting plug-and-play expansion of charging piles; optimizing the network topology in real time by periodically constructing signal attenuation heatmaps, and achieving fault self-healing based on multi-path heartbeat packets and dynamic master node switching (child nodes autonomously elect a new master node when disconnected), ensuring high reliability without blind spots; blocking unauthorized access through device fingerprint authentication, and dynamically isolating risks through encrypted tunnel transmission and a pre-trained channel allocation model, simultaneously improving real-time performance (low-latency load balancing) and security without hardware redundancy, thereby supporting the large-scale and efficient deployment of distributed charging scenarios.

[0102] Please refer to Figures 1 to 2 As shown, a preferred embodiment of the self-organizing network method for charging piles according to the present invention includes the following steps:

[0103] Step S1: Each charging pile is pre-configured with a unique device fingerprint and a pre-trained channel allocation model; each charging pile's wireless communication module is equipped with a 5G communication unit and a Wi-SUN communication unit; the 5G communication unit is used to communicate with the server, and only the 5G communication unit of the charging pile acting as the master node has communication permissions, while the other 5G communication units only have the permission to scan the wireless environment; the Wi-SUN communication unit is used to build a wireless mesh network.

[0104] Step S2: After the first charging pile is powered on and started, it performs initialization operations and establishes a connection with the server as the master node.

[0105] Step S3: After the charging pile is powered on and started, it performs an initialization operation and, as a child node, performs an authentication operation with the main node through the device fingerprint before connecting to the wireless Mesh network managed by the main node.

[0106] Step S4: Each charging pile in the wireless local area network periodically scans the wireless environment to construct a signal attenuation heat map, and re-elects a master node based on the signal attenuation heat map and node status data.

[0107] Step S5: The master node and each child node maintain a connection through multipath heartbeat packets. When the master node loses connection with each child node, each child node re-elects a master node based on the signal attenuation heatmap and node status data.

[0108] Step S6: When the master node disconnects from the server, the child node is dynamically selected to communicate with the server through a multi-hop relay mode, or the network disconnection fault is uploaded through an encrypted tunnel via a mobile terminal connected to the master node.

[0109] Step S7: The master node monitors the real-time operating data of each sub-node in real time, inputs the real-time operating data into the channel allocation model to obtain the real-time channel allocation strategy, and performs load balancing on the network of each sub-node based on the real-time channel allocation strategy.

[0110] By implementing a wireless mesh network architecture (Wi-SUN), self-organizing communication between devices is achieved, solving the networking problem in signal-deficient areas such as basements. Dynamic master node election (based on signal attenuation heatmaps and node status data) and a multi-hop relay mechanism ensure network robustness and significantly reduce cabling costs. Simultaneously, the use of national cryptographic algorithms (SM3 / SM9) and a triple bidirectional authentication process (including device fingerprint, dynamic random number, and timestamp verification), combined with real-time identity verification via heartbeat packets, constructs a financial-grade security protection system, effectively resisting replay attacks and man-in-the-middle threats. Its intelligent channel allocation model integrates CNN spatial interference features, GRU node load features, and user behavior clustering features, dynamically optimizing channel strategies through deep reinforcement learning to improve spectrum utilization and achieve load balancing. This results in significantly better coverage flexibility, network outage resistance, security, and operational costs compared to traditional centralized solutions.

[0111] In step S1, the device fingerprint generation process is as follows: the device serial number, MAC address and manufacturing date are normalized and concatenated to obtain concatenated data, the concatenated data is hashed using SHA-256 to obtain a first hash value, and bits 2-129 are extracted from the first hash value as the device fingerprint.

[0112] The channel allocation model is used to output a channel allocation strategy based on the running data, and is constructed based on a multi-source feature extraction layer, a dynamic decision layer, and a strategy optimization layer.

[0113] The multi-source feature extraction layer is constructed based on a wireless environment feature extraction module, a node state feature extraction module, and a user behavior feature extraction module. The wireless environment feature extraction module is used to extract spatial interference features from wireless environment data through a convolutional neural network (CNN) combined with a multi-head attention mechanism (CNN layer extracts local spectrum features (such as channel occupancy heatmap), and the multi-head attention layer captures long-distance interference dependence (such as cross-cell co-channel interference)). The node state feature extraction module is used to extract node load features from node state data through a gated recurrent unit (GRU) (learning the temporal variation law of load (such as burst traffic cycle)). The user behavior feature extraction module is used to extract user behavior pattern features from user behavior data through a clustering algorithm (K-means) and a fully connected layer (clustering identifies user group behavior patterns (such as high-density area video stream concentration), and the fully connected layer maps QoS requirements to channel priority weights).

[0114] The dynamic decision-making layer is constructed based on a feature fusion module and a policy generation module. The feature fusion module is used to fuse and concatenate spatial interference features, node load features, and user behavior pattern features through a graph neural network (GNN) to obtain a comprehensive feature vector (GNN aggregates node interference and load information to generate network state embedding; it concatenates user QoS weights to form a comprehensive feature vector). The policy generation module is used to reason about the comprehensive feature vector through a deep reinforcement learning (DRL) agent (PPO algorithm) and output a triplet policy (reward function: throughput gain + blocking rate penalty + interference suppression reward) that includes a channel allocation plan, load balancing instructions, and network optimization rules.

[0115] The strategy optimization layer is constructed based on a conflict detection module and an adaptive optimization module. The conflict detection module is used to verify the channel allocation plan based on interference-free constraints. The adaptive optimization module is used to optimize the triplet strategy through an online learner (Bandit algorithm, which uses an explore-exploitation mechanism to adjust channel binding or power control parameters), and outputs a channel allocation strategy carrying a channel allocation plan, load balancing instructions, and network optimization rules.

[0116] The operational data includes at least wireless environment data, node status data, and user behavior data; the wireless environment data includes at least a signal attenuation heatmap and heartbeat packet data; the node status data includes at least network load indicators, device operating status, and location and topology data; and the user behavior data includes at least charging behavior data and fault and event data.

[0117] Signal attenuation heatmap: Each charging station periodically scans the wireless environment (such as Wi-Fi signal strength and interference level) to construct a signal attenuation heatmap, showing the degree of signal attenuation at different locations (for example, attenuation caused by reinforced concrete structures can reach 30-50dB). The signal attenuation heatmap includes a signal strength matrix, the distribution of interference sources (such as 2.4GHz / 5GHz interference from other devices), and path loss data. This data helps the model evaluate channel quality and avoid communication blind spots caused by signal attenuation.

[0118] Heartbeat data: The master node and child nodes maintain the connection through heartbeats on multiple paths. Heartbeat data includes heartbeat success rate, end-to-end latency (e.g., measured latency is as high as 300-500ms), packet loss rate, and path stability metrics. These metrics reflect the real-time connection quality and are used to identify congested or faulty paths.

[0119] Network load metrics: Real-time bandwidth utilization, data throughput, channel occupancy, and queue latency of each sub-node. These data reveal network congestion points (such as excessive load on edge nodes during peak hours) and help the model dynamically adjust resources.

[0120] Equipment operating status: This includes the charging pile's CPU / memory usage, power status (power on / power off), error codes (such as communication module failure), and connection status (whether it is connected to the master node or server), which is used to prevent node failure.

[0121] Location and topology data: physical distance between nodes, network topology (such as relay hop count), and signal coverage. This data, combined with a signal attenuation heatmap, is used to evaluate the communication efficiency between nodes.

[0122] Charging behavior data, including the frequency of charging start / stop commands (such as surges during peak hours), charging session duration, real-time charging request rate, and user distribution patterns (e.g., concentrated access during peak hours in underground parking garages), is used to predict traffic peaks and avoid congestion.

[0123] Fault and event data, such as network outage events, authentication failure records, and mobile terminal intervention events, provide anomaly context and help the model optimize the channel during faults.

[0124] The channel allocation strategy includes at least a channel allocation plan, load balancing instructions, and network optimization rules. The channel allocation plan includes at least a channel allocation table and channel parameters; the load balancing instructions include at least a traffic scheduling strategy and relay node selection; and the network optimization rules include at least a congestion control mechanism and a security integration mechanism.

[0125] Channel allocation table: Assigns the optimal communication channel to each child node to avoid co-channel interference; for example, when the signal attenuation heatmap shows a high-interference area, the model assigns a low-collision channel.

[0126] Channel parameters include dynamically adjusting channel width (e.g., from 20MHz to 40MHz), transmit power level (enhanced or reduced to balance coverage and power consumption), and frequency band switching (e.g., from 2.4GHz to 5GHz) to optimize power consumption.

[0127] Traffic scheduling strategy: Based on node load metrics and user behavior data, allocate data transmission paths (e.g., redirect data from high-traffic nodes to low-load nodes). For example, during peak charging periods, the model outputs start / stop commands to idle channels to reduce packet loss.

[0128] Relay node selection: When the master node is disconnected, the strategy dynamically selects the best child node as the relay (multi-hop relay mode) based on the signal attenuation heatmap and node status data (such as bandwidth margin) to ensure uninterrupted connection and avoid single point bottlenecks.

[0129] Congestion control mechanisms include rate limiting (such as limiting node data rate during peak hours), priority scheduling (prioritizing real-time charging data), and failover contingency plans (such as designating a backup channel when the primary node fails).

[0130] Security integration mechanism: Although not the primary output, the policy may implicitly contain security rules (such as combining device fingerprint authentication) to ensure that the allocated channel is not exploited by malicious nodes.

[0131] The formula for the loss function of the channel allocation model is:

[0132] L=λ1*L_interf+λ2*L_balance+λ2*L_switch;

[0133] Where L represents the loss value of the loss function; L_interf represents the interference loss, using the normalized mean square error function; L_balance represents the load balancing loss, using the Gini coefficient imbalance function; L_switch represents the switching stability loss, using the cosine similarity function; λ1, λ2, and λ3 all represent weight coefficients.

[0134] Step S2 specifically involves:

[0135] After the first charging pile is powered on and started, it automatically performs initialization operations including at least hardware self-test and status confirmation, software system initialization, communication module initialization, and charging interface pre-test. It obtains the current timestamp as the request time, obtains the device serial number of the device, encrypts the device fingerprint, device serial number and request time into ciphertext information using the SM9 algorithm, calculates the second hash value of the ciphertext information using the SM3 algorithm, generates a connection request based on the ciphertext information and the second hash value, and uploads the connection request to the server through the 5G communication unit of the wireless communication module as the master node.

[0136] The server parses the received connection request to obtain encrypted information and a second hash value. After verifying the integrity of the encrypted information using the second hash value, it decrypts the encrypted information using the SM9 algorithm to obtain the device fingerprint, device serial number, and request time. After verifying the timeliness of the request, it matches the corresponding MAC address and manufacturing date from a preset device management table using the device serial number. After verifying the legality of the device fingerprint based on the device serial number, MAC address, and manufacturing date, it establishes a connection with the charging pile.

[0137] Step S3 specifically involves:

[0138] After the charging pile is powered on and started, it automatically performs initialization operations and, as a child node, creates a pair of public and private keys based on the RSA algorithm, generates a first random number, obtains the current timestamp as the first authentication time, and uses the device fingerprint of the local machine as the sub-fingerprint.

[0139] The child node encrypts the sub-fingerprint, the first random number, and the first authentication time using a private key to obtain ciphertext data A1. It then encrypts the ciphertext data A1 and the public key using the SM9 algorithm to obtain ciphertext data A2. Finally, it shifts each character of the ciphertext data A2 cyclically to the right by 3 bits to obtain ciphertext data A3. Based on the ciphertext data A3, it generates the first authentication request and sends it to the master node.

[0140] The master node parses the received first authentication request to obtain ciphertext data A3, shifts each character of ciphertext data A3 to the left by 3 bits to obtain ciphertext data A2, decrypts ciphertext data A2 into ciphertext data A1 and public key using the SM9 algorithm, decrypts ciphertext data A1 with the public key to obtain sub-fingerprint, first random number and first authentication time, performs timeliness verification using the first authentication time, adds 3 to the value of the first random number to obtain a second random number, obtains the current timestamp as the second authentication time, and uses the device fingerprint of the local machine as the master fingerprint;

[0141] The master node encrypts the master fingerprint, sub-fingerprint, second random number, and second authentication time using the public key to obtain ciphertext data B1. It then encrypts the ciphertext data B1 into ciphertext data B2 using the SM9 algorithm. Finally, it shifts each character of the ciphertext data B2 cyclically to the right by 5 bits to obtain ciphertext data B3. Based on the ciphertext data B3, it generates a second authentication request and sends it to the sub-node.

[0142] The child node parses the received second authentication request to obtain ciphertext data B3, shifts each character of ciphertext data B3 cyclically 5 bits to the left to obtain ciphertext data B2, decrypts ciphertext data B2 into ciphertext data B1 using the SM9 algorithm, decrypts ciphertext data B1 using the private key to obtain the master fingerprint, sub-fingerprint, second random number, and second authentication time, performs timeliness verification using the second authentication time, verifies the second random number using the first random number, verifies the decrypted sub-fingerprint using the sub-fingerprint of the local machine, adds 5 to the value of the second random number to obtain the third random number, and obtains the current timestamp as the third authentication time;

[0143] The child node encrypts the master fingerprint, the third random number, and the third authentication time using a private key to obtain ciphertext data A4. It then encrypts the ciphertext data A4 into ciphertext data A5 using the SM9 algorithm. Finally, it shifts each character of the ciphertext data A5 cyclically to the right by 3 bits to obtain ciphertext data A6. Based on the ciphertext data A6, it generates a third authentication request and sends it to the master node.

[0144] The master node parses the received third authentication request to obtain ciphertext data A6. It then shifts each character of ciphertext data A6 cyclically 3 bits to the left to obtain ciphertext data A5. Using the SM9 algorithm, it decrypts ciphertext data A5 into ciphertext data A4. The master node then decrypts ciphertext data A4 using the public key to obtain the master fingerprint, the third random number, and the third authentication time. After verifying the timeliness of the third authentication time, it verifies the third random number using the second random number. Finally, it verifies the decrypted master fingerprint using the master fingerprint of the local machine. If the verification passes, the bidirectional authentication operation is completed, and the node accesses the wireless Mesh network managed by the master node through the Wi-SUN communication unit of the wireless communication module.

[0145] Step S4 specifically involves:

[0146] Each charging station within the wireless local area network periodically scans the target signal strength and interference signal strength in the wireless environment, and broadcasts the target signal strength, interference signal strength, and the charging station's location (location and topology data) within the wireless local area network. A signal attenuation heatmap is constructed based on the target signal strength, interference signal strength, and charging station location of each charging station; the target signal is a 5G signal.

[0147] Each charging station will broadcast its node status data, including at least network load indicators, device operating status, and location and topology data, within the wireless local area network;

[0148] Based on the signal attenuation heatmap and node status data, each charging pile re-elects a master node through a weighted scoring algorithm (e.g., selecting a charging pile with high target signal strength and idle resources as the master node);

[0149] Step S5 specifically involves:

[0150] The master node and each child node maintain connection through multi-path heartbeat packets (i.e., the master node and child nodes send heartbeat packets simultaneously through multiple independent physical paths (such as WiFi 2.4GHz + 5GHz dual-band, LoRa + BLE hybrid links) to avoid interruption due to signal attenuation or interference on a single path). The heartbeat packets embed device fingerprints and temporal encrypted hash values ​​for real-time node authentication (Temporal Encrypted Hash Value is a secure digest value generated by combining timestamps, dynamic random numbers, and encrypted hash algorithms. It is mainly used to solve security problems such as identity forgery and replay attacks. Its core is to enhance the anti-tampering capability of traditional hashes through the time dimension. It is suitable for scenarios such as real-time communication and node authentication).

[0151] When the disconnection between the master node and each child node exceeds a preset time threshold, each child node re-elects a master node based on the signal attenuation heatmap and node status data.

[0152] Step S6 specifically involves:

[0153] When the master node disconnects from the server, based on the signal attenuation heatmap and node status data, it dynamically selects a child node to communicate with the server through a multi-hop relay mode, or establishes an end-to-end encrypted tunnel to upload the network outage fault to the server through a mobile terminal connected to the master node via Bluetooth.

[0154] A preferred embodiment of the self-organizing network system for charging piles according to the present invention includes the following modules:

[0155] The initial setup module is used to pre-configure a unique device fingerprint and a pre-trained channel allocation model for each charging pile; each charging pile's wireless communication module is equipped with a 5G communication unit and a Wi-SUN communication unit; the 5G communication unit is used to communicate with the server, and only the 5G communication unit of the charging pile acting as the master node has communication permissions, while the other 5G communication units only have the permission to scan the wireless environment; the Wi-SUN communication unit is used to build a wireless mesh network.

[0156] The master node connection module is used to perform initialization operations after the first charging pile is powered on and started, and to establish a connection with the server as the master node.

[0157] The wireless mesh networking module is used to perform initialization operations after the charging pile is powered on and started, and as a child node, it performs authentication operations with the main node through the device fingerprint and then accesses the wireless mesh network managed by the main node.

[0158] The first master node election module is used to periodically scan the wireless environment of each charging pile in the wireless local area network to construct a signal attenuation heat map, and re-elect the master node based on the signal attenuation heat map and node status data.

[0159] The second master node election module is used to maintain the connection between the master node and each child node through multi-path heartbeat packets. When the master node loses connection with each child node, each child node re-elects the master node based on the signal attenuation heatmap and node status data.

[0160] The network outage fault upload module is used to dynamically select a child node to communicate with the server through a multi-hop relay mode when the master node loses connection with the server, or to upload the network outage fault through a mobile terminal connected to the master node in an encrypted tunnel.

[0161] The load balancing module is used by the master node to monitor the real-time operating data of each sub-node, input the real-time operating data into the channel allocation model to obtain the real-time channel allocation strategy, and perform load balancing on the network of each sub-node based on the real-time channel allocation strategy.

[0162] By implementing a wireless mesh network architecture (Wi-SUN), self-organizing communication between devices is achieved, solving the networking problem in signal-deficient areas such as basements. Dynamic master node election (based on signal attenuation heatmaps and node status data) and a multi-hop relay mechanism ensure network robustness and significantly reduce cabling costs. Simultaneously, the use of national cryptographic algorithms (SM3 / SM9) and a triple bidirectional authentication process (including device fingerprint, dynamic random number, and timestamp verification), combined with real-time identity verification via heartbeat packets, constructs a financial-grade security protection system, effectively resisting replay attacks and man-in-the-middle threats. Its intelligent channel allocation model integrates CNN spatial interference features, GRU node load features, and user behavior clustering features, dynamically optimizing channel strategies through deep reinforcement learning to improve spectrum utilization and achieve load balancing. This results in significantly better coverage flexibility, network outage resistance, security, and operational costs compared to traditional centralized solutions.

[0163] In the initial setup module, the device fingerprint generation process is as follows: the device serial number, MAC address and manufacturing date are normalized and concatenated to obtain concatenated data, the concatenated data is hashed using SHA-256 to obtain a first hash value, and bits 2-129 are extracted from the first hash value as the device fingerprint.

[0164] The channel allocation model is used to output a channel allocation strategy based on the running data, and is constructed based on a multi-source feature extraction layer, a dynamic decision layer, and a strategy optimization layer.

[0165] The multi-source feature extraction layer is constructed based on a wireless environment feature extraction module, a node state feature extraction module, and a user behavior feature extraction module. The wireless environment feature extraction module is used to extract spatial interference features from wireless environment data through a convolutional neural network (CNN) combined with a multi-head attention mechanism (CNN layer extracts local spectrum features (such as channel occupancy heatmap), and the multi-head attention layer captures long-distance interference dependence (such as cross-cell co-channel interference)). The node state feature extraction module is used to extract node load features from node state data through a gated recurrent unit (GRU) (learning the temporal variation law of load (such as burst traffic cycle)). The user behavior feature extraction module is used to extract user behavior pattern features from user behavior data through a clustering algorithm (K-means) and a fully connected layer (clustering identifies user group behavior patterns (such as high-density area video stream concentration), and the fully connected layer maps QoS requirements to channel priority weights).

[0166] The dynamic decision-making layer is constructed based on a feature fusion module and a policy generation module. The feature fusion module is used to fuse and concatenate spatial interference features, node load features, and user behavior pattern features through a graph neural network (GNN) to obtain a comprehensive feature vector (GNN aggregates node interference and load information to generate network state embedding; it concatenates user QoS weights to form a comprehensive feature vector). The policy generation module is used to reason about the comprehensive feature vector through a deep reinforcement learning (DRL) agent (PPO algorithm) and output a triplet policy (reward function: throughput gain + blocking rate penalty + interference suppression reward) that includes a channel allocation plan, load balancing instructions, and network optimization rules.

[0167] The strategy optimization layer is constructed based on a conflict detection module and an adaptive optimization module. The conflict detection module is used to verify the channel allocation plan based on interference-free constraints. The adaptive optimization module is used to optimize the triplet strategy through an online learner (Bandit algorithm, which uses an explore-exploitation mechanism to adjust channel binding or power control parameters), and outputs a channel allocation strategy carrying a channel allocation plan, load balancing instructions, and network optimization rules.

[0168] The operational data includes at least wireless environment data, node status data, and user behavior data; the wireless environment data includes at least a signal attenuation heatmap and heartbeat packet data; the node status data includes at least network load indicators, device operating status, and location and topology data; and the user behavior data includes at least charging behavior data and fault and event data.

[0169] Signal attenuation heatmap: Each charging station periodically scans the wireless environment (such as Wi-Fi signal strength and interference level) to construct a signal attenuation heatmap, showing the degree of signal attenuation at different locations (for example, attenuation caused by reinforced concrete structures can reach 30-50dB). The signal attenuation heatmap includes a signal strength matrix, the distribution of interference sources (such as 2.4GHz / 5GHz interference from other devices), and path loss data. This data helps the model evaluate channel quality and avoid communication blind spots caused by signal attenuation.

[0170] Heartbeat data: The master node and child nodes maintain the connection through heartbeats on multiple paths. Heartbeat data includes heartbeat success rate, end-to-end latency (e.g., measured latency is as high as 300-500ms), packet loss rate, and path stability metrics. These metrics reflect the real-time connection quality and are used to identify congested or faulty paths.

[0171] Network load metrics: Real-time bandwidth utilization, data throughput, channel occupancy, and queue latency of each sub-node. These data reveal network congestion points (such as excessive load on edge nodes during peak hours) and help the model dynamically adjust resources.

[0172] Equipment operating status: This includes the charging pile's CPU / memory usage, power status (power on / power off), error codes (such as communication module failure), and connection status (whether it is connected to the master node or server), which is used to prevent node failure.

[0173] Location and topology data: physical distance between nodes, network topology (such as relay hop count), and signal coverage. This data, combined with a signal attenuation heatmap, is used to evaluate the communication efficiency between nodes.

[0174] Charging behavior data, including the frequency of charging start / stop commands (such as surges during peak hours), charging session duration, real-time charging request rate, and user distribution patterns (e.g., concentrated access during peak hours in underground parking garages), is used to predict traffic peaks and avoid congestion.

[0175] Fault and event data, such as network outage events, authentication failure records, and mobile terminal intervention events, provide anomaly context and help the model optimize the channel during faults.

[0176] The channel allocation strategy includes at least a channel allocation plan, load balancing instructions, and network optimization rules. The channel allocation plan includes at least a channel allocation table and channel parameters; the load balancing instructions include at least a traffic scheduling strategy and relay node selection; and the network optimization rules include at least a congestion control mechanism and a security integration mechanism.

[0177] Channel allocation table: Assigns the optimal communication channel to each child node to avoid co-channel interference; for example, when the signal attenuation heatmap shows a high-interference area, the model assigns a low-collision channel.

[0178] Channel parameters include dynamically adjusting channel width (e.g., from 20MHz to 40MHz), transmit power level (enhanced or reduced to balance coverage and power consumption), and frequency band switching (e.g., from 2.4GHz to 5GHz) to optimize power consumption.

[0179] Traffic scheduling strategy: Based on node load metrics and user behavior data, allocate data transmission paths (e.g., redirect data from high-traffic nodes to low-load nodes). For example, during peak charging periods, the model outputs start / stop commands to idle channels to reduce packet loss.

[0180] Relay node selection: When the master node is disconnected, the strategy dynamically selects the best child node as the relay (multi-hop relay mode) based on the signal attenuation heatmap and node status data (such as bandwidth margin) to ensure uninterrupted connection and avoid single point bottlenecks.

[0181] Congestion control mechanisms include rate limiting (such as limiting node data rate during peak hours), priority scheduling (prioritizing real-time charging data), and failover contingency plans (such as designating a backup channel when the primary node fails).

[0182] Security integration mechanism: Although not the primary output, the policy may implicitly contain security rules (such as combining device fingerprint authentication) to ensure that the allocated channel is not exploited by malicious nodes.

[0183] The formula for the loss function of the channel allocation model is:

[0184] L=λ1*L_interf+λ2*L_balance+λ3*L_switch;

[0185] Where L represents the loss value of the loss function; L_interf represents the interference loss, using the normalized mean square error function; L_balance represents the load balancing loss, using the Gini coefficient imbalance function; L_switch represents the switching stability loss, using the cosine similarity function; λ1, λ2, and λ3 all represent weight coefficients.

[0186] The master node connection module is specifically used for:

[0187] After the first charging pile is powered on and started, it automatically performs initialization operations including at least hardware self-test and status confirmation, software system initialization, communication module initialization, and charging interface pre-test. It obtains the current timestamp as the request time, obtains the device serial number of the device, encrypts the device fingerprint, device serial number and request time into ciphertext information using the SM9 algorithm, calculates the second hash value of the ciphertext information using the SM3 algorithm, generates a connection request based on the ciphertext information and the second hash value, and uploads the connection request to the server through the 5G communication unit of the wireless communication module as the master node.

[0188] The server parses the received connection request to obtain encrypted information and a second hash value. After verifying the integrity of the encrypted information using the second hash value, it decrypts the encrypted information using the SM9 algorithm to obtain the device fingerprint, device serial number, and request time. After verifying the timeliness of the request, it matches the corresponding MAC address and manufacturing date from a preset device management table using the device serial number. After verifying the legality of the device fingerprint based on the device serial number, MAC address, and manufacturing date, it establishes a connection with the charging pile.

[0189] The wireless mesh networking module is specifically used for:

[0190] After the charging pile is powered on and started, it automatically performs initialization operations and, as a child node, creates a pair of public and private keys based on the RSA algorithm, generates a first random number, obtains the current timestamp as the first authentication time, and uses the device fingerprint of the local machine as the sub-fingerprint.

[0191] The child node encrypts the sub-fingerprint, the first random number, and the first authentication time using a private key to obtain ciphertext data A1. It then encrypts the ciphertext data A1 and the public key using the SM9 algorithm to obtain ciphertext data A2. Finally, it shifts each character of the ciphertext data A2 cyclically to the right by 3 bits to obtain ciphertext data A3. Based on the ciphertext data A3, it generates the first authentication request and sends it to the master node.

[0192] The master node parses the received first authentication request to obtain ciphertext data A3, shifts each character of ciphertext data A3 to the left by 3 bits to obtain ciphertext data A2, decrypts ciphertext data A2 into ciphertext data A1 and public key using the SM9 algorithm, decrypts ciphertext data A1 with the public key to obtain sub-fingerprint, first random number and first authentication time, performs timeliness verification using the first authentication time, adds 3 to the value of the first random number to obtain a second random number, obtains the current timestamp as the second authentication time, and uses the device fingerprint of the local machine as the master fingerprint;

[0193] The master node encrypts the master fingerprint, sub-fingerprint, second random number, and second authentication time using the public key to obtain ciphertext data B1. It then encrypts the ciphertext data B1 into ciphertext data B2 using the SM9 algorithm. Finally, it shifts each character of the ciphertext data B2 cyclically to the right by 5 bits to obtain ciphertext data B3. Based on the ciphertext data B3, it generates a second authentication request and sends it to the sub-node.

[0194] The child node parses the received second authentication request to obtain ciphertext data B3, shifts each character of ciphertext data B3 cyclically 5 bits to the left to obtain ciphertext data B2, decrypts ciphertext data B2 into ciphertext data B1 using the SM9 algorithm, decrypts ciphertext data B1 using the private key to obtain the master fingerprint, sub-fingerprint, second random number, and second authentication time, performs timeliness verification using the second authentication time, verifies the second random number using the first random number, verifies the decrypted sub-fingerprint using the sub-fingerprint of the local machine, adds 5 to the value of the second random number to obtain the third random number, and obtains the current timestamp as the third authentication time;

[0195] The child node encrypts the master fingerprint, the third random number, and the third authentication time using a private key to obtain ciphertext data A4. It then encrypts the ciphertext data A4 into ciphertext data A5 using the SM9 algorithm. Finally, it shifts each character of the ciphertext data A5 cyclically to the right by 3 bits to obtain ciphertext data A6. Based on the ciphertext data A6, it generates a third authentication request and sends it to the master node.

[0196] The master node parses the received third authentication request to obtain ciphertext data A6. It then shifts each character of ciphertext data A6 cyclically 3 bits to the left to obtain ciphertext data A5. Using the SM9 algorithm, it decrypts ciphertext data A5 into ciphertext data A4. The master node then decrypts ciphertext data A4 using the public key to obtain the master fingerprint, the third random number, and the third authentication time. After verifying the timeliness of the third authentication time, it verifies the third random number using the second random number. Finally, it verifies the decrypted master fingerprint using the master fingerprint of the local machine. If the verification passes, the bidirectional authentication operation is completed, and the node accesses the wireless Mesh network managed by the master node through the Wi-SUN communication unit of the wireless communication module.

[0197] The first master node election module is specifically used for:

[0198] Each charging station within the wireless local area network periodically scans the target signal strength and interference signal strength in the wireless environment, and broadcasts the target signal strength, interference signal strength, and the charging station's location (location and topology data) within the wireless local area network. A signal attenuation heatmap is constructed based on the target signal strength, interference signal strength, and charging station location of each charging station; the target signal is a 5G signal.

[0199] Each charging station will broadcast its node status data, including at least network load indicators, device operating status, and location and topology data, within the wireless local area network;

[0200] Based on the signal attenuation heatmap and node status data, each charging pile re-elects a master node through a weighted scoring algorithm (e.g., selecting a charging pile with high target signal strength and idle resources as the master node);

[0201] The second master node election module is specifically used for:

[0202] The master node and each child node maintain connection through multi-path heartbeat packets (i.e., the master node and child nodes send heartbeat packets simultaneously through multiple independent physical paths (such as WiFi 2.4GHz + 5GHz dual-band, LoRa + BLE hybrid links) to avoid interruption due to signal attenuation or interference on a single path). The heartbeat packets embed device fingerprints and temporal encrypted hash values ​​for real-time node authentication (Temporal Encrypted Hash Value is a secure digest value generated by combining timestamps, dynamic random numbers, and encrypted hash algorithms. It is mainly used to solve security problems such as identity forgery and replay attacks. Its core is to enhance the anti-tampering capability of traditional hashes through the time dimension. It is suitable for scenarios such as real-time communication and node authentication).

[0203] When the disconnection between the master node and each child node exceeds a preset time threshold, each child node re-elects a master node based on the signal attenuation heatmap and node status data.

[0204] The network outage fault upload module is specifically used for:

[0205] When the master node disconnects from the server, based on the signal attenuation heatmap and node status data, it dynamically selects a child node to communicate with the server through a multi-hop relay mode, or establishes an end-to-end encrypted tunnel to upload the network outage fault to the server through a mobile terminal connected to the master node via Bluetooth.

[0206] In summary, the advantages of this invention are:

[0207] 1. Each charging pile is pre-configured with device fingerprints and a channel allocation model. Upon power-on, the first charging pile performs initialization and establishes a connection with the server as the master node. Subsequent charging piles perform initialization upon power-on and, as child nodes, authenticate with the master node using their device fingerprints before accessing the wireless Mesh network managed by the master node. Each charging pile in the wireless LAN periodically scans the wireless environment to construct a signal attenuation heatmap, and re-elects a master node based on the heatmap and node status data. The master node maintains connection with each child node via multi-path heartbeat packets. When the master node loses connection with any child node, each child node re-elects a master node based on the heatmap and node status data. When the master node loses connection with the server, it dynamically selects a child node to communicate with the server using a multi-hop relay mode, or uploads the network failure information via an encrypted tunnel through a mobile terminal connected to the master node. The master node monitors the real-time operating data of each child node. The system inputs real-time operational data into the channel allocation model to obtain a real-time channel allocation strategy. Based on this strategy, load balancing is performed on the network of each sub-node. Specifically, it replaces traditional wired cabling with a wireless mesh network architecture, combining device fingerprint authentication and a dynamic master node election mechanism to solve the collaborative optimization challenge while ensuring low-cost deployment and high scalability. This is achieved by eliminating cable laying costs through a wireless self-organizing network, supporting plug-and-play expansion of charging piles; optimizing the network topology in real time by periodically constructing signal attenuation heatmaps, and achieving fault self-healing based on multi-path heartbeat packets and dynamic master node switching (sub-nodes autonomously elect a new master node when disconnected), ensuring high reliability without blind spots; blocking unauthorized access through device fingerprint authentication, and dynamically isolating risks through encrypted tunnel transmission and a pre-trained channel allocation model. Ultimately, this system solves the collaborative optimization challenge of reliability, real-time performance, and security while ensuring low-cost deployment and high scalability, supporting the large-scale deployment of distributed charging scenarios.

[0208] 2. Device fingerprints are generated using hardware information such as device serial numbers and MAC addresses through normalization and SHA-256 hashing, ensuring uniqueness and immutability. When sub-nodes connect, a triple two-way authentication process (RSA asymmetric encryption + SM9 national cryptographic algorithm + random number shift transformation) is adopted. Through timestamp verification, random number progressive verification, and two-way verification of device fingerprints, replay attacks and man-in-the-middle attacks are effectively resisted, significantly improving network security. Random numbers and timestamps are dynamically generated for each authentication, and the complexity of the ciphertext is increased through shift transformation (such as character cyclic shift), further reducing the risk of key leakage, which is superior to traditional static key authentication methods.

[0209] 3. The system elects a master node based on a weighted average of signal attenuation heatmap and node status data, enabling real-time optimization of network topology. For example, when the master node's signal attenuates or its load is too high, the system automatically switches to a better node to ensure network stability. The heartbeat packet embeds device fingerprints and time-series encrypted hash values, which maintains the connection between nodes and simultaneously completes real-time identity verification to prevent unauthorized nodes from accessing the network.

[0210] 4. When the master node loses connection with the server, the system dynamically selects child nodes to communicate with the server through multi-hop relay based on the signal heatmap, solving the industry pain point of being unable to connect to the network in areas with no signal, such as basements; it supports establishing an end-to-end encrypted tunnel to upload faults via Bluetooth connection to mobile terminals, providing dual protection for fault reporting capabilities in network outage scenarios and improving system robustness.

[0211] 5. The channel allocation model integrates wireless environment features (CNN + multi-head attention to extract spatial interference), node state features (GRU to extract load), and user behavior features (clustering to extract behavior patterns) to achieve comprehensive analysis of high-dimensional features. It generates triple strategies (channel allocation, load balancing, and network optimization) through deep reinforcement learning. After being verified by conflict detection and interference-free constraints, the online learner adaptively adjusts the strategy, which significantly improves spectrum utilization and network throughput.

[0212] 6. When the first node is powered on, server security authentication is completed through SM3 / SM9 encryption and timeliness verification to avoid manual configuration errors and shorten deployment time; subsequent nodes automatically connect to the Mesh network through a standardized authentication process, supporting the rapid expansion of large-scale charging pile clusters and reducing operation and maintenance costs.

[0213] 7. Multi-source feature extraction layer (spatial interference, load, behavioral patterns) → dynamic decision layer (graph neural network feature fusion + reinforcement learning policy generation) → policy optimization layer (conflict detection + online optimization) to achieve closed-loop optimization from data to policy, reducing the need for manual parameter tuning; GRU is used to replace traditional RNN to process time-series load data, reducing computational overhead and adapting to resource-constrained charging pile edge devices.

[0214] 8. By implementing a wireless mesh network architecture (Wi-SUN), self-organizing communication between devices is achieved, solving the networking problem in signal-deficient areas such as basements. Dynamic master node election (based on signal attenuation heatmaps and node status data) and a multi-hop relay mechanism ensure network robustness and significantly reduce cabling costs. Simultaneously, the use of national cryptographic algorithms (SM3 / SM9) and a triple bidirectional authentication process (including device fingerprint, dynamic random number, and timestamp verification), combined with real-time identity verification via heartbeat packets, constructs a financial-grade security protection system, effectively resisting replay attacks and man-in-the-middle threats. Its intelligent channel allocation model integrates CNN spatial interference features, GRU node load features, and user behavior clustering features, dynamically optimizing channel strategies through deep reinforcement learning to improve spectrum utilization and achieve load balancing. This results in significantly better coverage flexibility, network outage resistance, security, and maintenance costs compared to traditional centralized solutions.

[0215] While specific embodiments of the present invention have been described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and not intended to limit the scope of the present invention. Equivalent modifications and variations made by those skilled in the art in accordance with the spirit of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for self-organizing a charging pile network, characterized in that: Includes the following steps: Step S1: Each charging pile is pre-set with a unique device fingerprint and a pre-trained channel allocation model. Step S2: After the first charging pile is powered on and started, it performs initialization operations and establishes a connection with the server as the master node. Step S3: After the charging pile is powered on and started, it performs an initialization operation and, as a child node, performs an authentication operation with the main node through the device fingerprint before connecting to the wireless Mesh network managed by the main node. Step S4: Each charging pile in the wireless local area network periodically scans the wireless environment to construct a signal attenuation heat map, and re-elects a master node based on the signal attenuation heat map and node status data. Step S5: The master node and each child node maintain a connection through multipath heartbeat packets. When the master node loses connection with each child node, each child node re-elects a master node based on the signal attenuation heatmap and node status data. Step S6: When the master node disconnects from the server, the child node is dynamically selected to communicate with the server through a multi-hop relay mode, or the network disconnection fault is uploaded through an encrypted tunnel via a mobile terminal connected to the master node. Step S7: The master node monitors the real-time operating data of each sub-node in real time, inputs the real-time operating data into the channel allocation model to obtain the real-time channel allocation strategy, and performs load balancing on the network of each sub-node based on the real-time channel allocation strategy.

2. The self-organizing network method for charging piles as described in claim 1, characterized in that: In step S1, the device fingerprint generation process is as follows: the device serial number, MAC address and manufacturing date are normalized and concatenated to obtain concatenated data, the concatenated data is hashed using SHA-256 to obtain a first hash value, and bits 2-129 are extracted from the first hash value as the device fingerprint. The channel allocation model is used to output a channel allocation strategy based on the running data, and is constructed based on a multi-source feature extraction layer, a dynamic decision layer, and a strategy optimization layer. The multi-source feature extraction layer is constructed based on a wireless environment feature extraction module, a node state feature extraction module, and a user behavior feature extraction module. The wireless environment feature extraction module is used to extract spatial interference features from wireless environment data through a convolutional neural network combined with a multi-head attention mechanism. The node state feature extraction module is used to extract node load features from node state data through a gated recurrent unit. The user behavior feature extraction module is used to extract user behavior pattern features from user behavior data through a clustering algorithm and a fully connected layer. The dynamic decision-making layer is constructed based on a feature fusion module and a policy generation module. The feature fusion module is used to fuse and concatenate spatial interference features, node load features, and user behavior pattern features through a graph neural network to obtain a comprehensive feature vector. The policy generation module is used to reason about the comprehensive feature vector through a deep reinforcement learning agent and output a triplet policy including a channel allocation plan, a load balancing instruction, and a network optimization rule. The strategy optimization layer is constructed based on a conflict detection module and an adaptive optimization module; the conflict detection module is used to verify the channel allocation plan based on interference-free constraints; the adaptive optimization module is used to optimize the triplet strategy through an online learner and output a channel allocation strategy carrying a channel allocation plan, load balancing instructions and network optimization rules. The operational data includes at least wireless environment data, node status data, and user behavior data. The channel allocation strategy includes at least a channel allocation plan, load balancing instructions, and network optimization rules.

3. The self-organizing network method for charging piles as described in claim 1, characterized in that: Step S2 specifically involves: After the first charging pile is powered on and started, it automatically performs initialization operations including at least hardware self-test and status confirmation, software system initialization, communication module initialization, and charging interface pre-test. It obtains the current timestamp as the request time, obtains the device serial number of the device, encrypts the device fingerprint, device serial number and request time into ciphertext information using the SM9 algorithm, calculates the second hash value of the ciphertext information using the SM3 algorithm, generates a connection request based on the ciphertext information and the second hash value, and uploads the connection request to the server through the 5G communication unit of the wireless communication module as the master node. The server parses the received connection request to obtain encrypted information and a second hash value. After verifying the integrity of the encrypted information using the second hash value, it decrypts the encrypted information using the SM9 algorithm to obtain the device fingerprint, device serial number, and request time. After verifying the timeliness of the request, it matches the corresponding MAC address and manufacturing date from a preset device management table using the device serial number. After verifying the legality of the device fingerprint based on the device serial number, MAC address, and manufacturing date, it establishes a connection with the charging pile.

4. The self-organizing network method for charging piles as described in claim 1, characterized in that: Step S3 specifically involves: After the charging pile is powered on and started, it automatically performs initialization operations and, as a child node, creates a pair of public and private keys based on the RSA algorithm, generates a first random number, obtains the current timestamp as the first authentication time, and uses the device fingerprint of the local machine as the sub-fingerprint. The child node encrypts the sub-fingerprint, the first random number, and the first authentication time using a private key to obtain ciphertext data A1. It then encrypts the ciphertext data A1 and the public key using the SM9 algorithm to obtain ciphertext data A2. Finally, it shifts each character of the ciphertext data A2 cyclically to the right by 3 bits to obtain ciphertext data A3. Based on the ciphertext data A3, it generates the first authentication request and sends it to the master node. The master node parses the received first authentication request to obtain ciphertext data A3, shifts each character of ciphertext data A3 to the left by 3 bits to obtain ciphertext data A2, decrypts ciphertext data A2 into ciphertext data A1 and public key using the SM9 algorithm, decrypts ciphertext data A1 with the public key to obtain sub-fingerprint, first random number and first authentication time, performs timeliness verification using the first authentication time, adds 3 to the value of the first random number to obtain a second random number, obtains the current timestamp as the second authentication time, and uses the device fingerprint of the local machine as the master fingerprint; The master node encrypts the master fingerprint, sub-fingerprint, second random number, and second authentication time using the public key to obtain ciphertext data B1. It then encrypts the ciphertext data B1 into ciphertext data B2 using the SM9 algorithm. Finally, it shifts each character of the ciphertext data B2 cyclically to the right by 5 bits to obtain ciphertext data B3. Based on the ciphertext data B3, it generates a second authentication request and sends it to the sub-node. The child node parses the received second authentication request to obtain ciphertext data B3, shifts each character of ciphertext data B3 cyclically 5 bits to the left to obtain ciphertext data B2, decrypts ciphertext data B2 into ciphertext data B1 using the SM9 algorithm, decrypts ciphertext data B1 using the private key to obtain the master fingerprint, sub-fingerprint, second random number, and second authentication time, performs timeliness verification using the second authentication time, verifies the second random number using the first random number, verifies the decrypted sub-fingerprint using the sub-fingerprint of the local machine, adds 5 to the value of the second random number to obtain the third random number, and obtains the current timestamp as the third authentication time; The child node encrypts the master fingerprint, the third random number, and the third authentication time using a private key to obtain ciphertext data A4. It then encrypts the ciphertext data A4 into ciphertext data A5 using the SM9 algorithm. Finally, it shifts each character of the ciphertext data A5 cyclically to the right by 3 bits to obtain ciphertext data A6. Based on the ciphertext data A6, it generates a third authentication request and sends it to the master node. The master node parses the received third authentication request to obtain ciphertext data A6. It then shifts each character of ciphertext data A6 cyclically 3 bits to the left to obtain ciphertext data A5. Using the SM9 algorithm, it decrypts ciphertext data A5 into ciphertext data A4. The master node then decrypts ciphertext data A4 using the public key to obtain the master fingerprint, the third random number, and the third authentication time. After verifying the timeliness of the third authentication time, it verifies the third random number using the second random number. Finally, it verifies the decrypted master fingerprint using the master fingerprint of the local machine. If the verification passes, the bidirectional authentication operation is completed, and the node accesses the wireless Mesh network managed by the master node through the Wi-SUN communication unit of the wireless communication module.

5. The self-organizing network method for charging piles as described in claim 1, characterized in that: Step S4 specifically involves: Each charging station within the wireless local area network periodically scans the target signal strength and interference signal strength in the wireless environment, broadcasts the target signal strength, interference signal strength, and the charging station's location within the wireless local area network, and constructs a signal attenuation heatmap based on the target signal strength, interference signal strength, and charging station location of each charging station; the target signal is a 5G signal. Each charging station will broadcast its node status data, including at least network load indicators, device operating status, and location and topology data, within the wireless local area network; Based on the signal attenuation heatmap and node status data, each charging pile re-elects a master node using a weighted scoring algorithm; Step S5 specifically involves: The master node and each child node maintain a connection through a multi-path heartbeat packet, which embeds a device fingerprint and a time-series encrypted hash value for real-time node authentication. When the disconnection between the master node and each child node exceeds a preset time threshold, each child node re-elects a master node based on the signal attenuation heatmap and node status data. Step S6 specifically involves: When the master node disconnects from the server, based on the signal attenuation heatmap and node status data, it dynamically selects a child node to communicate with the server through a multi-hop relay mode, or establishes an end-to-end encrypted tunnel to upload the network outage fault to the server through a mobile terminal connected to the master node via Bluetooth.

6. A self-organizing network system for charging piles, characterized in that: Includes the following modules: The initial setup module is used to pre-configure a unique device fingerprint and a pre-trained channel allocation model for each charging pile. The master node connection module is used to perform initialization operations after the first charging pile is powered on and started, and to establish a connection with the server as the master node. The wireless mesh networking module is used to perform initialization operations after the charging pile is powered on and started, and as a child node, it performs authentication operations with the main node through the device fingerprint and then accesses the wireless mesh network managed by the main node. The first master node election module is used to periodically scan the wireless environment of each charging pile in the wireless local area network to construct a signal attenuation heat map, and re-elect the master node based on the signal attenuation heat map and node status data. The second master node election module is used to maintain the connection between the master node and each child node through multi-path heartbeat packets. When the master node loses connection with each child node, each child node re-elects the master node based on the signal attenuation heatmap and node status data. The network outage fault upload module is used to dynamically select a child node to communicate with the server through a multi-hop relay mode when the master node loses connection with the server, or to upload the network outage fault through a mobile terminal connected to the master node in an encrypted tunnel. The load balancing module is used by the master node to monitor the real-time operating data of each sub-node, input the real-time operating data into the channel allocation model to obtain the real-time channel allocation strategy, and perform load balancing on the network of each sub-node based on the real-time channel allocation strategy.

7. A charging pile self-organizing network system as described in claim 6, characterized in that: In the initial setup module, the device fingerprint generation process is as follows: the device serial number, MAC address and manufacturing date are normalized and concatenated to obtain concatenated data, the concatenated data is hashed using SHA-256 to obtain a first hash value, and bits 2-129 are extracted from the first hash value as the device fingerprint. The channel allocation model is used to output a channel allocation strategy based on the running data, and is constructed based on a multi-source feature extraction layer, a dynamic decision layer, and a strategy optimization layer. The multi-source feature extraction layer is constructed based on a wireless environment feature extraction module, a node state feature extraction module, and a user behavior feature extraction module. The wireless environment feature extraction module is used to extract spatial interference features from wireless environment data through a convolutional neural network combined with a multi-head attention mechanism. The node state feature extraction module is used to extract node load features from node state data through a gated recurrent unit. The user behavior feature extraction module is used to extract user behavior pattern features from user behavior data through a clustering algorithm and a fully connected layer. The dynamic decision-making layer is constructed based on a feature fusion module and a policy generation module. The feature fusion module is used to fuse and concatenate spatial interference features, node load features, and user behavior pattern features through a graph neural network to obtain a comprehensive feature vector. The policy generation module is used to reason about the comprehensive feature vector through a deep reinforcement learning agent and output a triplet policy including a channel allocation plan, a load balancing instruction, and a network optimization rule. The strategy optimization layer is constructed based on a conflict detection module and an adaptive optimization module; the conflict detection module is used to verify the channel allocation plan based on interference-free constraints; the adaptive optimization module is used to optimize the triplet strategy through an online learner and output a channel allocation strategy carrying a channel allocation plan, load balancing instructions and network optimization rules. The operational data includes at least wireless environment data, node status data, and user behavior data. The channel allocation strategy includes at least a channel allocation plan, load balancing instructions, and network optimization rules.

8. A charging pile self-organizing network system as described in claim 6, characterized in that: The master node connection module is specifically used for: After the first charging pile is powered on and started, it automatically performs initialization operations including at least hardware self-test and status confirmation, software system initialization, communication module initialization, and charging interface pre-test. It obtains the current timestamp as the request time, obtains the device serial number of the device, encrypts the device fingerprint, device serial number and request time into ciphertext information using the SM9 algorithm, calculates the second hash value of the ciphertext information using the SM3 algorithm, generates a connection request based on the ciphertext information and the second hash value, and uploads the connection request to the server through the 5G communication unit of the wireless communication module as the master node. The server parses the received connection request to obtain encrypted information and a second hash value. After verifying the integrity of the encrypted information using the second hash value, it decrypts the encrypted information using the SM9 algorithm to obtain the device fingerprint, device serial number, and request time. After verifying the timeliness of the request, it matches the corresponding MAC address and manufacturing date from a preset device management table using the device serial number. After verifying the legality of the device fingerprint based on the device serial number, MAC address, and manufacturing date, it establishes a connection with the charging pile.

9. A charging pile self-organizing network system as described in claim 6, characterized in that: The wireless mesh networking module is specifically used for: After the charging pile is powered on and started, it automatically performs initialization operations and, as a child node, creates a pair of public and private keys based on the RSA algorithm, generates a first random number, obtains the current timestamp as the first authentication time, and uses the device fingerprint of the local machine as the sub-fingerprint. The child node encrypts the sub-fingerprint, the first random number, and the first authentication time using a private key to obtain ciphertext data A1. It then encrypts the ciphertext data A1 and the public key using the SM9 algorithm to obtain ciphertext data A2. Finally, it shifts each character of the ciphertext data A2 cyclically to the right by 3 bits to obtain ciphertext data A3. Based on the ciphertext data A3, it generates the first authentication request and sends it to the master node. The master node parses the received first authentication request to obtain ciphertext data A3, shifts each character of ciphertext data A3 to the left by 3 bits to obtain ciphertext data A2, decrypts ciphertext data A2 into ciphertext data A1 and public key using the SM9 algorithm, decrypts ciphertext data A1 with the public key to obtain sub-fingerprint, first random number and first authentication time, performs timeliness verification using the first authentication time, adds 3 to the value of the first random number to obtain a second random number, obtains the current timestamp as the second authentication time, and uses the device fingerprint of the local machine as the master fingerprint; The master node encrypts the master fingerprint, sub-fingerprint, second random number, and second authentication time using the public key to obtain ciphertext data B1. It then encrypts the ciphertext data B1 into ciphertext data B2 using the SM9 algorithm. Finally, it shifts each character of the ciphertext data B2 cyclically to the right by 5 bits to obtain ciphertext data B3. Based on the ciphertext data B3, it generates a second authentication request and sends it to the sub-node. The child node parses the received second authentication request to obtain ciphertext data B3, shifts each character of ciphertext data B3 cyclically 5 bits to the left to obtain ciphertext data B2, decrypts ciphertext data B2 into ciphertext data B1 using the SM9 algorithm, decrypts ciphertext data B1 using the private key to obtain the master fingerprint, sub-fingerprint, second random number, and second authentication time, performs timeliness verification using the second authentication time, verifies the second random number using the first random number, verifies the decrypted sub-fingerprint using the sub-fingerprint of the local machine, adds 5 to the value of the second random number to obtain the third random number, and obtains the current timestamp as the third authentication time; The child node encrypts the master fingerprint, the third random number, and the third authentication time using a private key to obtain ciphertext data A4. It then encrypts the ciphertext data A4 into ciphertext data A5 using the SM9 algorithm. Finally, it shifts each character of the ciphertext data A5 cyclically to the right by 3 bits to obtain ciphertext data A6. Based on the ciphertext data A6, it generates a third authentication request and sends it to the master node. The master node parses the received third authentication request to obtain ciphertext data A6. It then shifts each character of ciphertext data A6 cyclically 3 bits to the left to obtain ciphertext data A5. Using the SM9 algorithm, it decrypts ciphertext data A5 into ciphertext data A4. The master node then decrypts ciphertext data A4 using the public key to obtain the master fingerprint, the third random number, and the third authentication time. After verifying the timeliness of the third authentication time, it verifies the third random number using the second random number. Finally, it verifies the decrypted master fingerprint using the master fingerprint of the local machine. If the verification passes, the bidirectional authentication operation is completed, and the node accesses the wireless Mesh network managed by the master node through the Wi-SUN communication unit of the wireless communication module.

10. A charging pile self-organizing network system as described in claim 6, characterized in that: The first master node election module is specifically used for: Each charging station within the wireless local area network periodically scans the target signal strength and interference signal strength in the wireless environment, broadcasts the target signal strength, interference signal strength, and the charging station's location within the wireless local area network, and constructs a signal attenuation heatmap based on the target signal strength, interference signal strength, and charging station location of each charging station; the target signal is a 5G signal. Each charging station will broadcast its node status data, including at least network load indicators, device operating status, and location and topology data, within the wireless local area network; Based on the signal attenuation heatmap and node status data, each charging pile re-elects a master node using a weighted scoring algorithm; The second master node election module is specifically used for: The master node and each child node maintain a connection through a multi-path heartbeat packet, which embeds a device fingerprint and a time-series encrypted hash value for real-time node authentication. When the disconnection between the master node and each child node exceeds a preset time threshold, each child node re-elects a master node based on the signal attenuation heatmap and node status data. The network outage fault upload module is specifically used for: When the master node disconnects from the server, based on the signal attenuation heatmap and node status data, it dynamically selects a child node to communicate with the server through a multi-hop relay mode, or establishes an end-to-end encrypted tunnel to upload the network outage fault to the server through a mobile terminal connected to the master node via Bluetooth.

Images