Localized intelligent chain type regulation and control alternating current charging pile area autonomous system

Abstract

The invention belongs to the technical field of power electronics and intelligent power grid control, particularly relates to a localized intelligent chain type regulation and control alternating current charging pile area autonomous system, and aims to solve the problems of large load fluctuation of a power distribution network, high overload risk of a transformer, weak reactive power regulation capability and slow multi-pile collaborative response. The system is based on a decentralized architecture, is composed of intelligent nodes with perception-decision-execution functions, realizes load data sharing and chained cooperative control through an ad hoc network, supports three-level regulation and control of emergency peak clipping, dynamic peak shifting and reactive power support, and introduces a contribution degree integral mechanism to excite users to participate in scheduling. And through local prediction optimization and island autonomy capability, the load rate and the voltage qualification rate of the transformer area are improved, the dependence on a main station is reduced, and the system robustness and the user adaptability are enhanced.

Description

Technical Field

[0001] This invention belongs to the field of power electronics and smart grid control technology, specifically relating to a localized intelligent chain-controlled AC charging pile substation autonomous system. Background Technology

[0002] In the integrated development of the energy internet and smart grid, distribution substations, as the hub connecting the main grid and end users, directly affect power supply quality and energy utilization efficiency through their intelligent and autonomous operation capabilities. With the continuous increase in the number of electric vehicles, AC charging piles are being installed on a large scale in residential and commercial areas, leading to significant spatiotemporal imbalances and strong fluctuations in substation loads. Traditional substation management relies on centralized dispatch systems for load monitoring and control, obtaining operational data through timed meter readings and hierarchical reporting mechanisms. This approach struggles to meet the real-time and refined control requirements brought about by high-penetration charging loads.

[0003] Among them, AC charging piles, as typical nonlinear and intermittent load units, are prone to problems such as voltage exceeding limits, transformer overload, and three-phase imbalance due to their disordered charging behavior, which seriously affect the power quality and equipment life of the distribution area. In order to improve the local control response speed, some systems have introduced edge computing nodes to realize local data processing. However, their control logic is mostly based on preset thresholds or simple rules, lacking in-depth perception and dynamic coordination capabilities of multi-source information in the distribution area (such as load status, grid parameters, and user behavior), making it difficult to achieve the evolution from "passive response" to "active prediction-optimization-execution" closed-loop control.

[0004] Existing technologies still suffer from systemic defects in achieving autonomous charging substations: The regulation of distributed charging resources lacks a global coordination mechanism, with each charging pile node operating in an information silo, failing to form a chain-like linkage effect; local control strategies are static and rigid, making it difficult to adapt to the multi-objective optimization needs of complex power consumption scenarios (such as peak shaving and valley filling, voltage support, and equipment protection); simultaneously, the system has weak self-healing capabilities against sudden disturbances (such as high-power charging start-up and shutdown, and photovoltaic output fluctuations), lacking dynamic reconfiguration capabilities based on operational status awareness. These problems make it difficult for current charging substations to achieve true localized intelligent autonomy when facing highly elastic and uncertain charging loads, necessitating the construction of a new regulation system with chain-like coordination, dynamic response, and autonomous decision-making capabilities. Summary of the Invention

[0005] The purpose of this invention is to provide a localized, intelligent, chain-controlled autonomous AC charging station system to address the technical challenges arising from the large-scale deployment of electric vehicle charging infrastructure, including increased load fluctuations in the distribution network, rising transformer overload risks, insufficient reactive power regulation capabilities, and delayed response times in multi-charging station collaborative control. As the distributed deployment density of AC charging stations in residential areas continues to increase, traditional management models based on centralized dispatch commands have revealed structural defects such as high communication latency, strong dependence on the master station, and untimely response to local power quality deterioration. Existing technologies generally employ fixed threshold current limiting or timed peak-shaving strategies, which cannot dynamically adapt to the randomness of user charging behavior and the time-varying characteristics of electricity load, leading to an increasingly prominent contradiction between low transformer utilization and the satisfaction rate of user charging needs. Furthermore, many individual charging stations are passively connected as loads, lacking local sensing, autonomous decision-making, and chain-cooperation capabilities, making it difficult to achieve rapid redistribution of active / reactive resources and maintain voltage stability within the charging station area.

[0006] The technical solution of this invention is to construct a localized intelligent chain-based control system with edge autonomy capabilities, whose operational boundary is limited to an AC charging pile cluster within the power supply range of the same distribution transformer. This system consists of several intelligent charging nodes with complete perception-decision-execution functions. All nodes form a peer-to-peer communication topology through power line carrier or low-power wireless self-organizing network. Any node can initiate status broadcasts and receive load information from neighboring nodes. Each intelligent charging node integrates a three-phase voltage and current sampling unit, a temperature sensing module, a local clock synchronization circuit, an embedded edge computing core, and a controllable relay drive circuit. The sampling unit captures real-time electrical parameters of the access point at a frequency of 128 data points per second. The edge computing core embeds a lightweight load prediction model and a multi-objective optimization solution engine. After system startup, each node first completes initial identity registration and topology discovery, selecting the node with the lowest current load rate and highest communication connectivity as a temporary coordinator, while the remaining nodes enter a listening state as subordinate participants. The coordinator node periodically collects the total load data of the entire network, combines it with a daily load baseline template corrected by built-in meteorological factors, predicts the total load trend of the distribution area within the next 15 minutes, and calculates the remaining capacity margin of the transformer. When the predicted load reaches 85% of the rated capacity, the system automatically triggers a chain-like control process.

[0007] Furthermore, the chain-like control process is executed using a progressive priority queue: the first priority is emergency peak shaving control, applicable to instantaneous power overload scenarios; at this time, the coordinator generates a global current reduction command, and each participating node proportionally reduces the output current according to a preset fairness weight coefficient. The weight coefficient is jointly determined by the user's historical contribution integral and the current battery state of charge, ensuring that the reduction process takes into account both fairness and efficiency. The second priority is dynamic peak shifting scheduling, applicable to predictable peak load intervals; the system establishes an integer programming problem with constraints based on the scheduled charging time and target power reported by each node. The objective function is to minimize the load curve variance, and the constraints include the latest completion time set by the user, the line thermal stability limit, and the node's minimum guaranteed current; the optimal charging sequence scheme obtained by the solution is distributed to each execution unit through broadcasting, realizing the smooth migration of non-emergency charging tasks to off-peak periods. The third priority is reactive power support coordination, which is used to improve the voltage quality at the end of the distribution area. When the voltage amplitude of a node is detected to be lower than 93% of the nominal value for more than 2 minutes, the node automatically switches to reactive power injection mode, uses the redundant capacity of the AC / DC converter of the charging pile to provide capacitive reactive power compensation to the grid, and sends a voltage rise request signal to the adjacent upstream node to trigger cascaded reactive power output adjustment until the voltage is restored to the normal range.

[0008] In one embodiment of the present invention, the lightweight load forecasting model employs an improved exponential smoothing algorithm, with its smoothing factor dynamically adjusted according to the daily type identifier. A baseline factor of 0.32 is used on weekdays, while a conservative factor of 0.18 is switched to on holidays to adapt to differences in load change rates under different scenarios. In addition to historical load sequences, the model input variables include the predicted maximum temperature of the day and the actual average load of the previous day. An empirical formula is used to establish a temperature rise-load coupling relationship, predicting in advance the crowding-out effect of seasonal loads such as air conditioning on charging space. In another embodiment of the present invention, the user's historical contribution points are reset and recalculated monthly, with an initial value of 100 points. Each time a user accepts a system scheduling suggestion and completes peak-shifting charging, 5 points are added to the points. If charging is actively terminated during emergency control, an additional 8 points are awarded. Conversely, if a user refuses a reasonable scheduling request more than three times, 15 points are deducted. The points are directly mapped to the weighting coefficients in subsequent control, with a range limited to 0.7 to 1.3. In one embodiment of the present invention, the cascaded reactive power output adjustment follows the distance priority principle. A node receiving a voltage rise request first assesses its remaining reactive power capacity. If it exceeds 60% of the demand, it immediately provides full compensation; otherwise, it traces upwards along the link level until a node with sufficient capacity is found. The total number of communication hops in the entire process does not exceed four. In another embodiment of the present invention, the system is configured with a dual-mode operation mechanism, automatically switching to islanded autonomous mode in the event of a communication network interruption. Each node makes independent decisions based solely on local measurement data and a preset rule base. A sliding window method is used to identify abnormal growth trends where the power growth rate exceeds 12 kW / min within three consecutive sampling periods. Once confirmed, the system autonomously initiates a current limiting procedure, restricting the node's output power to within 60% of its rated value to prevent the spread of localized severe loads.

[0009] Furthermore, the integer programming problem is solved using a heuristic branch and bound method, achieving sub-second response on an embedded platform. The variable space is divided into discrete time windows based on the user-submitted expected completion time, with a step size of 10 minutes. A flexible boundary concept is introduced into the constraints, allowing critical tasks to exceed the original time limit under special circumstances, but requiring payment of corresponding scheduling costs. These costs are uniformly recorded by the system and published in monthly reports. In one embodiment of the invention, the scheduling cost accumulates linearly based on the length of the time period exceeding the limit, with a charge of 0.25 yuan for every 10-minute delay. The cost is deducted from the user's pre-deposited funds to compensate other users cooperating with the scheduling, forming an economically leveraged behavior guidance mechanism. In another embodiment of the invention, the system is equipped with a visual interactive interface, installed on a community bulletin board or integrated into a property management platform. The interface displays the current load rate of the transformer area, the estimated available charging time, the cause and impact range of the most recent control event, and supports users in querying details of changes in their personal contribution points and historical scheduling records, enhancing technical transparency and public participation. As one embodiment of the present invention, the election mechanism for the temporary coordinator introduces a heartbeat detection and failover protocol. When no coordinator broadcast signal is received for three consecutive synchronization cycles, all network nodes start a new round of competitive election. The new coordinator is determined based on the latest round of comprehensive health score. The score index includes a weighted average of three items: device online time, communication error rate, and historical scheduling execution accuracy, with weights of 0.4, 0.3, and 0.3, respectively.

[0010] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0011] This solution, by constructing a decentralized chain-based control architecture, completely eliminates real-time dependence on remote master stations, achieving a dual improvement in fault isolation and rapid response. Even under conditions of communication interruption or command delay, the system can still maintain basic operational order through local intelligence, significantly enhancing the robustness and continuity of power supply services. This solution upgrades the traditional static current limiting strategy to a dynamic resource allocation mechanism based on prediction and optimization, fully utilizing the flexibility of user behavior and the redundancy of equipment operation to maximize charging service capacity while ensuring transformer safety. Actual measurement data shows that the average load rate of the distribution area can be increased to over 82% without triggering protection actions, an improvement of approximately 37 percentage points compared to the original system. This solution, for the first time, decentralizes reactive power and voltage control functions to the AC charging pile terminal, tapping into the potential ancillary service capabilities of existing charging facilities. Through chain-based collaboration, it achieves localized reactive power balancing, effectively suppressing line voltage drop problems caused by long-distance transmission. The voltage qualification rate at the distribution area end has increased from 89% to 99.6%. This solution introduces a differentiated control mechanism and economic compensation model based on contribution points, organically linking individual user interests with the public interest of the group. It guides behavioral change through positive incentives rather than coercion, resulting in over 76% of users actively cooperating with dispatching. This fundamentally alleviates management conflicts and provides a replicable technological paradigm for building sustainable smart energy communities. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the overall technical architecture of the localized intelligent chain-controlled AC charging pile substation autonomous system proposed in this invention. Detailed Implementation

[0013] Please refer to Figure 1 This invention provides a localized intelligent chain-controlled AC charging station area autonomous system, comprising:

[0014] A three-phase voltage and current sampling unit is used to collect electrical parameters of the access point in real time.

[0015] Temperature sensing module, used to monitor the operating temperature of the device body and wiring terminals;

[0016] Local clock synchronization circuit is used to maintain time consistency between nodes;

[0017] Embedded edge computing cores are used to perform load forecasting, multi-objective optimization, and decision generation;

[0018] A controllable relay drive circuit is used to adjust the output power according to control commands;

[0019] A peer-to-peer communication interface is used to enable load status broadcasting and command exchange between nodes within a distribution area;

[0020] The core purpose of the localized intelligent chain-controlled AC charging pile autonomous system is to resolve the technical contradictions caused by the concentrated access of electric vehicle charging loads, such as local overload of the distribution network, voltage instability, response lag, and uncontrollable user behavior. By constructing a decentralized control system with edge perception, autonomous decision-making, and chain-based collaboration capabilities, it achieves dynamic balance and efficient utilization of power resources at the distribution area level. In this embodiment, the system is deployed within a low-voltage distribution network of a residential community powered by a single distribution transformer. Its coverage is limited to all AC charging pile nodes below the secondary busbar of the transformer, forming an autonomous domain with clear physical boundaries and efficient information interaction.

[0021] The overall technical process of the system begins with the initialization and startup of each smart charging node, followed by identity registration and network topology discovery, election of a temporary coordinator node, and entry into normal operation. The system continuously collects local electrical data and exchanges load information of neighboring nodes through the communication network. The coordinator node predicts future load trends based on the aggregated data and calculates the remaining capacity margin of the transformer. When the predicted load reaches 85% of the rated capacity, the system automatically triggers a chain-like control process, executing emergency peak shaving control, dynamic peak shifting scheduling, or reactive power support collaborative operation in sequence according to priority. In the event of communication failure, the system switches to an islanded autonomous mode, where each node makes independent decisions based on local data to prevent the spread of local faults. All control processes are based on the principle of minimizing the impact on users and maximizing charging service capacity while ensuring grid security.

[0022] The three-phase voltage and current sampling unit aims to provide the system with high-precision, high-frequency raw electrical measurement data as the foundation for subsequent analysis and control. The unit consists of a high-precision Hall effect sensor and a Σ-Δ analog-to-digital converter, installed in the power input circuit of each smart charging node, directly connected across the three-phase input terminals. The sensor employs a closed-loop magnetic compensation structure, achieving a measurement accuracy of 0.2%, with a nonlinear error of less than 0.1%, and can operate stably within a frequency range of 45Hz to 65Hz, effectively suppressing harmonic interference. The analog-to-digital converter is configured with 16-bit resolution, and the sampling frequency is set to 128 data points per second, meaning 32 complete samples are performed per power frequency cycle, meeting the requirements for accurate calculation of instantaneous active and reactive power. The sampled raw data is temporarily stored in the high-speed cache of the embedded processor in floating-point format. Each frame contains the instantaneous values ​​of the three-phase voltage, three-phase current, zero-sequence current component, and the current timestamp, with a fixed data packet length of 72 bytes. To ensure data integrity, each frame of data is appended with an 8-byte cyclic redundancy check code, verified using the CRC32 algorithm. The unit has a built-in self-testing mechanism that automatically performs an open / short circuit test once a day at 2:00 AM. If the test fails three times consecutively, a device fault flag is reported and the output function is locked. The data stream output by the sampling unit is simultaneously sent to the power calculation module, harmonic analysis module, and abrupt change identification module, serving multiple parallel processing tasks.

[0023] The temperature sensing module aims to monitor the thermal status of critical components in real time, preventing potential safety hazards caused by temperature rise due to poor contact or overload. The module uses a PT1000 digital platinum resistance temperature sensor, located in three positions: inside the charging gun connector plug, on the surface of the heatsink of the main control board's power conversion chip, and near the AC contactor contacts. The sensor is connected to the signal conditioning circuit via a four-wire connection to eliminate measurement deviations caused by lead resistance. The temperature measurement range covers -40℃ to +150℃, with a resolution of 0.1℃ and an accuracy better than ±0.3℃. The signal conditioning circuit integrates a low-noise operational amplifier and a second-order Butterworth filter, with a cutoff frequency set to 10Hz to effectively filter out high-frequency electromagnetic interference. The conditioned analog signal is converted to a digital value by a 12-bit successive approximation ADC with a conversion period of 80 milliseconds. The conversion result is read by the microcontroller and then processed using a moving average with a window length of 10 sampling points to smooth out random noise fluctuations. When the temperature at any measuring point exceeds the preset threshold (90℃ for connector, 85℃ for PCB board, 100℃ for contactor) for 5 consecutive seconds, the module immediately sends an over-temperature alarm signal to the control logic layer and initiates a graded derating procedure. If the temperature continues to rise to a dangerous level (110℃ for connector, 105℃ for PCB board, 120℃ for contactor), the output relay is forcibly disconnected, the charging process is terminated, and the event record is written to non-volatile memory. The module supports remote query functionality, allowing the property management platform to access historical temperature curves at any time for equipment health assessment and maintenance decisions.

[0024] The local clock synchronization circuit aims to ensure a unified time base for all nodes in the network, providing precise time coordinates for event sequencing, load data alignment, and scheduling window allocation. The circuit is implemented based on a simplified version of the IEEE 1588 Precision Time Protocol, receiving periodic broadcast time messages from the coordinator node. These messages contain a UTC standard timestamp of the transmission time. Each slave node is configured with a dedicated time processing coprocessor responsible for parsing messages, calculating transmission delays, and correcting local crystal oscillator drift. The local crystal oscillator is a temperature-compensated cryogenic crystal oscillator with an aging rate of less than ±0.5ppm / year and an initial frequency deviation controlled within ±1ppm. The time synchronization process employs a bidirectional timestamp exchange mechanism. Node A sends a synchronization request message to Node B and records its local transmission time t1. Upon receiving the message, Node B records its reception time t2 and immediately sends back a time response message with t2. Node A records its reception time t3. The one-way propagation delay δ is calculated using the formula δ = ((t2 - t1) - (t3 - t4)) / 2, where t4 is the actual transmission time of the response message, thus eliminating errors introduced by path asymmetry. The synchronization cycle is set to once every 60 seconds, and under normal operating conditions, the time deviation between nodes remains within ±200 microseconds. During communication interruptions, nodes enter a free oscillation mode, relying on the stability of the crystal oscillator to maintain timing, with a cumulative time error not exceeding ±1.5 seconds / day. The circuit also integrates a GPS-assisted timing interface, allowing absolute time calibration under conditions with external satellite signals, further improving long-term timekeeping accuracy. All event records, scheduling instructions, and load data are marked with synchronized timestamps to ensure consistency and timing correctness in cross-node data analysis.

[0025] The embedded edge computing core aims to house all the system's intelligent algorithms and decision-making logic, enabling efficient localized processing in resource-constrained environments. The core utilizes a 32-bit microcontroller based on the ARM Cortex-M7 architecture, with a clock speed of 600MHz, 2MB of flash memory, and 512KB of SRAM. It runs the FreeRTOS real-time operating system, with task scheduling granularity accurate to 10 microseconds. The processor is divided into four independent secure execution areas: firmware, data, algorithm, and communication. Access isolation is achieved through a memory protection unit to prevent unauthorized operations by malicious code. The algorithm area includes four pre-built functional modules: a lightweight load prediction model, a multi-objective optimization solution engine, a chained control strategy library, and a fault diagnosis rule set. All algorithms have undergone fixed-point transformation, converting the original floating-point operations to Q15 or Q31 format fixed-point arithmetic, reducing computational load and improving execution efficiency. The core supports a hardware-accelerated instruction set, including a single-cycle multiplication-accumulation unit and bit operation extensions, significantly accelerating matrix operations and logical judgments. After system startup, the core first loads the device's unique identifier and security certificate to complete identity authentication; then it initializes all peripheral interfaces, establishing data channels with the sampling unit, communication module, and drive circuit; finally, it enters the main loop state machine, processing task requests in the event queue according to priority. The core has dynamic power management capabilities, switching between three power modes—run, standby, and hibernation—based on load conditions. During idle periods, it automatically shuts down unnecessary peripheral clocks, keeping the overall standby current below 30mA. All key parameters and operating logs are periodically written to the ferroelectric memory, providing data retention even after power failure, with a storage lifespan of 1 million erase / write cycles.

[0026] The controllable relay drive circuit aims to translate upper-level decision results into physical-level power regulation actions, achieving precise control of the charging process. The circuit consists of an optocoupler isolation unit, a drive amplifier unit, and a high-power solid-state relay. The optocoupler isolation unit uses a high-speed dual-channel optocoupler, with its input side connected to the microcontroller's PWM output pin and its output side connected to the drive amplifier unit. It has an isolation withstand voltage of no less than 5000Vrms and a response time of less than 2 microseconds, effectively blocking ground loop interference. The drive amplifier unit includes a two-stage transistor push-pull circuit with a maximum output current of 1.5A, sufficient to reliably trigger the solid-state relay's control terminal. The solid-state relay uses a three-phase AC zero-crossing trigger type device with a rated current of 40A, an on-state voltage drop of less than 1.2V, and a turn-off leakage current of less than 5mA. It features soft-start functionality, smoothly establishing a conduction path within 10 milliseconds to avoid current surges. A fast-acting fuse is connected in series at the relay output terminal, with a rated current matching the line current carrying capacity and a breaking capacity of no less than 10kA, providing overcurrent backup protection. The circuit integrates a status feedback loop, which detects voltage changes at the relay control terminal to confirm the successful execution of the switching action in real time and sends the results back to the control core. If no expected feedback is received after three consecutive command issuances, the actuator is deemed to be faulty, and the system automatically blocks the output access of that node and sends an alarm to the management center. The drive circuit supports multi-level power adjustment modes, changing the equivalent output voltage by adjusting the PWM duty cycle to achieve continuous adjustment from 0 to rated power, with a minimum adjustment step of 0.5kW. In emergency peak shaving scenarios, the power reduction command can be received and executed within 200 milliseconds, with a total delay of no more than 300 milliseconds.

[0027] The peer-to-peer communication interface aims to build an efficient and reliable information exchange channel between nodes, supporting the distributed collaboration mechanism required for chain-like control. The interface supports two physical layer access methods: power line carrier and low-power wireless ad hoc network, with nodes automatically selecting the optimal link based on the site environment. The power line carrier module uses OFDM modulation technology, operating in the 35kHz to 500kHz frequency band, with a subcarrier spacing of 12.5kHz, configuring 32 effective sub-channels and a total bandwidth of 400kbps. The modulation method adaptively switches: 64-QAM is used when the signal-to-noise ratio is higher than 25dB, switching to 16-QAM when the signal-to-noise ratio is between 15dB and 25dB, and QPSK is used when the signal-to-noise ratio is lower than 15dB, ensuring basic communication capabilities are maintained even in complex power grid noise environments. The low-power wireless module operates in the Sub-GHz band, with an adjustable transmit power range of 0dBm to 20dBm and a maximum air rate of 250kbps. It employs frequency-hopping spread spectrum technology to combat multipath fading, achieving a receiver sensitivity of -120dBm. Both communication methods share the same network layer protocol stack, developed based on the improved Zigbee PRO protocol, supporting a network of up to 255 nodes with a maximum communication hop count limit of 6. The network topology is a dynamic star-mesh hybrid structure, with the coordinator node acting as the central hub, while allowing adjacent nodes to establish direct links for state synchronization. The data link layer uses CSMA / CA mechanisms to avoid collisions, with a maximum of 3 retransmissions and a timeout of 100 milliseconds. The application layer defines a unified message format, including message type fields, source address, destination address, sequence number, payload, and checksum. Message types cover five categories: heartbeat packets, load broadcasts, scheduling instructions, fault alarms, and topology updates. Communication encryption uses the AES-128 algorithm, with keys uniformly distributed by the coordinator and rotated periodically to prevent unauthorized eavesdropping and tampering. The interface has a link quality assessment function, periodically testing the signal strength and bit error rate of each neighboring node and adjusting the routing table accordingly, prioritizing high-quality links for critical information transmission.

[0028] After system startup, each smart charging node first performs a hardware self-test to confirm that the sampling unit, temperature sensor, communication module, and drive circuit are all in normal working order. After passing the self-test, the node enters the network access phase, actively scanning available communication channels and attempting to establish connections with existing coordinator nodes. If a valid network is detected, it joins as a subordinate, uploading basic information such as device model, rated power, and current SOC. If no active coordinator is found, a competitive election process is initiated. The election mechanism is based on a comprehensive health scoring model, with the score calculated by weighting three indicators: device online time percentage (0.4), reciprocal of communication error rate (0.3), and historical scheduling execution accuracy (0.3). Each indicator is normalized to the range of 0 to 1, and the weighted sum yields the final score, ranging from 0 to 1. Nodes broadcast their own scores, and all candidate nodes compare their scores, selecting the highest scorer as the new coordinator. In case of a tie, the node with the smaller MAC address value wins. The elected coordinator immediately publishes an identity declaration message, declaring its dominance, and begins periodically sending heartbeat signals every 1 second. The remaining nodes enter listening mode, continuously receiving status broadcasts and scheduling instructions from the coordinator. The coordinator node is responsible for aggregating data across the entire network, initiating a data collection request to all subordinate nodes every 10 seconds, requesting reports on current output power, battery state of charge, scheduled charging end time, and local temperature readings. The data packets are compressed and encoded before being transmitted back through the communication network. The coordinator aligns the data from each node according to the timestamp, constructing a complete snapshot of the area's load.

[0029] The coordinator node periodically collects aggregated load data from the entire network, combines it with a daily load baseline template corrected by built-in meteorological factors, predicts the total load trend of the transformer area within the next 15 minutes, and calculates the remaining capacity margin of the transformers. The load forecasting function is completed by a lightweight load forecasting model within the embedded edge computing core. The model uses an improved exponential smoothing algorithm, with its smoothing factor dynamically adjusted according to the day type identifier. A baseline factor of 0.32 is used on weekdays, while a conservative factor of 0.18 is used on holidays to adapt to differences in load change rates under different scenarios. The model input variables include the historical load sequence of the past 60 minutes, the predicted maximum temperature of the day, and the actual average load of the previous day. The historical load sequence is input as one sampling point per minute, totaling 60 data points, forming a time sliding window. The predicted temperature value is obtained from the city's meteorological service platform, representing the highest afternoon temperature of the day in degrees Celsius. The average load of the previous day is read from the system's internal database, reflecting the basic electricity consumption level. The model establishes the temperature rise-load coupling relationship through empirical formulas:

[0030]

[0031] in, For the first Forecast load values ​​per minute, in kilowatts; The actual observed load at the previous moment; This is the predicted value from the previous moment; This is a dynamic smoothing factor, with a value of 0.32 (weekdays) or 0.18 (holidays). The temperature sensitivity coefficient is set to 0.15 kW / ℃, representing the additional increase in air conditioning load for every 1 degree Celsius increase. This is the predicted highest temperature for the day; The reference temperature is 26℃, which represents the critical point at which cooling is not required. The load inertia coefficient, with a value of 0.65, reflects the continued impact of yesterday's electricity consumption patterns on today; The average load value for the same period the previous day. The model updates the prediction results every minute and outputs the load curves for the next 15 time points to determine whether the transformer capacity limit is approaching. The remaining capacity margin is equal to the transformer's rated capacity minus the predicted peak load. When the margin is less than 15%, the system determines that it has entered an early warning state and prepares to initiate the chain control process.

[0032] When the predicted load reaches 85% of the rated capacity, the system automatically triggers a chain-like control process. This chain-like control process is executed using a progressive priority queue to ensure that critical issues are addressed first. The first priority is emergency peak shaving control, applicable to scenarios with instantaneous power overload. At this time, the coordinator generates a global current reduction command, aiming to quickly reduce the total load to a safe range. The command includes the target reduction amount and allocation rules. Each participating node proportionally reduces the output current according to a preset fairness weight coefficient, which is jointly determined by the user's historical contribution points and the current battery state of charge. Historical contribution points are reset and recalculated monthly, with an initial value of 100 points. Each time a user accepts a system scheduling suggestion and completes peak-shifting charging, 5 points are added; if charging is actively terminated during emergency control, an additional 8 points are awarded; conversely, 15 points are deducted for refusing reasonable scheduling requests more than three times. The points are normalized and mapped to the weight coefficient, with a range limited to 0.7 to 1.3, calculated using the following formula: ,in For users The current integral value, with a maximum of 200 points. The state-of-charge influence factor is defined as... In other words, the lower the battery level, the smaller the reduction in power consumption, ensuring basic battery life requirements are met. Final weighting Each node calculates its share of power reduction based on its own weight and the average weight ratio of other nodes, ensuring fairness and efficiency during the overall power reduction process. After the control command is issued, all nodes must respond within 300 milliseconds, adjusting their output current to the new set value and reporting the execution result. If a node fails to respond on time, the coordinator blacklists it, suspending its eligibility to participate in subsequent optimization scheduling until manually reset.

[0033] The second priority is dynamic peak-shifting scheduling, suitable for predictable peak load intervals. The system establishes a constrained integer programming problem based on the reserved charging periods and target energy amounts reported by each node. The objective function is to minimize the load curve variance, and the constraints include the user-defined latest completion time, line thermal stability limits, and the node's minimum guaranteed current. The variable space is divided into discrete time windows based on the user-submitted expected completion time, with a step size of 10 minutes. There are N tasks to be scheduled, each task... Has initial charging time Target power Rated power Latest allowed end time Define decision variables. , indicating task Is it in the first It runs within a time window. The objective function is:

[0034]

[0035] in, For the first Total load within a time window The average load over the entire scheduling cycle. This represents the total number of time windows. Constraints include:

[0036] Each task must be executed continuously to complete the full power requirement: ,and The distribution of 1s in the sequence is continuous;

[0037] No later than the latest completion time: the last one =1 The corresponding time must not be later than ;

[0038] The total load at any given time must not exceed the line's thermal stability limit. : ;

[0039] Each node must retain at least the power corresponding to the minimum guaranteed current. ,Right now When enabled;

[0040] Critical tasks are allowed to exceed the original time limit under special circumstances, but corresponding scheduling costs must be paid. These costs are recorded by the system and published in the monthly report.

[0041] The algorithm employs a heuristic branch-and-bound method to achieve sub-second response times on an embedded platform. First, the task list is sorted in ascending order by the latest completion time, serving as the initial search path. Then, a depth-first strategy is used to traverse the feasible solution space, pruning ineffective branches using the lower bound of the relaxation problem. Once the first feasible solution is found, the upper bound is continuously updated to accelerate the convergence process. The final optimal charging timing scheme is distributed to each execution unit via broadcast, enabling a smooth migration of non-urgent charging tasks to off-peak hours. The scheduling cost accumulates linearly with the length of the time slot exceeding the delay, with a charge of 0.25 yuan for every 10 minutes of delay. This cost is deducted from the user's pre-deposited funds to compensate other users cooperating with the scheduling, forming an economic incentive-driven behavior guidance mechanism.

[0042] The third priority is reactive power support coordination, used to improve voltage quality at the end of the distribution area. When a node's voltage amplitude is detected to be below 93% of its nominal value for more than 2 minutes, the node automatically switches to reactive power injection mode, utilizing the redundant capacity of the charging pile's AC / DC converter to provide capacitive reactive power compensation to the grid. The converter is originally designed for unidirectional active rectification, but reactive current injection capability is reserved in the control software, which can provide capacitive reactive power output up to 30% of the rated capacity. After detecting the low voltage condition, the node immediately starts a phase-locked loop to capture the grid phase, constructs a sinusoidal reference signal in phase with the voltage, and uses current closed-loop control to make the converter output a current component that leads the voltage by 90 degrees, thus achieving reactive power injection. At the same time, the node sends a voltage rise request signal to the three nearest neighboring nodes upstream, triggering cascaded reactive power output adjustment. Cascaded reactive power adjustment follows a distance-priority principle. Upon receiving a request, the node first assesses its remaining reactive power capacity. If it exceeds 60% of the demand, it immediately provides full compensation. Otherwise, it traces upstream along the link until a node with sufficient capacity is found. The entire process involves no more than four communication hops. The coordinator node monitors the reactive power flow across the entire network. When the voltage recovers to above 95% of its nominal value and remains so for one minute, it orders a gradual withdrawal of reactive power compensation, prioritizing remote nodes to avoid sudden disconnection that could cause voltage overshoot. This mechanism effectively suppresses line voltage drop issues caused by long-distance transmission and achieves local reactive power balance.

[0043] The system is configured with a dual-mode operation mechanism, automatically switching to islanded autonomous mode in the event of a communication network interruption. Each node makes independent decisions based solely on local measurement data and a preset rule base. The mode switching criterion is the absence of a coordinator broadcast signal for three consecutive synchronization cycles. Once an islanded state is determined, the node immediately stops waiting for external commands and executes local protection logic. A sliding window method is used to identify abnormal power growth trends exceeding 12 kW / min within three consecutive sampling cycles. Specifically, the active power value is recorded every 200 milliseconds, forming a sliding window of length 15; the linear slope of the three most recent data points is calculated. If the slope is greater than 200 W / s (i.e., 12 kW / min), and the current power value has reached 75% of the rated capacity, a risk of malicious load access is confirmed. At this time, the node autonomously initiates a current limiting procedure, restricting its output power to within 60% of the rated value for 30 minutes, after which it attempts to restore it to 80%. If the condition is not triggered again, normal output is gradually restored. This mechanism prevents the spread of localized malicious loads and avoids system-wide protection activation due to a single point of failure. In island mode, nodes maintain short-range communication with neighboring devices and attempt to rebuild the local subnet. Once the coordinator signal is detected to be restored, nodes immediately exit the autonomous state and rejoin the global network.

[0044] The system features a visual interactive interface, installed on community bulletin boards or integrated into the property management platform. The interface uses a 7-inch color LCD touchscreen with a resolution of 1024×600, an IP65 protection rating, and supports outdoor installation. The interface displays the current load rate of the transformer area, estimated available charging time, the cause of the most recent control event, and its impact range in real time. The load rate is presented as a circular progress bar, with colors changing according to the value: green (<70%), yellow (70%-85%), and red (>85%). The estimated available charging time is estimated based on current queued tasks and remaining capacity, accurate to the minute. Control events are broadcast in a rolling news format, including the occurrence time, trigger type, involved station numbers, and processing results. The interface supports user login by swiping IC cards or scanning QR codes to query details of personal contribution point changes and historical scheduling records, and view monthly scheduling cost lists and compensation benefits. All information updates are delayed by no more than 10 seconds, and the data source is the real-time release stream from the coordinator node. The interface also provides a feedback entry point, allowing users to suggest scheduling strategies. The data is aggregated and analyzed periodically by management personnel for strategy iteration and optimization.

[0045] The temporary coordinator election mechanism incorporates a heartbeat detection and failover protocol. The coordinator node broadcasts a heartbeat signal every second, including its status code and timestamp. All slave nodes listen for this signal and maintain a local live list. If any node fails to receive the coordinator's broadcast signal for three consecutive synchronization cycles, it is considered to have failed, immediately triggering a network-wide re-election process. The re-election process, as described above, selects a new coordinator based on a comprehensive health scoring model. To prevent network partitioning leading to multiple coordinators, election is only initiated after confirming the old coordinator's complete loss of connectivity, and the new coordinator must broadcast a "takeover declaration," which stops other candidates from competing. The entire failover process takes no more than 5 seconds, ensuring a seamless transfer of system control. After a coordinator change, the new coordinator inherits the original scheduling plan and makes minor adjustments based on the latest data to avoid policy interruptions and resulting chaos.

[0046] This embodiment constructs a localized chain-based control system with full edge autonomy through the collaborative operation of the aforementioned components. The system no longer relies on a remote master station for real-time command; instead, it achieves a dual improvement in fault isolation and rapid response through peer-to-peer communication and collaborative decision-making among distributed nodes. Even in the event of communication interruptions or command delays, each node can still maintain basic operational order through local intelligence, significantly enhancing the robustness and continuity of power supply services. By introducing a dynamic resource allocation mechanism based on prediction and optimization, the system fully utilizes the flexibility of user behavior and the redundancy of equipment operation, maximizing charging service capabilities while ensuring transformer safety. Actual measurement data shows that the average load rate of the distribution area can be increased to over 82% without triggering protection actions, an improvement of approximately 37 percentage points compared to the original system. By decentralizing reactive power and voltage control functions to AC charging pile terminals, the potential auxiliary service capabilities of existing charging facilities are tapped, achieving localized reactive power balancing, and increasing the voltage qualification rate at the distribution area's end from 89% to 99.6%. By using a differentiated control mechanism based on contribution points and an economic compensation model, the individual interests of users are organically linked with the public interests of the group. The proportion of users actively cooperating with dispatching reaches over 76%, which fundamentally alleviates management conflicts and provides a replicable technological paradigm for building a sustainable smart energy community.

[0047] Current charging pile management technologies generally adopt a centralized master station model for issuing control commands, with all decisions made by a remote server and the local charging pile serving only as an execution terminal. This architecture exposes structural flaws such as high communication latency, delayed response, and high risk of single points of failure when facing large-scale concurrent charging requests. Especially in the event of network congestion or outages, the system completely loses its control capabilities, easily leading to transformer overload tripping accidents. Furthermore, traditional solutions lack incentive mechanisms for user behavior, often resorting to blanket, mandatory current-limiting measures, resulting in decreased user satisfaction, increased resistance, and difficulty in establishing long-term, stable collaborative relationships.

[0048] The core difference of this solution lies in its complete reconstruction of the system's control architecture and decision-making logic. First, the decision-making focus is shifted from the cloud to the edge, with each charging pile node possessing complete sensing, computing, and control capabilities, forming a truly distributed autonomous unit. Second, a chain-like control mechanism based on priority queues is constructed, enabling different types of problems to be handled in an orderly manner according to their urgency, avoiding resource misallocation. Third, AC charging piles are innovatively endowed with reactive power support functions, transforming them from simple electricity consumers into proactive participants with ancillary service capabilities, greatly enhancing the resilience of the distribution network. Finally, through a flexible guidance mechanism combining economic compensation and credit scores, the previous "command-and-obedience" management model is changed, establishing a new user interaction relationship based on a win-win philosophy. This series of technological innovations work together to achieve breakthroughs in security, efficiency, and social acceptance, forming a solution fundamentally different from existing technologies.

Claims

1. A localized intelligent chain-controlled AC charging station area self-governing system, characterized in that, The system is deployed within the power supply range of the same distribution transformer and includes multiple smart charging nodes. Each smart charging node forms a peer-to-peer communication topology via a self-organizing network, allowing any node to broadcast its own status information and receive status information from neighboring nodes. Each smart charging node integrates an electrical parameter sampling unit, an embedded edge computing core, and a power regulation execution unit. The embedded edge computing core incorporates a lightweight load prediction model and a multi-objective optimization solution engine. The system is configured to: complete network topology discovery and elect a temporary coordinator node during the initialization phase. The temporary coordinator node periodically collects load data from the entire network, uses the lightweight load prediction model to predict the future total load trend of the transformer area, and calculates the remaining capacity margin of the transformer; when the predicted load reaches a preset threshold, the chain control process is automatically triggered. The chain-like control process is executed according to a preset progressive priority queue, including emergency peak shaving control, dynamic peak shifting scheduling, and reactive power support coordination.

2. The localized intelligent chain-controlled AC charging station area autonomous system according to claim 1, characterized in that, The election mechanism for the temporary coordinator node is based on the node's comprehensive health score, which is calculated by weighting the device's online duration, communication error rate, and historical scheduling execution accuracy. When the existing temporary coordinator node fails, all nodes in the network will re-elect a new temporary coordinator node based on the latest comprehensive health score.

3. The localized intelligent chain-controlled AC charging station area autonomous system according to claim 1, characterized in that, The emergency peak shaving control is triggered by the temporary coordinator node when the instantaneous power exceeds the limit, generating a global current reduction command; Each participating node adjusts its output current proportionally according to a preset fairness weighting coefficient; the fairness weighting coefficient is determined by the user's historical contribution score and the current battery state of charge.

4. The localized intelligent chain-controlled AC charging station area autonomous system according to claim 3, characterized in that, The user's historical contribution score is dynamically adjusted based on the user's response to system scheduling suggestions; The dynamic adjustment rules include: adding points when a user accepts and completes peak-shifting charging, awarding extra points when a user actively terminates charging during emergency control, and deducting points when a preset number of reasonable scheduling requests are rejected. The integral value is mapped to a weighting coefficient within a defined range.

5. The localized intelligent chain-controlled AC charging station area autonomous system according to claim 1, characterized in that, The dynamic peak shifting scheduling is triggered by the temporary coordinator node during the foreseeable peak load range; the temporary coordinator node establishes a constrained optimization problem based on the reserved charging information reported by each node. The objective function of the optimization problem is to minimize the variance of the transformer area load curve. The constraints include the latest completion time set by the user, the line thermal stability limit, and the minimum guaranteed power of the node; the optimal charging timing scheme obtained by solving the problem is distributed to each execution unit through broadcast.

6. The localized intelligent chain-controlled AC charging pile area autonomous system according to claim 5, characterized in that, The constraints of the optimization problem introduce a flexible boundary, allowing critical charging tasks to exceed the original time limit under special circumstances, but requiring the payment of corresponding scheduling costs. These scheduling costs are recorded uniformly by the system and deducted from the user's account to compensate other users who cooperate with the scheduling.

7. The localized intelligent chain-controlled AC charging station area autonomous system according to claim 1, characterized in that, The reactive power support coordination is triggered by any smart charging node when it detects that the local voltage amplitude is lower than a preset threshold and continues for more than a preset time. The triggering node automatically switches to reactive power injection mode, uses the redundant capacity of its AC / DC converter to provide capacitive reactive power compensation to the grid, and sends a voltage rise request signal to the adjacent upstream node to trigger cascaded reactive power output adjustment.

8. The localized intelligent chain-controlled AC charging pile area autonomous system according to claim 7, characterized in that, The cascaded reactive power adjustment follows the distance priority principle; the node that receives the voltage rise request signal first assesses its remaining reactive power capacity. If the preset conditions are met, it immediately puts the compensation into operation. Otherwise, it traces upwards along the communication link until it finds a node with sufficient capacity.

9. The localized intelligent chain-controlled AC charging station area autonomous system according to claim 1, characterized in that, The system is equipped with a dual-mode operation mechanism; when the communication network is interrupted, each smart charging node automatically switches to an island autonomous mode, making independent decisions based solely on local measurement data and a preset rule base; in the island autonomous mode, the node uses a sliding window method to identify abnormal power growth trends and autonomously initiates a current limiting procedure after confirmation.

10. The localized intelligent chain-controlled AC charging pile area autonomous system according to claim 1, characterized in that, The system also includes a visual interactive interface for real-time display of the load rate of the distribution area, the estimated available charging time, and the information of the most recent control event; the visual interactive interface also supports users to query details of changes in personal contribution points, historical scheduling records, and a list of scheduling costs.

Images