Interaction control system and method for power module and overcharge pile, storage medium and computer program product

Abstract

The invention provides an interactive control system and method for a power module and an overcharge pile, a storage medium and a computer program product, and the method comprises the steps: obtaining a charging control parameter according to the input time sequence state data of a battery through a charging control parameter prediction model based on deep learning; generating a power instruction according to the charging control parameter and sending the power instruction to the power module; after the power supply module executes the power instruction, an instruction execution result and battery state data are transmitted back to the overcharge pile main control software; and the overcharge pile main control software adjusts the weight of the charging control parameter prediction model according to the received data or regenerates a power instruction based on the latest battery time sequence state data. According to the invention, the charging time can be shortened, the risks of overheating, overcharging and the like can be prevented, different battery states and complex working conditions can be adapted, and the real-time requirement of an overcharging scene on a control instruction and state data can be met.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of electric vehicle supercharging, and particularly relates to an interaction control system and method of a power module and a supercharging pile, a storage medium and a computer program product. BACKGROUND

[0002] With the rapid popularization of electric vehicles, high-power direct-current fast charging technology (such as supercharging piles) has become a key to improving user experience. Efficient, real-time, and safe interaction between supercharging piles and power modules is the core of realizing fast charging and battery protection. The current mainstream interaction technology relies on traditional bus communication (such as CAN) or wireless transmission (such as Wi-Fi), which has significant bottlenecks in real-time performance, dynamic optimization capability, compatibility, and safety, and urgently needs innovation in software, algorithms, and protocol levels.

[0003] One of the existing technologies is that the supercharging pile master control sends preset charging instructions (such as target voltage and current) to the power module through the CAN bus. The power module periodically feeds back battery parameters such as SOC and temperature through the CAN bus. The typical application mode is fixed power output or simple multi-module power distribution, but due to the limited bandwidth of the CAN bus (typically 1 Mbps), concurrent communication of multiple modules is prone to arbitration delay and data conflict, resulting in high delay (> 10 ms) of instruction transmission and feedback, which cannot support millisecond-level response dynamic power adjustment algorithms. Secondly, the communication mechanism and software strategy mainly support static or predefined charging curves, and lack software architecture and algorithm support for online, fast, and intelligent power adjustment according to instantaneous temperature fluctuations, nonlinear SOC changes, and other real-time battery states. Thirdly, the CAN protocol frame formats of different manufacturers are not unified, and integrating devices from different manufacturers requires developing complex software protocol conversion gateways or deep customization, increasing software development and maintenance costs.

[0004] The second existing technology is to use Wi-Fi to realize wireless communication between the power module and the supercharging pile master control and the cloud. It supports remote monitoring and basic remote control of the state. However, wireless signals are easily affected by strong electromagnetic interference generated by high-voltage equipment in the charging station and physical shielding by the metal shell, resulting in data packet loss, communication interruption, or high delay jitter, which seriously affects the reliability and real-time performance of control instructions and state feedback, and is not suitable for high-precision closed-loop control. Open wireless channels are vulnerable to man-in-the-middle attacks (MITM), instruction tampering, or forgery. It is a huge software challenge to efficiently implement strong encryption protocols (such as WPA3) and secure identity authentication mechanisms on resource-constrained embedded devices, and the system security is difficult to guarantee. Moreover, the Wi-Fi protocol itself lacks strict guarantees for transmission delay, and is not suitable for dynamic power control scenarios that have strict requirements for time synchronization and low delay. SUMMARY

[0005] To solve the problems raised in the background art, the application provides an interactive control system and method of a power supply module and a super charging pile, a storage medium and a computer program product.

[0006] The interactive control system of a power supply module and a super charging pile for achieving one of the objectives of the application comprises: The power prediction module is configured to obtain charging control parameters from input time-series state data of the battery by using a deep learning-based charging control parameter prediction model, wherein the charging control parameters include optimal charging power and / or current set value and / or voltage set value. The instruction generation module is configured to generate power instructions according to the charging control parameters and then send the power instructions to the power supply module. The result feedback module is configured to return instruction execution results and battery state data to the super charging pile master control software after the power supply module executes the power instructions, and the super charging pile master control software adjusts the weights of the charging control parameter prediction model or regenerates power instructions based on the latest battery time-series state data.

[0007] The technical effects include accurate prediction and real-time optimization of charging control parameters, solving the problem that the traditional fixed parameter charging mode cannot adapt to the dynamic changes of the battery state, and improving the matching degree of charging parameters and the real-time state of the battery, thereby improving the charging efficiency under the premise of safety.

[0008] Further, the first safety check module is configured to perform safety checks on the battery temperature, and the check method includes requesting to reduce the power to a preset power safety threshold or to suspend charging if the maximum single battery temperature is greater than a set temperature threshold. The technical effects include real-time monitoring and over-limit intervention of the maximum single battery temperature to avoid battery damage or safety accidents caused by excessive temperature, and to strengthen the thermal safety protection of the charging process.

[0009] Further, the second safety check module is configured to perform safety checks on the battery SOC, and the check method includes switching to a constant voltage charging mode if the SOC is greater than a set SOC threshold. After switching to the constant voltage mode, the power prediction module stops outputting the power / current set value and instead outputs a constant voltage set value, and the instruction generation module generates power instructions (voltage is fixed, and current is adaptively adjusted according to the battery state) accordingly. The result feedback module continuously monitors the current attenuation until the termination condition is met. The technical effects include fixing the voltage and reducing the current to ensure that the battery slowly approaches the full charge state, solving the problem that high current charging at the high SOC stage of the battery easily causes overcharging and polarization intensification. The technical effects include reducing battery loss in the later charging stage, prolonging the battery life, and ensuring a smooth end of the charging process.

[0010] Further, after switching to the constant voltage mode, the power prediction module stops outputting the power / current setting value and instead outputs a constant voltage setting value, and the instruction generation module generates a power instruction with a fixed voltage and an adaptively adjusted current according to the battery state, and the result feedback module continuously monitors the current attenuation until the termination condition is met.

[0011] Further, in the instruction generation module, the power instruction is encapsulated through a TSN application layer protocol stack deployed in the super-charging pile master control software. In the result feedback module, the instruction execution result and the battery state data are fed back to the super-charging pile master control software through a TSN protocol stack deployed in the super-charging pile master control software. Technical effects include: encapsulating the power instruction using the TSN application layer protocol stack, using the real-time and deterministic transmission characteristics of TSN (Time-Sensitive Network) to solve the problem of excessive delay or jitter in traditional Ethernet transmission, ensuring low delay and high reliability of power instruction transmission, and meeting the real-time requirements of the super-charging scene for control instructions; extending the TSN protocol stack to feedback data transmission to ensure that the super-charging pile master control software can obtain accurate battery state data and instruction execution results in a timely manner.

[0012] Further, in the result feedback module, if the deviation between the actual value of the battery state data and the target value of the corresponding battery state data in the power instruction is > the first set deviation threshold and ≤ the second set deviation threshold, the weight parameters of the input layer of the charging control parameter prediction model are adjusted; if the deviation between the actual value of the battery state data and the target value of the corresponding battery state data in the power instruction is > the second set deviation threshold, the power instruction is regenerated using the latest feedback battery state data. Technical effects include: the threshold adjustment strategy of small deviation adjusting the model weight and large deviation regenerating the power instruction solves the problem that a single adjustment method is difficult to balance the model optimization accuracy and real-time control response, and considers long-term optimization and short-term safety.

[0013] Further, it further includes a data acquisition and preprocessing module for acquiring SOC, temperature, voltage, and current data provided by the battery BMS in real time through the power module software, and performing filtering, normalization processing, and abnormal data rejection processing on the acquired data.

[0014] Further, the super-charging pile master control software adopts a layered software architecture, which includes: a basic communication layer, a core application layer, and a security software layer; The basic communication layer is used to realize data transmission between the super-charging pile and the power module; a TSN protocol stack is used as the underlying communication protocol, and the protocol stack runs on a hardware platform supporting standard Ethernet; The core application layer is an intermediate connecting layer between the security software layer and the basic communication layer; the core application layer comprises a power prediction module, an instruction generation module and a result feedback module, wherein the power prediction module is internally provided with the deep learning-based charging control parameter prediction model, and the result feedback module comprises an algorithm logic for adjusting the model weight according to the feedback data. The security software layer is located at the uppermost layer, and is used for performing digital signature and verification on the data frames transmitted by the TSN protocol stack by using a lightweight cryptography algorithm, performing identity authentication on the devices accessing the host software, and recording a security event log.

[0015] Technical effects include: the layered architecture solves the problems of maintenance difficulty and insufficient security caused by system function coupling, the basic communication layer guarantees the real-time performance of data transmission, the core application layer realizes the modularization of business logic and facilitates function extension, and the security software layer enhances the anti-attack capability of the system through encryption, authentication and other mechanisms.

[0016] The interaction control method of the power module and the super charging pile for achieving the second purpose of the application comprises the following steps: The deep learning-based charging control parameter prediction model is used to obtain the charging control parameters according to the input time sequence state data of the battery, and the charging control parameters comprise optimal charging power and / or current set value and / or voltage set value. The power instruction is sent to the power module after being generated according to the charging control parameters. After the power module executes the power instruction, the instruction execution result and the battery state data are fed back to the super charging pile host software; and the super charging pile host software adjusts the weight of the charging control parameter prediction model or generates a power instruction again based on the latest battery time sequence state data according to the received data.

[0017] The non-transitory computer readable storage medium for achieving the third purpose of the application has a computer program stored thereon, and the computer program is characterized in that the computer program is executed by a processor to realize the steps of the interaction control method of the power module and the super charging pile.

[0018] The computer program product for achieving the fourth purpose of the application comprises a computer program / instruction, and the computer program / instruction is executed by a processor to realize the steps of the interaction control method of the power module and the super charging pile.

[0019] The beneficial effects of the application include: The application dynamically optimizes charging parameters through a deep learning model, combines temperature, SOC and other safety check mechanisms, effectively prevents risks such as overheating and overcharging while shortening the charging time; based on feedback data, the model weight is adjusted in real time or the instruction is regenerated, which can adapt to different battery states and complex working conditions; the TSN protocol stack is used to realize low-delay transmission of the whole link, meeting the real-time requirements of control instructions and state data in the supercharging scene; and the anti-interference ability and scalability of the system are improved through the layered software architecture. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is a schematic diagram of the system embodiment of the application; Figure 2 is a flowchart of the method embodiment of the application. DETAILED DESCRIPTION

[0021] The following detailed description is used to explain the technical solutions of the application, so that those skilled in the art can understand the application. The protection scope of the application is not limited to the following specific implementation structures. The technical solutions of the application contained in the following specific implementation structures different from the following detailed description are also within the protection scope of the application.

[0022] An interactive control method of a power module and a supercharging pile, as shown in Figure 2 , comprising: 1. Data acquisition and preprocessing The power module software acquires the data of SOC, temperature (including the temperature of key monomers), voltage, current and the like provided by the battery BMS in real time. Filtering, normalization and other preprocessing are performed. The preprocessing process needs to eliminate abnormal data such as values exceeding the reasonable range due to sensor failure, to ensure the effectiveness of the input data.

[0023] 2. Perform basic safety rule check The preprocessed data is subjected to a basic safety rule check, and the rules include: If the highest monomer temperature of the battery > the set temperature threshold T_safe_limit, immediately request to reduce the power to a preset power safety threshold or suspend the charging; in this embodiment, the preferred value of the set temperature threshold T_safe_limit is 45°C, and the preferred value of the preset power safety threshold is 70%.

[0024] If the SOC > the set SOC threshold SOC_constant_voltage_threshold, switch to the constant voltage charging (Cnstant Voltage Mode) mode; in the constant voltage mode, the charging voltage is fixed at the highest safe voltage allowed by the battery, the charging voltage is kept constant, and the charging current is allowed to naturally decay with the increase of the battery capacity.

[0025] Finally, the real-time state data after the basic safety check and the optional historical segment, such as data within the last 10 seconds, are input into the core prediction model.

[0026] 3. Predicted power An LSTM-based neural network model is used as the core prediction engine. The model is trained offline using a large-scale battery charging and discharging experimental dataset on a cloud or high-performance platform. The training goal is to minimize the charging time while meeting the temperature rise and aging constraints.

[0027] In an embodiment, the large-scale battery charging and discharging experimental dataset has the following specific specifications: a total of no less than 100,000 samples, covering 3 types of mainstream power batteries: ternary lithium batteries, iron phosphate lithium batteries, and solid-state batteries; 5 typical aging states: capacity attenuation rate 0%~20%, interval 5%; 8 temperature intervals: -20℃~60℃, interval 10℃; 6 charging rates: 0.5C~3.0C, interval 0.5C; during data collection, complete battery state including SOC, single cell temperature, voltage, and current is recorded every 100ms, and each sample is labeled with an optimal charging power label verified by real vehicle, which can be determined by comparing charging time and battery loss under different powers.

[0028] In an embodiment, the specific structure design of the LSTM neural network model is as follows: 1 input layer, 2 LSTM hidden layers, and 1 fully connected output layer; the input layer dimension matches the time series state data, i.e., 30 time steps x 8-dimensional features, corresponding to 100ms interval, 30 steps, and 8 core parameters; the first LSTM hidden layer contains 64 neurons, and the second LSTM hidden layer contains 32 neurons, both using Tanh activation function; the forget gate, input gate, and output gate all use Sigmoid activation function; the fully connected output layer contains 5 neurons, corresponding to the optimal charging power prediction value for the next 5 100ms periods. In an embodiment, the model is trained using the following self-defined weighted loss function: Loss=α×(charging time deviation) + β×(temperature rise threshold deviation) + γ×(aging rate deviation); wherein α, β, and γ are set weights for balancing charging efficiency and battery protection determined by orthogonal experiment, and the sum of the three is 1, which is dynamically adjusted according to the scene; for example, α=0.3, β=0.4, γ=0.3 to prioritize thermal safety; charging time deviation=(t 实际 -t 目标 ) / t 目标 ; t 实际 is the actual time consumed by the battery to charge from the current SOC to the target SOC; t目标 a preset target charging time; a temperature rise over threshold deviation is a cumulative deviation of a rise of a highest single cell temperature of the battery during the charging process exceeding a safety threshold, the temperature rise over threshold deviation = max(0, ), is an actual rise of the highest single cell temperature during the charging process; is a preset safety temperature rise threshold; an aging rate deviation is a relative deviation of a capacity attenuation rate of the battery caused by the charging process and an expected attenuation rate, the aging rate deviation = (CR 实际 -CR 预期 ) / CR 预期 ; CR 实际 is an actual capacity attenuation rate; CR 实际 = (C0-C1) / (C0xN); C0 is a battery capacity before charging, C1 is a battery capacity after charging, and N is an equivalent cycle number of this time charging; CR 预期 is a preset allowed aging rate.

[0029] In an embodiment, the optimizer selects an Adam optimizer, the initial learning rate is set to 1e-4, and is attenuated to 0.8 times of the previous time every 100 epochs; the training batch size is set to 64, the maximum iteration number is 500 times, and the training is stopped when the validation set loss does not decrease continuously for 20 epochs or the loss value is lower than 1e-3.

[0030] Model deployment: after the trained model is optimized by quantization, pruning, etc., it is deployed on an embedded processor of a power module or a super-charging pile master control.

[0031] In an embodiment, the specific optimization strategy includes: using 8-bit integer quantization to reduce model storage occupation, compressing the storage to 1 / 4 of the original model size, removing 30% redundant neurons through structured pruning to ensure that the inference accuracy loss is ≤5%; using the TensorFlow Lite inference engine, configuring 2 inference threads, limiting the memory occupation to 512MB, and adapting to ARM Cortex-A53 / A72 architecture embedded chips. Run using lightweight inference engines such as TensorFlow Lite and ONNX Runtime. The model input is time series state data, and the output is the recommended optimal charging power or current / voltage setting value.

[0032] In an embodiment, the input time series state data of the model includes: Collecting a multi-dimensional battery state data sequence of 30 time steps continuously at a collection time interval of 100 ms, and each time step contains the following 8 core feature parameters: Battery current SOC value (unit: %); highest single battery temperature (unit: °C); lowest single battery temperature (unit: °C); total battery voltage (unit: V); highest single battery voltage (unit: mV); lowest single battery voltage (unit: mV); charging current (unit: A); charging current fluctuation amplitude within 1 second (unit: A, calculation method: difference between maximum and minimum current values within the second).

[0033] 4. Generating and sending instructions The super-charging pile master control software first integrates the recommended optimal charging power / current / voltage set value output by the LSTM model with the basic safety rules in step 2 according to the basic safety rule priority principle; if the model recommended value meets the safety rules, the instruction is generated based on the recommended value; if the model recommended value exceeds the safety rule limit, the instruction parameters are automatically adjusted to the maximum value defined by the safety rule to ensure the safety of the instruction.

[0034] The generated dynamic power instruction contains the following parameters: target power value, voltage upper limit and current upper limit, instruction effective period, and then the instruction is encapsulated through the above customized TSN application layer protocol stack, the protocol frame structure includes: instruction identification field for uniquely identifying each frame of instruction, parameter field, safety verification field, timestamp field for ensuring timing synchronization. In an embodiment, the safety verification field generates a digital signature for the instruction content using the SHA-256 algorithm to prevent the instruction from being tampered with. After encapsulation, the instruction is periodically sent to the power module.

[0035] 5. Result feedback After the power module receives the dynamic power instruction, it adjusts the charging output to the battery according to the charging control parameters in the instruction, and collects the following feedback data: instruction execution result and new battery state data, the instruction execution result includes: instruction execution state, actual output power value, actual output voltage value, actual output current value, actual value and instruction target value deviation percentage, wherein the instruction execution state includes success and failure, and the reason is marked when it fails, such as hardware failure, parameter overrun; the new battery state data includes: current SOC value, highest / lowest single battery temperature, highest / lowest single battery voltage, current charging current fluctuation amplitude.

[0036] The feedback data is returned to the super-charging pile master control through the TSN protocol stack of the same source as the instruction sending, and the protocol frame structure includes the feedback identification field, the execution result field, the new battery state field, and the data verification field. The feedback period is synchronized with the instruction sending period.

[0037] In an embodiment, after the super-charging pile master control receives the feedback data, the method for adjusting the model parameters includes: 1. Calculate the core deviation index Δ according to the feedback data returned by the power module; Core deviation index Δ = max(ΔP, ΔI, ΔV) Wherein: ΔP is the deviation percentage of actual output power and command target power, i.e. ΔP = |(actual power-target power) / target power| x 100%; ΔI is the deviation percentage of actual output current and command target current, i.e. ΔI = |(actual current-target current) / target current| x 100%; ΔV is the deviation percentage of actual output voltage and command target voltage, ΔV = |(actual voltage-target voltage) / target voltage| x 100%; 2, adjust the feature weight, the adjustment method comprising: 2.1, when the first set deviation threshold 5% < core deviation index Δ ≤ second set deviation threshold 10%: deviation contribution ratio-feature weight quantization adjustment method is used to fine-tune the input layer feature weight of the LSTM model, specifically including: The contribution ratio of each deviation is calculated according to the following formula: Power deviation contribution ratio Kp = ΔP / (ΔP+ΔI+ΔV); Current deviation contribution ratio Ki = ΔI / (ΔP+ΔI+ΔV); Voltage deviation contribution ratio Kv = ΔV / (ΔP+ΔI+ΔV); 2.1.1, when Kp≠Ki≠Kv and max(Kp, Ki, Kv)>Δd, the weight parameters of the charge control parameter prediction model are adjusted according to the following formula or the power command is regenerated based on the latest battery time series state data: W F1’ =W F1 ×(1+Δ×0.01×Δd), W F2’ =W F2 ×(1+Δ×0.01×(1-Δd); Wherein: When max(Kp, Ki, Kv)=Kp, W F1 and W F1’ are the weights before and after adjusting the charge current fluctuation amplitude feature respectively; W F2 and W F2’ are the weights before and after adjusting the total battery voltage feature respectively; When max(Kp, Ki, Kv)=Ki, W F1 and W F1’ are the weights before and after adjusting the charge current feature respectively; W F2 and W F2’ are the weights before and after adjusting the minimum single cell voltage feature respectively; When max(Kp, Ki, Kv)=Kv, W F1and W F1’ These are the weights for the battery total voltage characteristics before and after adjustment, respectively; W F2 and W F2’ These are the weights for the highest single-cell voltage characteristics before and after adjustment, respectively. 2.1.2 When max(Kp, Ki, Kv) < Δd, calculate the following characteristic change rate according to the following formula: Maximum single-cell temperature change rate ΔT = Current maximum single-cell temperature - Maximum single-cell temperature in the previous feedback cycle (unit: °C); SOC change rate ΔSOC = Current SOC - Previous feedback period SOC (unit: %) When ΔT < 2℃ / cycle and ΔSOC < 1% / cycle: W F1’ =W F1 ×(1+Δ×0.01×0.5), W F2’ =W F2 ×(1+Δ×0.01×0.5); When ΔT ≥ 2℃ / cycle and ΔSOC < 1% / cycle: W F1’ =W F1 ×(1+Δ×0.01×0.6), W F2’ =W F2 ×(1+Δ×0.01×0.5); When ΔT < 2℃ / cycle and ΔSOC ≥ 1% / cycle: W F1’ =W F1 ×(1+Δ×0.01×0.5), W F2’ =W F2 ×(1+Δ×0.01×0.6); When ΔT ≥ 2℃ / cycle and ΔSOC ≥ 1% / cycle: W F1’ =W F1 ×(1+Δ×0.01×0.55), W F2’ =W F2 ×(1+Δ×0.01×0.55); W F1 and W F1’ These are the weights for the highest monomer temperature characteristic before and after adjustment, respectively; W F2 and W F2’ These are the weights of the SOC features before and after adjustment, respectively. 2.1.3 When two of Kp, Ki, and Kv are equal and both are at their maximum values, adjust the weight parameters of the charging control parameter prediction model according to the following formula or regenerate the power command based on the latest battery timing data: W F1’ =W F1 ×(1+Δ×0.01×0.4); W F2’ =W F2 ×(1+Δ×0.01×0.3); W F3’ =W F3 ×(1+Δ×0.01×0.3); In the formula, when Kp=Ki>Kv, W F1 and W F1’ are the weights before and after adjustment of the charging current fluctuation amplitude feature respectively; W F2 and W F2’ are the weights before and after adjustment of the total battery voltage feature respectively; W F3 and W F3’ are the weights before and after adjustment of the charging current feature respectively; When Kp=Kv>Ki, W F1 and W F1’ are the weights before and after adjustment of the charging current fluctuation amplitude feature respectively; W F2 and W F2’ are the weights before and after adjustment of the total battery voltage feature respectively; W F3 and W F3’ are the weights before and after adjustment of the highest single cell voltage feature respectively; When Ki=Kv>Kp, W F1 and W F1’ are the weights before and after adjustment of the charging current feature respectively; W F2 and W F2’ are the weights before and after adjustment of the lowest single cell voltage feature respectively; W F3 and W F3’ are the weights before and after adjustment of the highest single cell voltage feature respectively; 2.1.4, when Kp=Ki=Kv, the weight parameters of the charging control parameter prediction model are adjusted according to the following formula or the power instruction is regenerated based on the latest battery timing state data: W F1’ =W F1 ×(1+Δ×0.01×0.3); W F2’ =W F2 ×(1+Δ×0.01×0.3); W F3’ =W F3 ×(1+Δ×0.01×0.2); W F4’ =W F4× (1 + Δ × 0.01 × 0.2); W F1 and W F1’ are the weights of the charging current fluctuation amplitude feature before and after adjustment, respectively; W F2 and W F2’ are the weights of the charging current feature before and after adjustment, respectively; W F3 and W F3’ are the weights of the total battery voltage feature before and after adjustment, respectively; W F3 and W F3’ are the weights of the highest single cell temperature feature before and after adjustment, respectively.

[0038] 2.2, when the core deviation index Δ > the second set deviation threshold 10%: execute model local parameter reset and enhanced input calibration, specifically including: merge the battery state data of the latest N (such as 5) feedback periods with the original M (such as 30) time step input sequence to form an enhanced time sequence of M+N time steps; Let the original bias term of the second hidden layer be B, and the reset bias term B' = B × (1-Δ×0.005), and the value range of B' is limited to [-0.5, 0.5]; input the enhanced time sequence into the model after the reset bias term to generate the calibrated prediction value; if the deviation between the calibrated prediction value and the current actual output value is ≤ the first set deviation threshold 5%, use the reset bias term; if the deviation > the second set deviation threshold 5%, load the deviation correction coefficient table pre-stored in the super-charging pile master control Flash, the correction coefficient table pre-stores the corresponding relationship between Δ and the correction coefficient K: when 10% < Δ ≤ 15%, K = 0.92, when 15% < Δ ≤ 20%, K = 0.88, when Δ > 20%, K = 0.85, multiply the calibrated prediction value by K as the final prediction value.

[0039] In an embodiment, the physical implementation of the control method includes: Through embedded software (C / C++), the software code of the control method realizes data acquisition, preprocessing, safety rule logic, communication protocol stack, and model inference interface.

[0040] The lightweight machine learning inference engine (such as TensorFlow Lite Micro) as the model inference interface part is responsible for efficiently executing the prediction of the LSTM-based parameter prediction model in a resource-constrained environment.

[0041] The model parameters are stored in the non-volatile memory (Flash) of the device.

[0042] In an embodiment, the architecture and communication protocol stack of the embedded software comprises: a basic communication layer, a core application layer, and a security software layer.

[0043] The basic communication layer is at the bottom layer and is the underlying communication guarantee layer of the embedded software, and is also the physical and protocol basis for data transmission of all upper layers. A TSN (Time-Sensitive Network) protocol stack is used as the underlying communication protocol, which runs on a hardware platform supporting standard Ethernet. In some embodiments, the IEEE 802.1AS-rev time synchronization protocol is used to guarantee timing synchronization, and the IEEE 802.1Qbv traffic shaping protocol is used to guarantee high-priority data transmission, thereby providing a low-delay and high-reliability transmission channel for the data frames of the core application layer.

[0044] The core application layer is at the middle layer and is responsible for processing application-level business logic, and is an intermediate connecting layer between the security software layer and the basic communication layer. It includes a power prediction module, an instruction generation module, and a result feedback module. The power prediction module has the deep learning-based charging control parameter prediction model built-in, and the result feedback module includes algorithm logic for adjusting model weights according to feedback data. The data of the core application layer adopts a unified and standardized application layer data frame format, which includes mandatory fields and variable fields. The mandatory fields include instruction type, timestamp, source / destination address, and security signature. The variable fields include target voltage / current, real-time SOC, temperature, and health status. The design of the data frame format ensures that devices from different manufacturers can be parsed through software configuration, achieving compatibility. The core application layer is also used to generate specific business data frames, such as dynamic power instruction frames for super-charging pile master controls and battery status frames for power modules, and to pass the data frames to the security software layer for adding signatures and then to the basic communication layer for transmission.

[0045] The security software layer is at the top layer and is responsible for securing critical data transmitted by the core application layer. It integrates a digital signature mechanism based on lightweight cryptography algorithms such as ECDSA in the application layer protocol. All critical instructions and status frames must carry valid signatures to prevent instruction tampering and data forgery, ensuring interaction security.

[0046] The super-charging pile master software sends dynamic power instruction frames with time sensitivity and high priority labels to designated power modules through the TSN network. The instruction frames contain accurate timestamps and encrypted signatures. The power module software periodically or event-triggered sends battery status frames containing accurate timestamps to the super-charging pile master through the TSN network.

[0047] The software implements a device identity authentication process to ensure that both parties are legitimate. Key management is the responsibility of the software security module.

[0048] It should be understood that the size of the serial number of each step in the above embodiments does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0049] The embodiments of the present application also provide an interaction control system of a power supply module and a super charging pile, as shown in the figure, comprising: Figure 1 a power prediction module: configured to obtain a charging control parameter according to input time sequence state data of a battery by using a deep learning-based charging control parameter prediction model, wherein the charging control parameter comprises: an optimal charging power and / or a current setting value and / or a voltage setting value; an instruction generation module: configured to generate a power instruction according to the charging control parameter and then send the power instruction to the power supply module; a result feedback module: configured to return an instruction execution result and battery state data to a super charging pile master control software after the power supply module executes the power instruction; and the super charging pile master control software adjusts a weight of the charging control parameter prediction model or regenerates a power instruction based on the latest battery time sequence state data according to the received data.

[0050] In an embodiment, a first safety check module is further included, configured to perform safety check on a battery temperature according to collected battery data, and the check method comprises: if the highest single battery temperature is greater than a set temperature threshold, requesting to reduce the power to a preset power safety threshold or to suspend the charging.

[0051] In an embodiment, a second safety check module is further included, configured to perform safety check on a battery SOC according to collected battery data, and the check method comprises: if the SOC is greater than a set SOC threshold, switching to a constant voltage charging mode. After switching to the constant voltage mode, the power prediction module stops outputting the power / current setting value and instead outputs a constant voltage setting value, the instruction generation module generates a power instruction (voltage is fixed and current is adaptively adjusted according to the battery state) according to the constant voltage setting value, and the result feedback module continuously monitors the current attenuation until a termination condition is met.

[0052] In an embodiment, after switching to the constant voltage mode, the power prediction module stops outputting the power / current setting value and instead outputs a constant voltage setting value, the instruction generation module generates a power instruction (voltage is fixed and current is adaptively adjusted according to the battery state) according to the constant voltage setting value, and the result feedback module continuously monitors the current attenuation until a termination condition is met.

[0053] In an embodiment, the power instruction in the instruction generation module is encapsulated by a TSN application layer protocol stack deployed in the super charging pile master control software. In the result feedback module, the instruction execution result and the battery state data are returned to the super charging pile master control software through the TSN protocol stack deployed in the super charging pile master control software.

[0054] ​In an embodiment, in the result feedback module, if the deviation of the actual value of the battery state data from the target value of the corresponding battery state data in the power instruction is > the first set deviation threshold and ≤ the second set deviation threshold, the weight parameter of the input layer of the charging control parameter prediction model is adjusted; if the deviation of the actual value of the battery state data from the target value of the corresponding battery state data in the power instruction is > the second set deviation threshold, the power instruction is regenerated using the latest fed back battery state data.

[0055] In an embodiment, the data acquisition and preprocessing module is further included, which is configured to acquire the SOC, temperature, voltage and current data provided by the battery BMS in real time through the power module software, and perform filtering, normalization processing and abnormal data elimination processing on the acquired data.

[0056] In an embodiment, the super charging pile master software adopts a layered software architecture, which includes: a basic communication layer, a core application layer and a security software layer; The basic communication layer is configured to realize data transmission between the super charging pile and the power module; a TSN protocol stack is adopted as a bottom layer communication protocol, and the protocol stack runs on a hardware platform supporting a standard Ethernet; The core application layer is an intermediate connecting layer between the security software layer and the basic communication layer; the core application layer includes a power prediction module, an instruction generation module and a result feedback module, wherein the power prediction module is internally provided with the deep learning based charging control parameter prediction model, and the result feedback module contains an algorithm logic for adjusting the model weight according to the feedback data; The security software layer is located at the uppermost layer, and is configured to perform digital signature and verification on the data frames transmitted by the TSN protocol stack based on a lightweight cryptography algorithm, perform identity authentication on the devices accessing the master software, and record security event logs.

[0057] The embodiment of the application further provides an interactive control method of a power module and a super charging pile, which includes: The deep learning based charging control parameter prediction model is adopted to obtain charging control parameters according to the input time sequence state data of the battery, wherein the charging control parameters include optimal charging power and / or current set value and / or voltage set value; The power instruction is generated according to the charging control parameters and then sent to the power module; After the power module executes the power instruction, the instruction execution result and the battery state data are fed back to the super charging pile master software; the super charging pile master software adjusts the weight of the charging control parameter prediction model or regenerates the power instruction based on the latest battery time sequence state data according to the received data.

[0058] The embodiment of the present application also provides a computer program product, comprising computer programs / instructions, which, when executed by a processor, implement each step of the method of the present application.

[0059] The embodiment of the present application also provides a non-transitory computer readable storage medium, which stores a computer program, and the computer program comprises program instructions, which, when executed by a processor, implement each step of the method of the present application, which will not be repeated here.

[0060] The computer readable storage medium can be an internal storage unit of the data transmission device or the computer device, for example, a hard disk or a memory of the computer device. The computer readable storage medium can also be an external storage device of the computer device, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card and the like.

[0061] Further, the computer readable storage medium can include both the internal storage unit and the external storage device of the computer device. The computer readable storage medium is used for storing the computer program and other programs and data required by the computer device. The computer readable storage medium can also be used for temporarily storing data to be output or data that has been output.

[0062] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage and the like) containing computer-usable program code.

[0063] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system) and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device implemented in the flowcharts and / or block diagrams. Figure 1 The device for implementing the function specified in one flow or multiple flows and / or blocks. Figure 1 The device for implementing the function specified in one flow or multiple flows and / or blocks.

[0064] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0065] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0066] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

Claims

1. An interactive control system of a power module and a supercharging pile, characterized in that, Comprise: Power prediction module: for using deep learning-based charging control parameter prediction model to obtain charging control parameters according to input battery time-series state data; Instruction generation module: for generating power instructions according to charging control parameters and sending them to power module; Result feedback module: for feeding back instruction execution results and battery state data to super-charging pile master control software after power module executes the power instructions; the super-charging pile master control software adjusts the weight of the charging control parameter prediction model or regenerates power instructions based on the latest battery time-series state data according to the received data.

2. The interactive control system of a power module and a super-charging pile of claim 1, wherein, Also includes a first safety check module for performing safety checks on battery temperature, the checking method comprising: if the highest single battery temperature > set temperature threshold, request to reduce power to a preset power safety threshold or suspend charging.

3. The interactive control system of a power module and a super-charging pile according to claim 1 or 2, characterized in that, Including a second safety check module for performing safety checks on battery SOC, the checking method comprising: if SOC > set SOC threshold, switch to constant voltage charging mode.

4. The interactive control system of a power module and a super-charging pile of claim 1, wherein, In the instruction generation module, the power instructions are encapsulated through the TSN application layer protocol stack deployed in the super-charging pile master control software.

5. The interactive control system of a power module and a super-charging pile according to claim 1 or 4, characterized in that, In the result feedback module, the instruction execution results and battery state data are fed back to the super-charging pile master control software through the TSN protocol stack deployed in the super-charging pile master control software.

6. The interactive control system of a power module and a super-charging pile of claim 1, wherein, In the result feedback module, if the deviation between the actual value of the battery state data and the target value of the corresponding battery state data in the power instruction > first set deviation threshold and ≤ second set deviation threshold, adjust the weight parameters of the input layer of the charging control parameter prediction model; if the deviation between the actual value of the battery state data and the target value of the corresponding battery state data in the power instruction > second set deviation threshold, regenerate power instructions using the latest feedback battery state data.

7. The interactive control system of a power module and a super-charging pile according to claim 1 or 4, wherein, The super-charging pile master control software adopts a layered software architecture, which comprises: Basic communication layer, core application layer and security software layer; The basic communication layer is used to realize data transmission between the super-charging pile and the power module; the TSN protocol stack is used as the underlying communication protocol, which runs on a hardware platform supporting standard Ethernet; The core application layer is an intermediate layer between the security software layer and the basic communication layer; it includes power prediction module, instruction generation module and result feedback module, wherein the power prediction module has the deep learning-based charging control parameter prediction model built-in, and the result feedback module contains algorithm logic for adjusting model weight according to feedback data; The security software layer is located at the topmost layer, which is used to digitally sign and verify the data frames transmitted by the TSN protocol stack based on lightweight cryptography algorithms, authenticate the identity of devices accessing the master control software, and record security event logs.

8. A method for the interaction control of a power module and a super-charge pile of the system of claim 1, characterized in that, Comprise: Using deep learning-based charging control parameter prediction model to obtain charging control parameters according to input battery time-series state data; Generating power instructions according to charging control parameters and sending them to power module; After the power module executes the power instruction, the instruction execution result and the battery state data are returned to the super-charging pile master control software; the super-charging pile master control software adjusts the weight of the charging control parameter prediction model or regenerates the power instruction based on the latest battery timing state data according to the received data. 9.A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by a processor to implement the steps of the interactive control method of the power module and the super-charging pile in claim 8.

10. A computer program product comprising computer programs / instructions, characterized in that, The computer program / instruction is executed by a processor to implement the steps of the interactive control method of the power module and the super-charging pile in claim 8.

Images