Lithium battery charging system

Abstract

The invention relates to a lithium battery charging system, and relates to the technical field of battery charging, the lithium battery charging system comprises a charging module, a detection module and a data processing module, and also comprises a battery health assessment module used for collecting battery cycle index, internal resistance and capacity attenuation parameters, and constructing a battery health state assessment model; the power grid interaction module is used for acquiring real-time electricity price data and a power grid load state; the environment sensing module comprises an environment temperature and humidity sensor and is used for collecting environment temperature and humidity parameters; and the AI prediction module is constructed based on a deep learning network and is used for predicting a battery state change trend. According to the invention, the battery health evaluation module is arranged to collect parameters such as cycle index, internal resistance and capacity fading rate to construct the SOH evaluation model, the problem of single dimension of battery health state evaluation in the prior art is solved, the aging degree of the battery can be accurately identified, and the risk of overcharge and undercharge of the aged battery due to improper power adaptation is avoided.

Description

Technical Field

[0001] This application relates to the field of battery charging technology, and in particular to a lithium battery charging system. Background Technology

[0002] Lithium battery charging systems are core equipment that provides power to lithium batteries. They are widely used in electric vehicles, energy storage power stations, and industrial mobile equipment (such as forklifts and AGVs). By adjusting parameters such as charging power and duration, they can achieve rapid battery charging and lifespan maintenance, playing a key role in new energy power applications.

[0003] Publication No. (CN116061747B) discloses an intelligent time-sharing charging system, which includes a charging module (containing several charging piles), a detection module (used to detect the remaining battery power, loss, temperature and the actual output power value of the charging piles), and a data processing module (including a monitoring unit, a first control unit, a second control unit and a scheduling unit). By constructing the actual output power change curve of the charging piles, analyzing the average slope of its rising segment to divide the charging state, and then adjusting the output power within the remaining charging time for different states, it has a certain charging optimization capability.

[0004] However, this intelligent time-sharing charging system has the following drawbacks: Although it can determine the charging status and adjust the power through power curve slope analysis, the assessment of battery health status (SOH) relies solely on the current battery wear and tear, without integrating aging characteristic parameters such as cycle count, internal resistance change, and capacity decay rate. This makes it impossible to accurately identify the degree of battery aging, which can easily lead to overcharging or undercharging risks due to improper power matching of aging batteries. Furthermore, it does not integrate real-time electricity price and grid load data, making it impossible to prioritize fast charging during off-peak hours and reduce power during peak hours. This increases user electricity costs and makes it difficult to coordinate with the grid to achieve peak-valley load balancing. Simultaneously, the detection module only monitors battery temperature, ignoring the impact of ambient temperature and humidity on the charging process. In extreme environments such as high temperature, high humidity, and low temperature, charging safety and efficiency are difficult to guarantee. In addition, relying on real-time data to adjust power without utilizing historical charging data to build predictive models makes it impossible to optimize charging strategies in advance, resulting in insufficient dynamic adaptability and global optimization capabilities. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a lithium battery charging system that solves the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a lithium battery charging system, comprising a charging module, a detection module, and a data processing module, characterized in that the system further comprises: The battery health assessment module is used to collect battery cycle count, internal resistance, and capacity decay parameters to build a battery health status assessment model. The power grid interaction module is used to obtain real-time electricity price data and power grid load status; The environmental sensing module includes an environmental temperature and humidity sensor for collecting environmental temperature and humidity parameters; The AI ​​prediction module, built on a deep learning network, is used to predict battery state change trends. The data processing module is connected to the battery health assessment module, the power grid interaction module, the environmental perception module, and the AI ​​prediction module, respectively, and is configured as follows: The upper limit of charging power is dynamically adjusted based on the battery health status assessment results. Charging priority periods are determined based on real-time electricity prices and grid load conditions. The charging strategy is adjusted based on ambient temperature and humidity parameters and battery temperature threshold. Adjust the charging control parameters based on the predicted battery state change trend.

[0007] Preferably, the battery health assessment module performs the following steps: Collect the battery cycle count N, current internal resistance R, and capacity decay rate D; The battery health status is calculated using the formula SOH=α·(C / C0)+β·(R0 / R)+γ·(1-D / N), where C is the current capacity, C0 is the nominal capacity, R0 is the initial internal resistance, and α, β, and γ are weighting coefficients. The SOH value is compared with a preset health threshold to generate a health level assessment result. The health level assessment result includes: when SOH ≥ 90%, the health level is excellent, corresponding to a green label. When 80% ≤ SOH < 90%, the product's health level is good, corresponding to the blue label; When 70% ≤ SOH < 80%, the health level is medium, corresponding to a yellow label. When SOH < 70%, the health level is poor, corresponding to a red label; The health level assessment results are used to indicate the health status of the battery and serve as the basis for the data processing module to adjust the upper limit of charging power.

[0008] Preferably, the power grid interaction module performs the following steps: Real-time acquisition of grid time-of-use electricity price data and regional load status; Based on the electricity price fluctuation curve, periods are divided into low electricity price periods, medium electricity price periods, and high electricity price periods; The upper limit of allocable charging power is determined based on the grid load status.

[0009] Preferably, the environmental perception module performs the following steps: Collect ambient temperature T e and relative humidity H; Combined with battery temperature T β Calculate the temperature difference ΔT = T e -T β ; The environmental impact coefficient K = f(ΔT,H) is generated based on the values ​​of ΔT and H.

[0010] Preferably, the AI ​​prediction module performs the following steps: Constructing an LSTM network model involves inputting a historical charging data sequence. The construction of the LSTM network model includes: Collect historical charging data, including at least voltage, current, temperature, and SOC, and perform normalization preprocessing. The preprocessed data is divided into time windows, with each time window containing m consecutive sampling points, forming the input sequence X = [x1, x2, ..., xm]. m ]; Construct an LSTM network that includes an input layer, a hidden layer, and an output layer. The hidden layer contains p LSTM units, and each LSTM unit contains a forget gate, an input gate, and an output gate structure. The model is trained using the Adam optimizer and mean squared error loss function, and the network weight parameters are adjusted using the backpropagation algorithm. The trained model predicts the battery voltage, temperature, and battery health status for the next n time steps. The model predictions include: Input the current charging data into the trained LSTM model to generate a prediction sequence for the next n time steps; The predicted sequence is denormalized to obtain the predicted values ​​of the actual physical quantities; The charging termination threshold and power slope limit are dynamically adjusted based on the prediction results. This dynamic adjustment includes: When the predicted battery voltage V(t) reaches 95% of the preset upper voltage limit Vmax, the charging termination threshold is adjusted from the initial value V0 to V0·(1+k1·(1-SOH)), where k1 is the voltage adjustment coefficient. When the predicted rate of temperature change dT / dt exceeds the safety threshold Tth, the power slope will be limited from the initial value P. s Adjust to P s ·(1-k2·min(1,dT / dt / Tth)), where k2 is the slope adjustment coefficient; Based on the predicted battery health state SOH(t), the switching point between the constant current stage and the constant voltage stage is dynamically adjusted. When SOH < 80%, the switching point SOC is advanced from 80% to 75%.

[0011] Preferably, the data processing module adjusts the upper limit of charging power based on the battery health status assessment results, including: When SOH ≥ 90%, set the power upper limit P. max =P0; When 70% ≤ SOH < 90%, set the power upper limit P. max =P0·(SOH / 90%); When SOH < 70%, set the power limit P. max =P0·0.7, and start the battery maintenance mode.

[0012] Preferably, the data processing module divides charging priority time periods as follows: During periods of low electricity prices, high-priority battery packs are prioritized for fast charging. During periods of medium electricity price, maintain regular charging power; During periods of high electricity prices, only the minimum guaranteed power is provided to battery packs in urgent need.

[0013] Preferably, the data processing module corrects the charging strategy by including: When the ambient temperature T e At temperatures above 40℃, the charging power is reduced to P = P0·(1-k1·(T)). e -40)); When the ambient temperature T e At temperatures below 0°C, the battery preheating program is activated, with a preheating power P. r =k2·P0; When the relative humidity H>85%, increase the charging termination voltage threshold ΔV=k3·H.

[0014] Preferably, the data processing module adjusts the charging control parameters including: The temperature protection threshold is dynamically adjusted based on the battery temperature change rate predicted by LSTM. Based on the predicted SOC change trend, optimize the constant current / constant voltage switching point; When the predicted voltage fluctuation exceeds the threshold, the battery balancing procedure is triggered.

[0015] Preferably, the system further includes a multi-stack collaborative scheduling module, which is configured as follows: Obtain the real-time load rate of each charging station; Based on the priority of the battery to be charged and the real-time load rate, an optimal allocation matrix is ​​calculated and generated. The optimal allocation matrix contains multiple allocation identifiers, each of which is used to indicate whether the corresponding charging pile is allocated charging resources. By calculating load constraints, the sum of the products of the real-time load rate of each charging pile and the corresponding allocation identifier is controlled within the preset maximum load threshold in order to allocate charging resources.

[0016] In summary, this application includes at least one of the following beneficial technical effects: 1. By setting up a battery health assessment module to collect parameters such as cycle count, internal resistance, and capacity decay rate to construct a SOH assessment model, the problem of the single dimension of battery health status assessment in existing technologies is solved. It can accurately identify the degree of battery aging and avoid the risk of overcharging or undercharging due to improper power matching of aging batteries. 2. By acquiring real-time electricity price and grid load data through the grid interaction module and dividing charging priority time periods by the data processing module, the problem of lack of multi-pile coordination and grid interaction capability in existing technologies is solved. It can prioritize fast charging during off-peak electricity price periods and reduce power during peak electricity price periods, thereby reducing users' electricity costs and cooperating with the grid to achieve peak-valley load balance. 3. By collecting ambient temperature and humidity parameters through the environmental sensing module, the data processing module adjusts the charging strategy accordingly, solving the problem of insufficient environmental adaptability of existing technologies and ensuring charging safety and efficiency in extreme environments such as high temperature, high humidity, and low temperature; 4. By building a model based on an LSTM network through the AI ​​prediction module, the battery state change trend is predicted using historical charging data, solving the problem of weak intelligent prediction and control capabilities of existing technologies, and can optimize the charging strategy in advance, improving the system's dynamic adaptability and global optimization capabilities. Attached Figure Description

[0017] Figure 1 This is the system architecture diagram of the present invention. Detailed Implementation

[0018] The following is in conjunction with the appendix Figure 1 This application will be described in further detail below.

[0019] Example: This invention provides a lithium battery charging system, including a charging module, a detection module, and a data processing module, characterized in that the system further includes: The battery health assessment module is used to collect battery cycle count, internal resistance, and capacity decay parameters to build a battery health status assessment model. The power grid interaction module is used to obtain real-time electricity price data and power grid load status; The environmental sensing module includes an environmental temperature and humidity sensor for collecting environmental temperature and humidity parameters; The AI ​​prediction module, built on a deep learning network, is used to predict battery state change trends. The data processing module is connected to the battery health assessment module, the power grid interaction module, the environmental perception module, and the AI ​​prediction module, respectively, and is configured as follows: The upper limit of charging power is dynamically adjusted based on the battery health status assessment results. Charging priority periods are determined based on real-time electricity prices and grid load conditions. The charging strategy is adjusted based on ambient temperature and humidity parameters and battery temperature threshold. Adjust the charging control parameters based on the predicted battery state change trend.

[0020] Specifically, a multi-dimensional data acquisition and processing system is constructed by adding modules for battery health, grid interaction, environmental perception, and AI prediction. The battery health module integrates aging characteristics such as cycle count and internal resistance, overcoming the limitations of assessing only loss and making power adjustment more closely match the actual battery life. The grid interaction module connects to real-time electricity prices and loads, enabling differentiated charging during peak and off-peak hours, reducing user costs and contributing to grid load balancing. The environmental perception module compensates for the shortcomings of only monitoring battery temperature, combining dynamic correction strategies based on ambient temperature and humidity to improve charging safety in extreme environments. The AI ​​prediction module predicts battery status based on historical data, upgrading charging control from "real-time feedback" to "advance prediction," enhancing the system's dynamic adaptability. These four modules, in collaboration with the data processing module, comprehensively optimize the charging process from four dimensions: battery health, energy cost, environmental adaptation, and intelligent control, addressing the shortcomings of existing technologies in terms of accuracy, collaboration, adaptability, and intelligence.

[0021] The battery health assessment module performs the following steps: Collect the battery cycle count N, current internal resistance R, and capacity decay rate D; The battery health status is calculated using the formula SOH=α·(C / C0)+β·(R0 / R)+γ·(1-D / N), where C is the current capacity, C0 is the nominal capacity, R0 is the initial internal resistance, and α, β, and γ are weighting coefficients. The SOH value is compared with a preset health threshold to generate a health level assessment result, which includes: When SOH ≥ 90%, the product's health level is excellent, corresponding to a green label; When 80% ≤ SOH < 90%, the product's health level is good, corresponding to the blue label; When 70% ≤ SOH < 80%, the health level is medium, corresponding to a yellow label. When SOH < 70%, the health level is poor, corresponding to a red label; The health level assessment results are used to indicate the health status of the battery and serve as the basis for the data processing module to adjust the upper limit of charging power.

[0022] Specifically, the battery health assessment module collects core aging parameters such as cycle count, internal resistance, and capacity decay rate to construct a multi-dimensional health model, overcoming the limitations of existing technologies that rely solely on loss measurement. By grading batteries (Excellent, Good, Average, Poor) and associating them with visual indicators, the battery health status becomes more intuitively identifiable. The data processing module dynamically adjusts the power limit based on this grade—healthy batteries can be fitted with higher power to meet fast charging needs, while aging batteries have their power reduced and a maintenance mode activated to prevent damage from overcharging or undercharging, thus extending battery lifespan from the root and solving the lifespan degradation problem caused by insufficient accuracy in battery health assessment.

[0023] The power grid interaction module performs the following steps: Real-time acquisition of grid time-of-use electricity price data and regional load status; Based on the electricity price fluctuation curve, periods are divided into low electricity price periods, medium electricity price periods, and high electricity price periods; The upper limit of allocable charging power is determined based on the grid load status.

[0024] Specifically, the grid interaction module breaks through the limitations of existing technologies that lack grid coordination by capturing real-time electricity price fluctuations and regional load changes. It divides peak and off-peak periods based on electricity prices, releasing higher power during low-price periods to prioritize fast charging needs, while limiting power during high-price periods and ensuring only emergency power supply. This not only meets users' demands for reduced electricity costs but also assists the grid in balancing peak and off-peak loads through power limit regulation, achieving a two-way benefit for both users and the grid. This solves the cost and load problems caused by the lack of coordination between multiple charging stations and the grid.

[0025] The environmental perception module performs the following steps: Collect ambient temperature T e and relative humidity H; Combined with battery temperature T β Calculate the temperature difference ΔT = T e -T β ; The environmental impact coefficient K = f(ΔT,H) is generated based on the values ​​of ΔT and H.

[0026] Specifically, the environmental sensing module adds temperature and humidity data collection to compensate for the shortcomings of existing technologies that only monitor battery temperature. High temperatures exacerbate battery thermal degradation, high humidity can easily induce insulation problems, and low temperatures lead to a decrease in charging efficiency. By collecting ambient temperature and humidity data and coupling it with battery temperature to calculate the impact coefficient, the data processing module can make adjustments to the charging strategy—such as reducing power at high temperatures, preheating at low temperatures, and adjusting voltage thresholds at high humidity. This ensures that the system can still guarantee charging safety and efficiency in complex scenarios such as open-air storage yards and cold / tropical environments, addressing the core pain point of insufficient environmental adaptability.

[0027] The AI ​​prediction module performs the following steps: Constructing an LSTM network model involves inputting a historical charging data sequence. The construction of the LSTM network model includes: Collect historical charging data, including at least voltage, current, temperature, and SOC, and perform normalization preprocessing. The preprocessed data is divided into time windows, with each time window containing m consecutive sampling points, forming the input sequence X = [x1, x2, ..., xm]. m ]; Construct an LSTM network that includes an input layer, a hidden layer, and an output layer. The hidden layer contains p LSTM units, and each LSTM unit contains a forget gate, an input gate, and an output gate structure. The model is trained using the Adam optimizer and mean squared error loss function, and the network weight parameters are adjusted using the backpropagation algorithm. The trained model predicts the battery voltage, temperature, and battery health status for the next n time steps. The model predictions include: Input the current charging data into the trained LSTM model to generate a prediction sequence for the next n time steps; The predicted sequence is denormalized to obtain the predicted values ​​of the actual physical quantities; The charging termination threshold and power slope limit are dynamically adjusted based on the prediction results. This dynamic adjustment includes: When the predicted battery voltage V(t) reaches 95% of the preset upper voltage limit Vmax, the charging termination threshold is adjusted from the initial value V0 to V0·(1+k1·(1-SOH)), where k1 is the voltage adjustment coefficient. When the predicted rate of temperature change dT / dt exceeds the safety threshold Tth, the power slope will be limited from the initial value P. s Adjust to P s ·(1-k2·min(1,dT / dt / Tth)), where k2 is the slope adjustment coefficient; Based on the predicted battery health state SOH(t), the switching point between the constant current stage and the constant voltage stage is dynamically adjusted. When SOH < 80%, the switching point SOC is advanced from 80% to 75%.

[0028] Specifically, the AI ​​prediction module, based on an LSTM network, mines trends and patterns in historical charging data, overcoming the lag inherent in existing "real-time feedback" control technologies. By constructing a network with structures such as forget gates and optimizing training parameters, it accurately predicts future changes in voltage, temperature, and state of equilibrium (SOH). Based on the prediction results, it adjusts the charging termination threshold (to avoid sudden voltage spikes), power slope (to suppress rapid temperature changes), and charging path switching points in advance (to adapt to the characteristics of aging batteries). This upgrades charging control from "passive response" to "active prediction," enhancing the system's adaptability to changes in battery state and addressing the weakness in intelligent prediction and control.

[0029] The data processing module adjusts the upper limit of charging power based on the battery health status assessment results, including: When SOH ≥ 90%, set the power upper limit P. max =P0; When 70% ≤ SOH < 90%, set the power upper limit P. max =P0·(SOH / 90%); When SOH < 70%, set the power limit P. max =P0·0.7, and start the battery maintenance mode.

[0030] Specifically, the data processing module sets differentiated power limits based on the battery's health level. Healthy batteries are adapted to high power to meet fast charging needs, moderately aged batteries have their power reduced to balance efficiency and lifespan, and severely aged batteries have their power further limited and a maintenance mode activated. Through dynamic matching of power and health status, fast charging efficiency is ensured while preventing accelerated degradation of aging batteries due to power overload.

[0031] The data processing module divides charging priority time periods as follows: During periods of low electricity prices, high-priority battery packs are prioritized for fast charging. During periods of medium electricity price, maintain regular charging power; During periods of high electricity prices, only the minimum guaranteed power is provided to battery packs in urgent need.

[0032] Specifically, the data processing module, in conjunction with the grid interaction module, constructs a collaborative mechanism of "time period-priority-power" based on time period division. During low-price periods, high-power is released to prioritize high-priority batteries, maximizing the utilization of low-cost energy; during medium-price periods, normal power is maintained to balance cost and efficiency; and during high-price periods, only emergency power is guaranteed, with strict energy consumption control. This strategy reduces user electricity costs, ensures core needs through priority scheduling, and assists the grid in peak shaving and valley filling, addressing the problem of missing charging strategies across multiple time periods.

[0033] The data processing module corrects the charging strategy by including: When the ambient temperature T e At temperatures above 40℃, the charging power is reduced to P = P0·(1-k1·(T)). e -40)); When the ambient temperature T e At temperatures below 0°C, the battery preheating program is activated, with a preheating power P. r =k2·P0; When the relative humidity H>85%, increase the charging termination voltage threshold ΔV=k3·H.

[0034] Specifically, the data processing module uses temperature and humidity data from the environmental sensing module to adjust strategies for extreme environmental characteristics: High temperatures significantly increase the risk of battery thermal runaway, so reducing power output can decrease heat generation; low temperatures reduce battery activity, so preheating improves charging efficiency; high humidity easily leads to insulation failures, so increasing the termination voltage threshold avoids undercharging risks. Through dynamic adaptation of the environment and strategies, the system can still ensure charging safety and efficiency in high-temperature, high-cold, and high-humidity scenarios, solving safety and efficiency problems caused by poor environmental adaptability.

[0035] The data processing module adjusts the charging control parameters including: The temperature protection threshold is dynamically adjusted based on the battery temperature change rate predicted by LSTM. Based on the predicted SOC change trend, optimize the constant current / constant voltage switching point; When the predicted voltage fluctuation exceeds the threshold, the battery balancing procedure is triggered.

[0036] Specifically, the data processing module leverages the state predictions from the AI ​​prediction module to optimize charging control parameters: rapid temperature changes trigger pre-emptive tightening of protection thresholds to suppress thermal runaway; fluctuating SOC trends dynamically adjust the constant current / constant voltage switching point to adapt to battery aging characteristics; and excessive voltage fluctuations trigger equalization to ensure battery pack consistency. Through the coupled adjustment of parameters and predicted states, the charging process more closely mirrors battery dynamics, avoiding issues such as excessive temperature, incomplete charging, and intra-pack imbalances, thereby improving battery life and system stability.

[0037] The system also includes a multi-stack collaborative scheduling module, which is configured as follows: Obtain the real-time load rate of each charging station; Based on the priority of the battery to be charged and the real-time load rate, an optimal allocation matrix is ​​calculated and generated. The optimal allocation matrix contains multiple allocation identifiers, each of which is used to indicate whether the corresponding charging pile is allocated charging resources. By calculating load constraints, the sum of the products of the real-time load rate of each charging pile and the corresponding allocation identifier is controlled within the preset maximum load threshold in order to allocate charging resources.

[0038] Specifically, by collecting the load rate of multiple charging piles and combining it with battery priority, an allocation matrix is ​​constructed to achieve optimal matching between charging piles and batteries. Load constraint calculations ensure that the total load of multiple charging piles does not exceed the limit, avoiding the coexistence of local overload and resource idleness. In multi-vehicle concurrent charging scenarios, this significantly improves the utilization efficiency of charging pile clusters, shortens the overall charging time, and solves the resource waste problem caused by the lack of multi-pile coordination.

[0039] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Identical components are represented by the same reference numerals. Therefore, all equivalent changes made to the structure, shape, and principle of this application should be covered within the scope of protection of this application.

Claims

1. A lithium battery charging system, comprising a charging module, a detection module, and a data processing module, characterized in that, The system also includes: The battery health assessment module is used to collect battery cycle count, internal resistance, and capacity decay parameters to build a battery health status assessment model. The power grid interaction module is used to obtain real-time electricity price data and power grid load status; The environmental sensing module includes an environmental temperature and humidity sensor for collecting environmental temperature and humidity parameters; The AI ​​prediction module, built on a deep learning network, is used to predict battery state change trends. The data processing module is connected to the battery health assessment module, the power grid interaction module, the environmental perception module, and the AI ​​prediction module, respectively, and is configured as follows: The upper limit of charging power is dynamically adjusted based on the battery health status assessment results. Charging priority periods are determined based on real-time electricity prices and grid load conditions. The charging strategy is adjusted based on ambient temperature and humidity parameters and battery temperature threshold. Adjust the charging control parameters based on the predicted battery state change trend.

2. The lithium battery charging system according to claim 1, characterized in that, The battery health assessment module performs the following steps: Collect the battery cycle count N, current internal resistance R, and capacity decay rate D; The battery health status is calculated using the formula SOH=α·(C / C0)+β·(R0 / R)+γ·(1-D / N), where C is the current capacity, C0 is the nominal capacity, R0 is the initial internal resistance, and α, β, and γ are weighting coefficients. The SOH value is compared with a preset health threshold to generate a health level assessment result. The health level assessment result includes: when SOH ≥ 90%, the health level is excellent, corresponding to a green label. When 80% ≤ SOH < 90%, the product's health level is good, corresponding to the blue label; When 70% ≤ SOH < 80%, the health level is medium, corresponding to a yellow label. When SOH < 70%, the health level is poor, corresponding to a red label; The health level assessment results are used to indicate the health status of the battery and serve as the basis for the data processing module to adjust the upper limit of charging power.

3. The lithium battery charging system according to claim 1, characterized in that, The power grid interaction module performs the following steps: Real-time acquisition of grid time-of-use electricity price data and regional load status; Based on the electricity price fluctuation curve, periods are divided into low electricity price periods, medium electricity price periods, and high electricity price periods; The upper limit of allocable charging power is determined based on the grid load status.

4. A lithium battery charging system according to claim 1, characterized in that, The environmental perception module performs the following steps: Collect ambient temperature T e and relative humidity H; Combined with battery temperature T β Calculate the temperature difference ΔT = T e -T β ; The environmental impact coefficient K = f(ΔT,H) is generated based on the values ​​of ΔT and H.

5. A lithium battery charging system according to claim 1, characterized in that, The AI ​​prediction module performs the following steps: Constructing an LSTM network model involves inputting a historical charging data sequence. The construction of the LSTM network model includes: Collect historical charging data, including at least voltage, current, temperature, and SOC, and perform normalization preprocessing. The preprocessed data is divided into time windows, with each time window containing m consecutive sampling points, forming the input sequence X = [x1, x2, ..., xm]. m ]; Construct an LSTM network that includes an input layer, a hidden layer, and an output layer. The hidden layer contains p LSTM units, and each LSTM unit contains a forget gate, an input gate, and an output gate structure. The model is trained using the Adam optimizer and mean squared error loss function, and the network weight parameters are adjusted using the backpropagation algorithm. The trained model predicts the battery voltage, temperature, and battery health status over the next n time steps. The model predictions include: Input the current charging data into the trained LSTM model to generate a prediction sequence for the next n time steps; The predicted sequence is denormalized to obtain the predicted values ​​of the actual physical quantities; The charging termination threshold and power slope limit are dynamically adjusted based on the prediction results. This dynamic adjustment includes: When the predicted battery voltage V(t) reaches 95% of the preset upper voltage limit Vmax, the charging termination threshold is adjusted from the initial value V0 to V0·(1+k1·(1-SOH)), where k1 is the voltage adjustment coefficient. When the predicted rate of temperature change dT / dt exceeds the safety threshold Tth, the power slope will be limited from the initial value P. s Adjust to P s ·(1-k2·min(1,dT / dt / Tth)), where k2 is the slope adjustment coefficient; Based on the predicted battery health state SOH(t), the switching point between the constant current stage and the constant voltage stage is dynamically adjusted. When SOH < 80%, the switching point SOC is advanced from 80% to 75%.

6. A lithium battery charging system according to claim 1, characterized in that, The data processing module adjusts the upper limit of charging power based on the battery health status assessment results, including: When SOH ≥ 90%, set the power upper limit P. max =P0; When 70% ≤ SOH < 90%, set the power upper limit P. max =P0·(SOH / 90%); When SOH < 70%, set the power limit P. max =P0·0.7, and start the battery maintenance mode.

7. A lithium battery charging system according to claim 1, characterized in that, The data processing module divides charging priority time periods as follows: During periods of low electricity prices, priority battery packs will be allocated for fast charging. During periods of medium electricity price, maintain regular charging power; During periods of high electricity prices, only the minimum guaranteed power is provided to battery packs in urgent need.

8. A lithium battery charging system according to claim 1, characterized in that, The data processing module corrects the charging strategy by including: When the ambient temperature T e At temperatures above 40℃, the charging power is reduced to P = P0·(1-k1·(T)). e -40)); When the ambient temperature T e At temperatures below 0°C, the battery preheating program is activated, with a preheating power P. r =k2·P0; When the relative humidity H>85%, increase the charging termination voltage threshold αV=k3·H.

9. A lithium battery charging system according to claim 1, characterized in that, The data processing module adjusts the charging control parameters including: The temperature protection threshold is dynamically adjusted based on the battery temperature change rate predicted by LSTM. Based on the predicted SOC change trend, optimize the constant current / constant voltage switching point; When the predicted voltage fluctuation exceeds the threshold, the battery balancing procedure is triggered.

10. A lithium battery charging system according to claim 1, characterized in that, The system also includes a multi-stack collaborative scheduling module, which is configured as follows: Obtain the real-time load rate of each charging station; Based on the priority of the battery to be charged and the real-time load rate, an optimal allocation matrix is ​​calculated and generated. The optimal allocation matrix contains multiple allocation identifiers, each of which is used to indicate whether the corresponding charging pile is allocated charging resources. By calculating load constraints, the sum of the products of the real-time load rate of each charging pile and the corresponding allocation identifier is controlled within the preset maximum load threshold in order to allocate charging resources.

Images