The rapid global advancement toward carbon neutrality has fundamentally restructured the transportation energy landscape. Electric vehicles (electric vehicles) have transitioned from niche technological demonstrations to mainstream mobility solutions, particularly within the context of China’s massive urban development and environmental policy frameworks. According to the Ministry of Ecology and Environment, vehicle emissions remain a significant contributor to atmospheric particulate matter, with the total pollutant load from mobile sources exceeding millions of tons annually. The expansion of the electric vehicle fleet, which surpassed 36 million vehicles nationwide by late 2025, necessitates a correspondingly robust charging infrastructure. While the total number of charging facilities has reached an impressive scale, the operational integrity of these devices—particularly the direct current (DC) fast-charging stations that dominate highway service areas and urban public depots—presents an increasingly critical technical challenge.
The operational environment for DC charging piles is inherently demanding. These systems are deployed outdoors, subjected to wide temperature fluctuations, humidity, and substantial electrical stress from high-power switching operations within their power modules. A significant portion of the existing charging pile inventory lacks the modern sensing capabilities required for comprehensive real-time condition monitoring. Traditional maintenance paradigms, which rely heavily on periodic manual inspections and offline calibration procedures, generate substantial time delays in fault detection. This reactive approach often only triggers a maintenance response after a user encounters a critical failure, such as a protocol handshake loop failure that prevents charging initiation or a software state deadlock that results in the inability to disconnect the charging gun. The development of an effective, low-cost, and non-invasive monitoring mechanism to retrofit existing piles without altering their core architecture is therefore of strategic importance.
1. System Architecture Design for Charging Infrastructure Monitoring
The fundamental design objective for this online monitoring system was placed on the seamless integration of multi-dimensional sensing with existing hardware constraints. To validate the design from both theoretical and engineering perspectives, an analysis of the physical deployment environment and communication protocols was undertaken. The architecture leverages the Controller Area Network (CAN) bus as the primary data acquisition interface due to its ubiquitous deployment in vehicle-to-charger communications. Simultaneously, a suite of environmental sensors provides a contextual layer of data regarding the operational health of the pile enclosure.
The monitoring terminal was designed to function as a purely passive, high-impedance listener on the SAE J1939-based communication link defined by GB/T 27930. This minimal invasive strategy physically intercepts the S+ (CAN_H) and S- (CAN_L) pins within the charging gun connector without interfering with the handshake signals, enabling data capture without compromising the original charging logic. The system integrates localized data storage, ensuring data continuity during temporary network outages, and wireless transmission through Wi-Fi to establish a chain of custody for data typically absent in legacy charging piles.
| Design Dimension | Existing Research/Commercial Solutions | Identified Limitations | Proposed Improvements |
|---|---|---|---|
| Integration Interface | Dependence on native pile controller APIs and specific cloud platforms | Significant compatibility gaps with early-generation and multi-vendor piles | Deployment of a bypass mechanism operating on the S+/S- CAN link without interfering with control logic |
| Monitoring Dimensions | Primary focus on electrical parameters and fault codes | Lack of integration between environmental risk factors and interaction packets | Fusion of decoded CAN packets, humidity/temperature sensing, smoke detection, and complete log archival |
| Data Processing | High reliance on central cloud platforms for analysis | Data integrity exposed to network fluctuations and downtime | Implementation of edge-level cache with classified storage of raw packet data |
| Deployment Strategy | Optimized for newly constructed piles with manufacturer support | High retrofit complexity and hardware costs for existing infrastructure | Destination-oriented third-party installation with a focus on compactness and low power draw |
| Boundary of Function | Occasional overlap with legal metrological calibration | Potential functional ambiguity regarding measurement authority | Clear positioning as a diagnostic and continuous monitoring agent, supplementary to legal calibration |
2. Physical Layer Communication Interfaces for Electric Vehicles
To establish a robust linkage between the embedded device and the charging network, the study initially delved into the specific electrical and physical characteristics of the charging system, particularly the high-voltage DC charging units. The connection interface between a DC charger and an electric vehicle is standardized. The charging gun connector features a specific pin arrangement for high-voltage transfer and low-voltage digital communication.
| Contact Label | Rated Voltage and Current | Functional Definition in DC Charging |
|---|---|---|
| S+ | 0 to 30V, 2A | CAN High line for charger-BMS Communication |
| S- | 0 to 30V, 2A | CAN Low line for charger-BMS Communication |
| CC1 | 0 to 30V, 2A | Charging connection confirmation |
| CC2 | 0 to 30V, 2A | Charging connection confirmation |
| DC+ | 750V/1000V, up to 250A | Positive high voltage DC output |
| DC- | 750V/1000V, up to 250A | Negative high voltage DC output |
| A+ | 0 to 30V, 2A | Low voltage auxiliary power source positive |
| A- | 0 to 30V, 2A | Low voltage auxiliary power source negative |
| PE | N/A | Protective Grounding |
The interaction protocol requires electric vehicle powertrains to exchange data with the charger. The CAN bus mechanism, which uses differential voltage signaling, is resilient to the harsh electromagnetic environment often generated by the silicon carbide (SiC) or insulated-gate bipolar transistor (IGBT) modules within the charging pile cabinet. The differential signaling logic primarily works on the voltage difference between CAN_High and CAN_Low. Specifically, when in the ‘dominant’ bit state (logic 0), the voltage difference is approximately 2V, while the ‘recessive’ bit state (logic ‘1’) has a voltage difference near 0V. This physical assurance allows for accurate capture of even the weakest ACK frames within a noisy background.
3. Hardware Implementation of the EVSE Monitoring Electronics
The physical realization of the monitoring terminal was strategic, ensuring that its operation in high-interference environments remained unaffected. The constraints of the charging pile cabinet—where heat and noise are prevalent—necessitate a hardware design that incorporates selection criteria for specific industrial and automotive-grade components.
3.1 Microcontroller Unit Selection and Minimal System Construction
Selecting a control core that balances computational throughput, peripheral integration, and power efficiency was essential for the system’s edge processing capabilities. A significant departure from conventional 8-bit architectures was necessary to handle the high frequency of CAN messages and the accompanying Fast Fourier Transform (FFT) calculations for sensor data vibrancy detection. By adopting the GD32F303RET6, I ensure the system runs on an ARM Cortex-M4 core with a clock frequency extending to 120MHz, enabling intricate decoding algorithms to be processed in real-time without lag. This MCU provides 512KB of flash memory and 64KB of SRAM, sufficient to stack a wireless protocol library confidently.
To guarantee clock stability in the face of variable temperature, a high-precision 8 MHz passive crystal oscillator was connected to the main clock input pins. The external capacitor network is meticulously paired with the crystal to form a stable Pierce oscillator configuration, providing the base timing for the CAN controller and for serial bit timing synchronization with the Wi-Fi module.
3.2 Can Transceiver and High-Impedance Listening
For robust tapping of data from the charging pile’s controller, the CA-IF1051HS high-speed CAN transceiver became the pivot of the design. Its compatibility with the 3.3V logic levels of the GD32 MCU simplified the circuit design and avoided the latency introduced by level shifters. Its bus pins can withstand -42V to +42V fault voltages, which resists ground shifts occurring during charge contactors opening and closing.
The MCU and transceiver are linked via designated RXD/TXD pins. The monitoring system is engineered to observe communication without participating in the network’s arbitration or acknowledgment mechanisms, utilizing a high-impedance setup on the physical layer so that the effective capacitance and resistance of the probe do not unbalance the original termination scheme of the vehicle and charger.
To mitigate the risk of the RF signal eroding the bus functions, a physically isolated grounding strategy was deployed. A dedicated star topology was used for power distribution. The power supply circuitry initially steps down the auxiliary voltage (usually 12V or 24V from the pile) via a low dropout regulator. This regulator is specifically configured to manage the impulsive current drawn by the Wi-Fi module’s high-frequency transmission bursts, ensuring voltage ripple management below 50mV, guaranteeing precision for the sensor’s analog-to-digital converter (ADC) quantization.
3.3 Environmental Sensing Integration
In line with the need to identify failures beyond electrical requirements, an on-board Si7021 integrated circuit measured relative humidity and ambient cabinet temperature, while an MQ-2 semiconductor sensor monitored smoke emission. The analog output of the smoke sensor was routed to the MCU’s ADC, allowing graduated alarming thresholds rather than simpler digital logic. The Si7021 was placed strategically apart from the central processor to prevent thermal coupling—a vital necessity as early winding overhead in busbars can be masked by thermal conduction if placed too close to an adjacent processor heatsink.
4. Wireless Data Concentration and Communication Dynamics
A robust wireless link integrates the processing node with the remote top-level monitoring suite. Considering the network topology of EV charging stations—, including underground facilities with significant signal path attenuation—a Wi-Fi interface was chosen. When pitted against cellular options (such as NB-IoT or 4G), Wi-Fi (particularly the 2.4GHz band) presents compelling advantages in cost efficiency and bandwidth. Recurring costs disappear while the high instantaneous bandwidth accommodates dense, high-resolution waveform logging during each charge session.
The design integrates the ESP32-WROOM-32 series module, which contains a dual-core processor operating at up to 240MHz and supports full TCP/IP stack. The management of this stack is independent of the main control MCU, ensuring the CAN data is overrun-proof. UART communication between the GD32 and the ESP32 was implemented with hardware flow control (RTS/CTS) to prevent buffer overrun during extensive data streams. The ESP32’s enable pin is toggled by a GPIO from the main processor, enabling an overcurrent protection response should the module fail to communicate. In critical operations, voltage sag compensation is provided through the use of a large capacitance local to the power amplifier stage, decoupling the radio emissions from the host logic system.
5. Embedded Software Architecture and Data Processing Logic
The software forms the bridge between the physical network and the interpretation of electric vehicle charging behavior. The enterprise software logic is structured around a series of state transitions, governed by a main loop scheduling and interrupt service routines (ISR). A circular ring buffer mechanism accomplishes the intermediate storage of incoming CAN packets to avoid message loss during burst communications (like TPCM_DT sequences) when parsing an electric vehicle’s multi-packet battery parameter declarations (BCP).
5.1 Decoding the Canned Message Data Frames
The protocol used in the charging process, based on SAE J1939’s transport protocol, uses 29-bit identifiers to distinguish up to various power management messages. In analysis, due to the distinction between different types of messages, I parse the CAN bus data in the Python and embedded processing stage. The algorithm isolates them by using the representative Priority, PDU Format (PF), and Destination Address (DA) segments. Prioritized bus access is achieved by affirming the binary traits of each bit position, where more critical messages have numerically lower identifiers.
To achieve absolute data integrity in parsing, the application structure ensures software logic does not misinterpret frame lengths. For the standard data-frame message length (8 bytes), PDU1 (destination-specific addressing) and PDU2 (broadcast) formats are executed through the following logical extraction relations:
$$ PF < 240 \rightarrow PGN = PF \times 256,\quad PS = \text{Destination Address} $$
$$ PF \geq 240 \rightarrow PGN = PF \times 256 + PS,\quad PS = \text{Group Extension} $$
This OSI Layer 2 mapping is the foundation behind successful multi-packet assembly. Transport Protocol (TP) frames, labeled TP.CM (Connection Management) and TP.DT (Data Transfer), enable long messages to be reassembled with a specific packet count cyclical redundancy check to detect lost frames, as evidenced by successful assembly of the Battery Charging Parameters (BCP) message.
$$ \text{SOC}_{\text{Display}} = \left(\sum_{i=1}^{n} \text{DATA\_Packet}_i\right)/\text{Total\_Distance} \times 100\%$$
For parameter coordination, the physical conversion of electric vehicles is done with defined resolutions and offsets. The highest charger output voltage (SPN 2824) arrives as a two-byte integer at a scale factor of 0.1V/bit, resulting in direct scaling to Volts. For current values such as SPN 2826 with a -400A offset, the algorithm accounts for the negative offset provided in the specification:
$$ \text{Voltage}_{V} = \text{RAW}_{dec} \times 0.1 $$
$$ \text{Current}_{A} = (\text{RAW}_{dec} – 4000) \times 0.1 $$
5.2 Sequence of Packet Transmission States
The software stack handles different charging phases. Through definition tables stored in code memory, each flow stage has unique protocol data unit rules. The logic modules inspect the movement of high-level data structures and the switchover from one stage to another:
| Charging Stage | Message Acronym | PGN (Physical Layer) | Priority | Data Length (Bytes) | Transmission path |
|---|---|---|---|---|---|
| Handshake Initiation | CHM / BHM | 002600H / 002700H | 6 | 3 | Charger to EV / EV to Charger |
| Handshake Identification | CRM / BRM | 000100H / 000200H | 6 / 7 | 8 / 41 | Charger to EV / EV to Charger |
| Parameter Configuration | BCP | 000600H | 7 | 13 | Electric Vehicle to Charger |
| Parameter Config. | CTS / CML | 000700H / 000800H | 6 | 7 / 8 | Charger to Electric Vehicle |
| Parameter Config. | BRO / CRO | 000900H / 000A00H | 4 | 1 / 1 | Mutual |
| Charging Execution | BCL | 001000H | 6 | 5 | Electric Vehicle (BMS) to Charger |
| Charging Execution | BCS / CCS | 001100H / 001200H | 7 / 6 | 9 / 8 | Bidirectional |
| Charging Execution | BSM / BMV / BMT | 001300H / 001500H / 001600H | 6 / 6 / 7 | Variable | Electric Vehicle to Charger |
| Termination | BST / CST | 001900H / 001A00H | 4 | 4 / 4 | One-Way Alarms |
| Post-Charge Statistics | BSD / CSD | 001C00H / 001D00H | 6 | 7 / 8 | Statistics Exchange |
6. Upper Computer and Platform Data Management Strategy
The user-facing management framework was designed within Python 3.13 with the PySide6 GUI framework. Distinct nuances exist between real-time monitoring and post-incident analysis. Therefore, the platform architecture utilizes a thread-based split: a back-end receiver running on a dedicated QThread listens to a TCP port for high-throughput UDP packages and writes classification into a segregated SQLite database. Concurrently, an average-case QTimer checks for transitions in vital battery health points each second, refreshing chart elements through synchronous signals.
The visualization output offers distinct views for analytical measurements and low-level records. By providing raw hex logs on a single page, field engineers gain the ability to perform in-depth packet inspection outside the GUI. Data storage aligns with an explicit hierarchy, separating raw stream data and parsed data to minimize database drift:
| Data Table / Processing | Key Schema Fields | Write Frequency / Trigger | Purpose and Indexing |
|---|---|---|---|
| raw_data | Timestamp, Frame Header, Identifier, Data Bytes, Source | Upon each valid CAN frame receipt | Facilitates forensic review of protocol logic at the bit-level |
| charging_data | Voltage, Current, SOC, Battery Temp, Power | Appending every second from parsed queue | Provides longitudinal voltage/current waveforms for the plotting engine |
| sensor_data | Temperature, Humidity, Smoke ppm, Sample Time | Written on a sampling interval while charging | Assessment of thermal runaway threats within the cabinet |
| fault_records | Fault Type, Threshold Value, Frame Snapshot, Status | Inserted when physical threshold or protocol error appears | Handles alarm notification loops and tracks to resolution |
| Data scrubbing | Timestamp validation, range integrity, redundant frame filters | Background thread with cyclic execution | Reduces data echoes to lower CPU and SQLite write contention |
The adoption of graphical representation simplifies the comprehension of electric vehicle battery response, offering immediate insight into the state of charge (SOC) trajectory and temperature progression across the charging timeline. The application interface is developed to show metrics for voltage and current, preserving the intended utility for remote, unskilled staff to monitor the health of electric vehicle fleets at scale.
7. System Integration Tests and Validation Using Test Bench
Prior to integration, the system executed isolated modular checks of hardware and software to ensure the integrity of the whole. A significant portion of this was confirming physical calibration of CAN transceivers. By using standard Windows compatible diagnostic software (CANTest) in a sent-in-loop test, the self-test ensured message transmission latency stayed consistent at roughly 30 ms and performed flawlessly under the RF interference of a nearby Wi-Fi transmitter.
For the experimental validation of the embedded code and remote station, I used the Chroma 8000 Automated Test System (ATS), a high precision programmable measurement unit that emulates different communication protocols of an electric vehicle’s battery management system. The monitoring terminal was wired to the CAN bus accessible in the CCS interface and powered from the test cabinet’s auxiliary power supply. With the Chroma test executing a scripted charging process based on the GB/T 27930 communication standard, the monitor observed the exchange.
The exactness of the telemetry was validated through actual concurrent measurement of the source data and telemetry-decoded output:
| Raw Frame Content (Hex) | Frame Identifier | Message Acronym | Decoded Value Linearization | Physical Meaning Verification |
|---|---|---|---|---|
| D2 0F 3C 0F 02 | 0x181056F4 | BCL | D2 0F -> 0FD2… val = 0x0FD2 * 0.1 | Voltage Demand: 405.0 V |
| E7 0E A0 0F 8E 01 38 | 0x1CEB56F4 | TPCM_DT | 0x0EE7 * 0.1 / 0x0FA0 -400 | Voltage = 381.5V, Current = 0A |
| 3C 00 FF FF | 0x1CEB56F4 | Second Packet | 0x003C = 60; Factor 1 | Estimated Remaining Time: 60 minutes |
The test demonstrated that the system not only captured but instantly rendered a clean curve on the platform display. The data resolution obtained from the embedded terminal matched the raw hex values logged by the Chroma system to an error margin of ±1 least significant bit, satisfying the operating requirements for safety diagnostics. The Micro SD card logging worked precisely in the offline state, storing aligned data files that were digestible after reconnection, without missing any charging messages.
8. Comparative Analysis for Engineering Implementation
Mapping the current output against commercial and research baselines reveals the project’s market positioning. Although the system will not replace legal charge meters, continuous monitoring delivered makes it wholly complementary, achieving a balance between low-cost installation and adequate reliability thresholds for routine warning of environmental hazards.
| Target Feature | Original Platform Interfaces | This Non-invasive Monitoring System |
|---|---|---|
| Continuity of record during transient wifi loss | Not possible in cloud-dependent systems | Local cache archives all CAN streams with time slots for later synchronization |
| Interference with the original EVSE controller | Software integration overhead affecting performance | Bypassing at physical layer yields zero added latency to critical BMS messages |
| Regulatory boundary | High-precision reference for legal verification | State-of-health trend analysis for early screening, not trade settlement |
| Retrofit process | Requires downtime and opening proprietary control link | Plug-in installation available, avoiding modification of pile software integrity |
9. Future Directions and Concluding Insights
The comprehensive design and integration of this monitoring system yields several conclusive deductions. It is technically feasible and practical to graft new generation internet-of-things telemetry to legacy static EV supply equipment. The research achieved a unified and disciplined electromagnetic layout that could withstand high voltage switching events while continually sampling ambient characteristics. The use of robust file buffering helped protect stored data in case of unexpected mains isolation.
Looking to future iterative cycles, the substantial amount of clean state data harvested during each charging event is a high-value asset. This data could function as input to machine learning classifiers that identify atypical operating envelope deviations for early charger module failure. Additional engineering attention could apply an auto-negotiating dynamic CAN bus arbitration that allows multiple diagnostic listeners without distorting protocol timing. The employment of Wi-Fi HaLow could push the coverage envelope for vast charging parks, providing secure deterministic latency at an expanded range. Fully remote accessible charging circuit diagnosis is promoted, ensuring maximum fleet uptime for the electric vehicle ecosystem. The result of this work demonstrates that the design maximizes the service value derived from each connection with the grid.