A centralized charging management system for lithium-ion batteries

Abstract

The present application relates to the technical field of battery charging, in particular to a lithium ion battery centralized charging management system, comprising: a state acquisition module, a state management module, a state retrieval module and a charging inspection module; the current charging target battery and the charging nodes of each battery are determined, and the current of each charging node, the voltage and the temperature are taken as the charging state of the target battery; based on the charging state of the target battery, the peak value conditions of the current and the voltage of each charging node are introduced, the transition probability and the peak probability of each charging state under the peak value conditions are analyzed; the target charging state is selected according to the charging state of the target battery and the peak probability, whether the target charging state is consistent at the charging nodes of adjacent two retrievals is compared, if the same, the charging node is recorded; the number of charging node records is compared, whether each charging node is abnormal is judged, and the retrieval of each charging node is completed; the efficiency of centralized battery charging and the accuracy of charging node selection are realized.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of battery charging, in particular to a lithium ion battery centralized charging management system. BACKGROUND

[0002] With the popularity of electric vehicles and energy storage systems, the demand for centralized charging of lithium ion batteries has increased dramatically. Traditional charging technology cannot meet the demand for large-scale, high-efficiency and high-safety charging. Because different batteries have different states, different charging nodes need to be used to adjust the current battery charging schedule to improve the efficiency of each charging node selection in the battery charging schedule solution.

[0003] For example, Chinese Patent Publication No. CN116154914A discloses a battery charging management method and device, wherein the method comprises: obtaining a proactive handshake signal of a battery to be charged before the battery to be charged starts charging; performing self-checking of the battery to be charged according to the proactive handshake signal to determine whether the battery to be charged is abnormal; if the battery to be charged is not abnormal, determining a corresponding charging mode to perform charging operation on the battery to be charged.

[0004] For example, Chinese Patent Publication No. CN115395613A discloses a battery charging management method, device, electronic equipment and storage medium, wherein the method comprises: obtaining historical charging logs of a target battery at at least one battery swap station; determining charging variation data and charging management attributes corresponding to at least one charging phase of the target battery according to the historical charging logs of the at least one battery swap station, wherein the charging management attributes include average charging voltage, average charging current, average battery temperature and average battery voltage corresponding to the corresponding charging phase; determining a charging strategy of a target charging pile for charging the target battery based on the charging variation data corresponding to each charging phase and the corresponding charging management attributes, to charge the target battery based on the charging strategy.

[0005] In the prior art, temperature information is used to match the state of the battery to identify abnormal conditions of the battery state, and recorded data during charging is used to explain the battery charging condition. However, in these battery processing, the different charging nodes required by each battery in the centralized charging scenario of multiple batteries are not considered, resulting in the need to schedule the current of multiple charging nodes, voltage data, etc. to find the most suitable node for the current battery charging to achieve the accuracy of each charging node scheduling in centralized charging. SUMMARY

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a centralized charging management system for lithium-ion batteries, comprising: a status acquisition module, used to determine the target battery currently being charged and the charging nodes of each battery, and using the current, voltage and temperature of each charging node as the charging status of the target battery.

[0007] The state management module is used to determine the state of charge of each charging node under the peak conditions of the current and voltage of each charging node based on the charging state of the target battery; analyze the transition probability of each charging state under the peak conditions; if the transition probability is greater than a preset threshold, obtain the charging cycle of each charging node; use the charging cycle to perform threshold retrieval on each charging point to determine the peak probability of each charging state.

[0008] The state retrieval module is used to control battery charging based on the peak probability of each charging state. It selects the target charging state based on the target battery's charging state and peak probability, and compares whether the target charging state is consistent in two adjacent retrievals of charging nodes. If they are the same, the charging node is recorded.

[0009] The charging check module is used to compare the number of times the charging node records, determine whether each charging node is abnormal, and complete the retrieval of each charging node.

[0010] The beneficial effects of this invention are as follows: First, this invention collects multi-dimensional data such as current, voltage, and temperature in real time, and constructs a global mapping using the connection relationships between nodes. Then, based on the connection relationships between the target battery and multiple charging nodes, it combines the target battery with edge-distributed charging nodes to ensure high-frequency sampling during data processing, providing a foundation for subsequent state management.

[0011] Second, this invention obtains the corresponding peak values, establishes state intervals, and quantifies the occurrence of peak values ​​of current and voltage in different intervals through transition probabilities. This verifies how each charging node performs corresponding state transitions when reaching peak values, enabling timely identification of the dynamics of multiple charging nodes related to the target battery in centralized charging scenarios. This facilitates the selection of appropriate charging nodes to adjust the charging status of the target battery under battery scheduling decisions. Simultaneously, the change trends of the target battery and related charging nodes can be identified based on the transition probabilities, facilitating early warning of peak values ​​exceeding the normal range. This enables scheduling and management of multiple charging nodes, anomaly detection in centralized charging scenarios, and node status query.

[0012] Third, this invention obtains the query records of charging nodes by using the target battery and the node type of each node, as well as the different situations of each charging node during retrieval, and generates the minimum common subset corresponding to each charging node. By identifying the probability of anomalies, the invention classifies the state of charging nodes and makes anomaly decisions, thereby enabling the retrieval of the optimal operating charging node in multiple scenarios such as charging scheduling. Attached Figure Description

[0013] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0014] Figure 1 This is a system framework diagram of a centralized charging management system for lithium-ion batteries.

[0015] Figure 2 This is a flowchart illustrating the state management module of a centralized charging management system for lithium-ion batteries.

[0016] Figure 3 This is a flowchart illustrating the status retrieval module of a centralized charging management system for lithium-ion batteries.

[0017] Figure 4 This is a flowchart illustrating the charging inspection module of a centralized charging management system for lithium-ion batteries. Detailed Implementation

[0018] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0019] See Figure 1 A centralized charging management system for lithium-ion batteries includes: a status acquisition module, a status management module, a status retrieval module, and a charging inspection module; wherein, the output terminal of the status acquisition module is connected to the status management module, the output terminal of the status management module is connected to the status retrieval module, and the output terminal of the status retrieval module is connected to the charging inspection module.

[0020] The status acquisition module is used to determine the target battery and the charging nodes of each battery, and uses the current, voltage and temperature of each charging node as the charging status of the target battery.

[0021] The state management module is used to determine the state of charge of each charging node under the peak conditions of the current and voltage of each charging node based on the charging state of the target battery; analyze the transition probability of each charging state under the peak conditions; if the transition probability is greater than a preset threshold, obtain the charging cycle of each charging node; use the charging cycle to perform threshold retrieval on each charging point to determine the peak probability of each charging state.

[0022] The state retrieval module is used to control battery charging based on the peak probability of each charging state. It selects the target charging state based on the target battery's charging state and peak probability, and compares whether the target charging state is consistent in two adjacent retrievals of charging nodes. If they are the same, the charging node is recorded.

[0023] The charging check module is used to compare the number of times the charging node records, determine whether each charging node is abnormal, and complete the retrieval of each charging node.

[0024] Each battery charging node represents a unit that can connect to the target battery, such as a physical data interface. This charging node can collect the battery's status data to execute battery control commands. Typically, a target battery corresponds to at least one charging node, and the target battery represents the main object controlled by the current charging strategy, including one or more target batteries.

[0025] When the data acquisition module is charging, it describes the charging status of the target battery by the location of each charging node and the collected current, voltage and temperature. It also records when the current and voltage reach their maximum values ​​and the number of cycles that a normal charging cycle is completed, in order to evaluate whether the battery charging status has changed.

[0026] Therefore, the implementation of the status acquisition module includes: calculating the average values ​​of current, voltage and temperature of each charging node based on the location of each charging node, and establishing a mapping relationship between the target battery, charging nodes and charging status based on the connection relationship between each charging node.

[0027] Each charging node is connected sequentially according to its average current, voltage, and temperature to obtain a mapping relationship between at least one charging node and the charging state; the charging nodes and charging states are combined according to the corresponding target battery to obtain a mapping relationship between the target battery, charging node, and charging state.

[0028] The connection relationships between the charging nodes at this point illustrate how the charging nodes for multiple target batteries are connected in a scenario of concentrated battery charging. These connections form the necessary mapping relationships, facilitating the detection of whether excessively high current or voltage peaks at a particular charging node will affect the charging of multiple target batteries and thus pose a charging risk. In addition to the average values, the collected current and voltage values ​​are also actively analyzed for peak values ​​near their upper and lower limits to verify whether current fluctuations are abnormal.

[0029] This section emphasizes evaluating and selecting the most suitable charging node for the target battery based on the real-time operating parameters of the charging node, such as current, voltage, and temperature. This ensures that the selected charging node not only meets the basic charging needs of the target battery but also charges under optimal conditions, thereby guaranteeing charging efficiency and battery health.

[0030] Preferably, when determining the target battery and the charging nodes of each battery, the method further includes: determining the number of target batteries and charging nodes, and sequentially detecting the current, voltage and temperature of the target batteries and charging nodes under different numbers; if the current, voltage and temperature are in an abnormal state, the data of the time period corresponding to the abnormal state is output as the charging state of the target battery.

[0031] In one embodiment of the present invention, the peak condition refers to the instantaneous abnormal state in which a certain parameter, such as voltage, current, or temperature, significantly exceeds its normal fluctuation range within a short period of time during which a current peak or voltage peak occurs at each charging node, reaching or approaching a safety threshold. The time point at which this abnormal state is reached, and the value of a certain parameter, are taken as the peak state identified at this time. Alternatively, a tag can be used to mark this instantaneous abnormal state, and the marked tag can be used as the peak condition.

[0032] For example, the peak conditions obtained are mainly the values ​​of current and voltage. When the moving average of voltage or current in a continuous time window exceeds its standard safety threshold and occurs for more than 3 consecutive sampling points, the data is recorded and marked as a peak condition. At the same time, the corresponding time point and the corresponding value range of the peak condition are recorded. At this time, the safety threshold can be set by adding three times the standard deviation of the historical data to verify the potential peak values ​​that exist, and the abnormal states that approach or exceed the safety threshold can be identified.

[0033] After segmenting the voltage and current within this time period, multiple state intervals are obtained, each representing a different range of voltage and current values. The state intervals are then used to calculate the voltage and the probability of voltage transition within each state interval. For example, after calculating the probability transition using an implicit Maldives model, the charging cycles existing in these voltage and current intervals are selected to identify charging time periods with voltage and current related peak problems.

[0034] like Figure 2 As shown, the transition probability of each charging state under peak conditions can be expressed as follows: perform regional planning for the charging nodes and encode the charging nodes according to a preset order; obtain at least one peak condition of the target battery, and establish the state interval between each charging node based on the numerical interval corresponding to the peak condition.

[0035] Based on the state interval where the peak condition of the charging node occurs, a transition probability model for a single charging node is established. Based on the charging state value of the target battery and the nearest adjacent charging node, a transition probability model for multiple charging nodes is established. Taking the state interval as the main factor, the peak conditions of each charging node under the same charging cycle are compared. If the peak conditions are consistent, the transition probability of the corresponding charging state is calculated using the transition probability model of a single charging node. If the peak conditions are inconsistent, the transition probability of the current adjacent charging cycle is calculated using the transition probability model of multiple charging nodes.

[0036] The single-node transition probability model is designed for a single charging node and directly establishes a transition probability model based on its peak condition state range. This model reflects the dynamic characteristics of the node when it operates independently by quantifying the transition probability of the node in different state ranges; that is, it describes the values ​​of current, voltage and temperature of a single node in different time periods during charging.

[0037] The transition probability model for multiple charging nodes extends to multiple charging nodes by constructing a joint transition probability model that integrates the charging state value of the target battery with the spatiotemporal correlation of adjacent nodes. This model describes the overall behavioral characteristics of the charging network by introducing the coupling relationship between nodes. It is mainly used to indicate whether there are anomalies in the current, voltage, and temperature of multiple sets of adjacent charging nodes at the nearest time, thus demonstrating that each charging node can perform charging allocation normally. That is, it identifies the adjacent charging node closest to the charging stage of the target battery on the charging time axis as the target of analysis, to determine the charging nodes that are closely related in time, space, or function. For example, when the target battery experiences voltage or current peaks, the charging parameters of adjacent nodes are adjusted to prevent system-level overload, thereby achieving the identification and control of the common state of multiple charging nodes.

[0038] Subsequently, the calculations from both models are evaluated using a unified charging cycle to synchronously verify the peak conditions of each charging node. If the peak conditions of all nodes are consistent, the model of a single charging node is used for calculation, and the cooperative transition probability between state intervals is determined based on the state intervals of all charging nodes. If inconsistent peak conditions occur, the model of multiple nodes is used to switch to the analysis of multiple adjacent groups of charging nodes in the near future to achieve cross-cycle state identification of the transition probability of their charging state.

[0039] At this point, the transition probability of a single node is calculated using all state intervals within the current charging cycle, employing Bayes' theorem. For multiple nodes, a Markov chain Monte Carlo method is introduced to estimate the transition probability distribution using sampled data from adjacent cycles, sequentially describing the relative distribution probabilities of the current, voltage, and temperature of the current target battery.

[0040] For the state interval where no peak condition occurs, the transition probability model of a single charging node is used to calculate the transition probability from the state interval where no peak condition occurs to the adjacent state interval, and this probability is used as the output transition probability.

[0041] For state intervals where no peak conditions occur, global state interval analysis is used to explain the probability of changes in the size of the numerical intervals corresponding to current, voltage, and temperature when no peak conditions exist, by using the transition probability from this state interval to adjacent state intervals.

[0042] Preferably, the transition probability model for a single charging node is expressed as: the current represented by the charging state. ,Voltage and temperature The transition probability of a single charging node is set by considering the posterior probability values ​​corresponding to Bayes' theorem. The transition probability of a single charging node represents the variable representing the state interval with peak conditions at a given time, transitioning from one state to another, such as a value indicating a data-driven change from 3.8V to 3.9V. Assume the state variables corresponding to current, voltage, and temperature are... This state variable represents the values ​​of current, voltage, and temperature at time t. Since it involves three variables, the calculation is performed by extracting values ​​from the observed time series of the state variable, assuming it belongs to any one of current, voltage, or temperature. The values ​​of current, voltage, and temperature observed at time t at this point are used to form the... , ;in, , , These represent the values ​​of current, voltage, and temperature at time t, respectively.

[0043] Then we obtain its posterior probability: ;in, Represents the posterior probability. This represents a time series of observations, representing the posterior probability of the corresponding state variable from the first value to the t-th value. Represents the likelihood probability, indicating that in The following observations The probability of; Let represent the prior probability, and let represent the probability without considering the current observation. At that time, state The probability, that is, the numerical value of its probability at t-1; This represents the evidence probability, a constant used for normalization, ensuring that the sum of its posterior probabilities is 1. This involves combining the corresponding likelihood probabilities and prior probabilities according to their respective state variables to obtain their product sum. After completing the calculation of the corresponding posterior probabilities, the relative probabilities of each value during transition can be known based on the multiple posterior probabilities obtained under the current peak condition's corresponding state interval. This allows us to understand whether the current, voltage, and temperature values ​​of the target battery change frequently, and whether there are any sudden increases, thus preventing fault problems at a single node.

[0044] The likelihood probability is obtained by calculating it using a Gaussian distribution based on the observed current, voltage, and temperature, and then representing it as the probability value corresponding to the Gaussian distribution, such as... ;in, This represents the average value of the current. The standard deviation of the current is used to set its likelihood probability when the state variable is current. The same method is used when the state variables are voltage and temperature. Finally, the product of the probabilities of these three factors is used as the output transition probability to represent the change of the corresponding individual charging node under the relevant data in its state interval.

[0045] Subsequently, for the state interval where no peak condition occurs, the transition probability model of a single charging node is adopted, and the adjacent state intervals are used as the main calculation intervals to obtain their calculated transition probabilities.

[0046] Preferably, the transition probability model for multiple charging nodes is represented by taking any two charging nodes as a group and obtaining multiple sampling points of their nearest moments in adjacent charging cycles to form their state vector. and state vector , ; These two state vectors indicate that the two selected charging nodes at this time satisfy the charging state value of the target battery and the nearest adjacent charging node. There are N elements in the state vector.

[0047] Then, its state distribution vector is calculated as the transition probability of the output.

[0048] ;in, The state distribution vector represents the probability of being in state vector x at time t+1, indicating that the system is in a certain state. The state distribution vector represents the probability of being in state vector y at time t. Being in state vector y at this time indicates the joint distribution probability of multiple collected values, which represents the correlation between multiple charging nodes in the near time. The summation of y represents the calculation method of summing and accumulating the elements existing in state vector y at this nearest time. This represents the probability of transitioning from state vector y to state vector x. This probability is the conditional probability, which is the conditional probability of choosing two values ​​at the same time at two charging nodes. This data can be calculated using historical data. At this time, the values ​​of the state distribution vectors calculated separately for current, voltage, and temperature are multiplied together to obtain the output transition probability.

[0049] The method for determining the peak probability under each charging state also includes: obtaining the maximum charging amount of the target battery in each state interval, and extracting the peak condition under the corresponding charging cycle based on the charging cycle required for the maximum charging amount.

[0050] The system manages each state interval in batches based on the state interval corresponding to the peak condition. When the same peak condition is met in each state interval, the peak probability for each charging state is recorded, and this peak probability is output. The peak probability represents the ratio of the amount of data corresponding to the peak condition to the amount of data contained in the current state interval, indicating the probability of that peak condition occurring.

[0051] Meanwhile, for the preset threshold where the transition probability is greater than a certain threshold, this preset threshold is set based on historical data, using 95% of the transition probability values ​​in the historical data as the preset threshold used here. It should be noted that the preset threshold for the transition probability will be adjusted according to the different models selected for the transition probability, to describe the data extraction for multiple models.

[0052] In one embodiment of the present invention, the implementation method of selecting the target charging state based on the charging state and peak probability of the target battery in the state retrieval module includes: calculating the peak probability of each target battery in each charging state, recording the probability of an abnormal event occurring in the target battery; taking the target battery that has an abnormal event as the starting point, selecting the charging state of the adjacent charging node of the target battery as the target charging state.

[0053] like Figure 3 As shown, the implementation method of comparing the target charging status in two adjacent searches in the status retrieval module includes: taking the charging node as input data, comparing whether the node type of each charging node is a candidate charging type, and retrieving the charging node in the first search for the target charging status.

[0054] The implementation method for comparing whether the node type of each charging node is a candidate charging type also includes: if it is a candidate charging type, the charging status of the corresponding charging node is extracted to form a state interval for each charging node. The similarity between each state interval is calculated based on the maximum peak probability in each state interval, and the charging node corresponding to the state interval with the highest similarity value is taken as the charging node for the first search.

[0055] If it is not a candidate charging type, the charging occupation cost of the corresponding charging node is obtained. The charging occupation cost is the proportion of the time that the current flows through the charging node when it is charging in the network corresponding to the overall charging nodes. Using the charging occupation cost as the search condition, the charging node corresponding to the target battery when it is charging is searched, and the corresponding charging node is used as the first searched charging node.

[0056] Based on the guiding charging sequence of each charging node, each charging node is searched to obtain the charging node with the target charging status in the second search.

[0057] Guided charging sequence refers to the process of allocating charging priority or queuing order for devices to be charged among multiple charging nodes. This process includes, but is not limited to, the device connection start time, the remaining charge of the target battery, the device type corresponding to the target battery, and the charging power of the target battery. These factors will determine the priority of the target battery in centralized charging management and affect the charging time of the target battery. The second search for charging nodes mainly measures whether the target node in the priority calculation is consistent with the target node searched in the first search. At this time, the power resources under centralized management are allocated by comparing the state range of the charging nodes or the charging occupation cost to prevent some charging nodes from being overused and some charging nodes from being restricted.

[0058] The second search involves selecting the relative priority of the charging nodes. The status of the charging nodes during the first search is compared with their priority to determine whether the charging nodes used by the current target battery can achieve the optimal result.

[0059] The first search primarily identifies any abnormal states, while the second search verifies node consistency to ensure that abnormal nodes are not selected repeatedly. This prevents issues such as asymmetry in actual detection data across nodes under global analysis and the problem of concentrated node selection when charging node data is relatively independent and not centrally integrated.

[0060] The second method of retrieving charging nodes includes using the device's connection start time, the remaining power of the target battery, the device type corresponding to the target battery, and the charging power of the target battery as the charging tag of the charging node.

[0061] The charging nodes are retrieved by charging tags, and the charging event stream of each charging node is obtained. The charging event stream indicates whether the current charging node is charging normally and the real-time charging amount when the charging node is judged based on the device connection start time, the remaining power of the target battery, the device type corresponding to the target battery, and the charging power of the target battery. The completion time of these data is combined with whether there are any abnormal events in the charging process to form the charging event stream at this time.

[0062] At least one management branch is generated using the charging event stream corresponding to the target charging state. The charging nodes within this management branch are used as the charging nodes for the second retrieval. Each management branch represents the ratio of the expected charging completion time to the normal completion time for multiple target batteries before and after charging is completed, within the charging event stream. The charging data for multiple groups of target batteries corresponding to this ratio are used as the current management branch to describe the charging nodes that complete charging at different priorities. These multiple groups of charging nodes that complete charging at different priorities are then compared with the charging nodes retrieved in the first retrieval to determine whether the current charging node can achieve optimal performance.

[0063] The charging nodes retrieved in the first search are compared with those retrieved in the second search. If any of the charging nodes retrieved in the first search are present in the second search, the corresponding charging nodes are extracted and recorded according to the data contained in the two searches.

[0064] If it does not exist, then based on the distribution of each charging node during the first search, starting from the charging node in the first search, connect the charging node in the first search with the charging node in the second search to obtain the shortest path between each charging node in the first search and the second search, and record the charging node corresponding to the shortest path.

[0065] In one embodiment of the present invention, during the charging check, the data marked by each charging node in each module is mainly used to analyze whether the current charging node can meet the charging requirements of the current target battery, and whether the data corresponding to the charging node can meet the fast charging of the target battery in a centralized charging scenario.

[0066] like Figure 4 As shown, the implementation of the charging inspection module includes: based on the number of times each charging node records data, extracting the least common subset of each charging node under multiple inspections, and constructing a deep query network using the elements corresponding to the least common subset of each charging node.

[0067] During each record retrieval, the current charging node is analyzed using the status retrieval module. Each retrieval checks the charging node's consistency based on data such as peak probability, alternative charging types, charging cost, the ratio of estimated charging completion time to normal completion time, and the connection time of related devices. These data are used to combine charging nodes to verify the minimum common subset corresponding to the retrieved content. This common subset includes the retrieved content. If a charging node is recorded frequently and its current, voltage, and temperature are stable, then the multiple charging nodes corresponding to the minimum common subset represent nodes that can be relied upon for current centralized charging. If a charging node is recorded infrequently, it is more likely to be a charging node designed for specific needs and can serve as an alternative node for different devices. Furthermore, if the minimum common subset exhibits unstable current, voltage, and temperature, the number of records can be used to identify problematic charging nodes and those exhibiting instability under specific conditions, enabling dynamic decision-making under various charging strategies.

[0068] For the deep query network, the least common subset of each charging node is used as the embedding vector. The embedding vectors are combined sequentially to form the input layer, hidden layer, and output layer. The output layer outputs the anomaly probability using the Sigmoid function. The convolutional kernel of the input layer is set to 3×3. The fully connected layers in the hidden layer are set sequentially from 256 nodes to 128 nodes to 64 nodes. After initializing and eliminating the dimensions of various data contained in the least common subset, the anomaly probability corresponding to the data is verified. Then, this anomaly probability is correlated with the number of records to find the more stable charging nodes and complete the retrieval of each charging node during the charging decision.

[0069] The deep query network is used to compare the anomaly probability of charging nodes under different number of records, and the anomaly probability of each charging node is output.

[0070] Preferably, when outputting the charging nodes with anomaly probabilities, the method further includes receiving the minimum common subset of each charging node, and classifying each charging node into stable extreme value nodes and unstable extreme value nodes according to the number of records and the anomaly probability, and in turn using the maximum and minimum values ​​of the corresponding data in the minimum common subset within the corresponding charging cycle. Stable extreme value nodes represent extreme value nodes with a high number of records and a low anomaly probability, which may represent extreme situations in the normal charging process. These nodes will represent the charging status of different charging nodes after the scheduling adjustment. These charging nodes will represent the part that each charging node can rely on after the scheduling. Unstable extreme value nodes represent extreme value nodes with a low number of records but a high anomaly probability, which may represent potential faults or abnormal charging behavior. These charging nodes represent the nodes that are mainly targeted when searching for charging nodes. At the same time, when obtaining stable extreme value nodes, it is also necessary to verify their maximum values ​​multiple times to prevent problems such as current overload and voltage overload after scheduling.

[0071] The values ​​of stable and unstable extreme nodes are monitored in real time. If there is an anomaly in the charging node corresponding to the stable or unstable extreme node, the corresponding abnormal node will be output; otherwise, the stable and unstable extreme nodes will be output.

[0072] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A centralized charging management system for lithium-ion batteries, characterized in that, include: The status acquisition module is used to determine the target battery and the charging nodes of each battery, and uses the current, voltage and temperature of each charging node as the charging status of the target battery. The state management module is used to determine the state of charge of each charging node under the peak conditions of the current and voltage of each charging node based on the charging state of the target battery; analyze the transition probability of each charging state under the peak conditions; if the transition probability is greater than a preset threshold, obtain the charging cycle of each charging node; use the charging cycle to perform threshold retrieval on each charging node to determine the peak probability of each charging state. The transition probability of each charging state under peak conditions is expressed as follows: regional planning of charging nodes and encoding of charging nodes according to a preset order; At least one peak condition of the target battery is obtained, and the state interval between each charging node is established based on the numerical interval corresponding to the peak condition. Based on the state interval where the charging node reaches its peak condition, a transition probability model for a single charging node is established. Based on the charging state value of the target battery and the nearest adjacent charging node, a transition probability model for multiple charging nodes is established. Based on the state interval, compare whether the peak conditions of each charging node are consistent in the same charging cycle. If the peak conditions are consistent, use the transition probability model of a single charging node to calculate the transition probability of the corresponding charging state. If the peak conditions are inconsistent, the transition probability of the current adjacent charging cycle is calculated using a transition probability model of multiple charging nodes. For the state interval where no peak condition occurs, the transition probability model of a single charging node is used to calculate the transition probability from the state interval where no peak condition occurs to the adjacent state interval, and this probability is used as the output transition probability. The state retrieval module is used to control battery charging based on the peak probability of each charging state. It selects the target charging state based on the target battery's charging state and peak probability, and compares whether the target charging state is consistent in two adjacent retrievals of the charging node. If they are the same, the charging node is recorded. The charging check module is used to compare the number of times the charging node records, determine whether each charging node is abnormal, and complete the retrieval of each charging node.

2. The centralized charging management system for lithium-ion batteries according to claim 1, characterized in that, The implementation methods of the status acquisition module include: Based on the location of each charging node, the average values ​​of current, voltage, and temperature of each charging node are calculated, and the connection relationship between each charging node is analyzed. Each charging node is connected sequentially according to its average current, voltage, and temperature to obtain a mapping relationship between at least one charging node and the charging state; the charging nodes and charging states are combined according to the corresponding target battery to obtain a mapping relationship between the target battery, charging node, and charging state.

3. The centralized charging management system for lithium-ion batteries according to claim 1, characterized in that, Determining the target battery for the current charge and the charging nodes of each battery also includes: Determine the number of target batteries and charging nodes, and sequentially detect the current, voltage, and temperature of target batteries and charging nodes with different numbers; if the current, voltage, and temperature are in an abnormal state, the data for the time period corresponding to the abnormal state will be output as the charging state of the target battery.

4. The centralized charging management system for lithium-ion batteries according to claim 1, characterized in that, Other methods for determining the peak probability under each charging state include: Obtain the maximum charge amount of the target battery in each state interval, and extract the peak condition under the corresponding charging cycle based on the charging cycle required for the maximum charge amount. The state intervals corresponding to the peak conditions are managed in batches. When the same peak condition is met in each state interval, the peak probability of each charging state is recorded and then output.

5. A centralized charging management system for lithium-ion batteries according to claim 1, characterized in that, The implementation methods for comparing the target charging state in two adjacent retrievals of the charging node in the state retrieval module include: Using charging nodes as input data, compare whether the node type of each charging node is a candidate charging type, and retrieve the charging node with the target charging status in the first retrieval. Based on the guiding charging sequence of each charging node, each charging node is searched to obtain the charging node with the target charging status in the second search. The charging nodes retrieved in the first search are compared with those retrieved in the second search. The comparison is made to see if any of the charging nodes included in the first search exist in the second search. If they do, the corresponding charging nodes are extracted and recorded according to the data included in the two searches. If it does not exist, then based on the distribution of each charging node during the first search, starting from the charging node in the first search, connect the charging node in the first search with the charging node in the second search to obtain the shortest path between each charging node in the first search and the second search, and record the charging node corresponding to the shortest path.

6. A centralized charging management system for lithium-ion batteries according to claim 5, characterized in that, The implementation methods for comparing whether the node type of each charging node is a candidate charging type also include: If it is a candidate charging type, the charging status of the corresponding charging node is extracted to form a state interval for each charging node. The similarity between each state interval is calculated based on the maximum peak probability in each state interval. The charging node corresponding to the state interval with the highest similarity value is taken as the charging node for the first search. If it is not a candidate charging type, the charging cost of the corresponding charging node is obtained. Using the charging cost as the search condition, the charging node corresponding to the target battery when it is charging is searched, and the corresponding charging node is used as the first search charging node.

7. A centralized charging management system for lithium-ion batteries according to claim 5, characterized in that, The second method for retrieving charging nodes includes: The charging node is tagged with the device's connection start time, the target battery's remaining power, the device type corresponding to the target battery, and the target battery's charging power. The charging nodes are retrieved by charging tags to obtain the charging event stream of each charging node; At least one management branch is generated using the charging event stream corresponding to the target charging state, and the charging nodes within the management branch are used as the charging nodes for the second retrieval.

8. A centralized charging management system for lithium-ion batteries according to claim 1, characterized in that, The charging check module can be implemented in the following ways: Based on the number of times each charging node is recorded, the least common subset of each charging node under multiple checks is extracted, and a deep query network is constructed using the elements corresponding to the least common subset of each charging node. The deep query network is used to compare the anomaly probability of charging nodes under different number of records, and the anomaly probability of each charging node is output.

9. A centralized charging management system for lithium-ion batteries according to claim 8, characterized in that, The method of outputting each charging node with anomaly probability also includes receiving the least common subset of each charging node, and classifying each charging node into stable extreme value nodes and unstable extreme value nodes according to the number of records and the anomaly probability, and in turn using the maximum and minimum values ​​of the corresponding data in the least common subset within the corresponding charging cycle. The values ​​of stable and unstable extreme nodes are monitored in real time. If there is an anomaly in the charging node corresponding to the stable or unstable extreme node, the corresponding abnormal node will be output; otherwise, the stable and unstable extreme nodes will be output.

Images