Intelligent operation and maintenance management platform and method for battery pack of electric carry-scraper
By using an intelligent operation and maintenance management platform, employing fuzzy matrix and membership calculation, adaptive clustering, and reinforcement learning, the problem of traditional battery management systems being unable to accurately predict battery life and health status in electric shovel loader has been solved, achieving efficient battery management and stable equipment operation.
Patent Information
- Application Number
- CN202511890812.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-15
- Publication Date
- 2026-02-06
AI Technical Summary
Traditional battery management systems cannot accurately predict battery life and health status, struggle to cope with battery damage and temperature accumulation caused by deep discharge, and are unable to achieve active control of temperature diffusion, leading to frequent malfunctions and reduced efficiency in electric shovel loader equipment.
The system employs a data acquisition and processing module, an anomaly detection module, and a strategy optimization module. It identifies abnormal data through fuzzy matrix and membership degree calculation, performs adaptive clustering and fault prediction, constructs a state reward and punishment table for reinforcement learning, and optimizes the charging and discharging strategy.
It enables flexible and accurate detection and grouping of battery status, and dynamic adjustment of charging and discharging behavior, thereby improving the efficiency of battery pack management and the overall working efficiency and economic benefits of electric loaders.
Smart Images
Figure CN121469313A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of battery management technology for electric loaders, and more specifically, to an intelligent operation and maintenance management platform and method for electric loader battery packs. Background Technology
[0002] With the widespread adoption of electrification in underground operations such as mining and tunneling, electric loaders are gradually replacing traditional fuel-powered equipment due to their environmental friendliness and low noise. However, the battery packs of electric loaders face unique and complex maintenance challenges in practical applications. These problems, due to their variability and unpredictability, are often difficult to be effectively solved by existing battery management systems.
[0003] Traditional battery management systems (BMS) can only record and display real-time battery data, but cannot predict battery life and health status. Under high-load operation, battery degradation accelerates, and traditional management systems lack predictive maintenance capabilities, making it difficult to accurately determine the remaining battery life or degradation status, leading to frequent equipment malfunctions and unplanned downtime. Electric loader battery packs frequently experience deep discharge during operation, which causes significant damage, especially to the internal active materials. Although equalization charging can partially restore battery performance, a single charging strategy is insufficient to address the performance degradation caused by deep discharge. Uneven recovery often results in a "pseudo-fully charged" phenomenon in individual cells within the battery pack after charging, meaning the battery surface voltage meets the requirements but the actual energy storage is insufficient, leading to a decrease in the overall efficiency and capacity of the battery pack. In actual use, the battery pack will exhibit uneven temperature distribution, especially under high power output conditions, where some battery cells will overheat, causing an increase in the temperature gradient. Traditional battery management systems only use temperature thresholds for alarms or passive cooling, and cannot achieve active control of temperature difference diffusion. The continuous accumulation of temperature difference can cause some batteries to overheat, affecting battery consistency, and in extreme cases, may lead to thermal runaway or fire.
[0004] In view of this, the present invention proposes an intelligent operation and maintenance management platform and method for electric shovel loader battery packs to solve the above problems. Summary of the Invention
[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: an intelligent operation and maintenance management platform for electric shovel loader battery packs, comprising: Data acquisition and processing module: Collects battery status data from the battery pack, preprocesses the battery status data, and obtains valid battery status data; Anomaly detection module: Adaptively clusters valid battery data to obtain N similar battery groups, and predicts the similar battery groups based on the constructed battery fault prediction model to obtain faulty battery groups and normal battery groups; Strategy optimization module: Constructs a state reward and punishment table based on the faulty battery pack, performs reinforcement learning on the normal battery pack based on the state reward and punishment table to obtain a charge and discharge optimization strategy, and optimizes the charge and discharge of the battery based on the charge and discharge optimization strategy to improve work efficiency.
[0006] Furthermore, the battery status data includes: battery temperature, voltage, current, impact intensity, and vibration frequency.
[0007] Furthermore, the method for preprocessing the battery state data includes: The standard score algorithm is used to normalize the battery state data to obtain a standard battery vector, and a system is built based on the standard battery vector. The fuzzy battery matrix, in which The data volume representing the standard battery vectors is used to perform fuzzy weighted calculations on different standard battery vectors to obtain fuzziness. This fuzziness is then mapped to a fuzzy battery matrix. The mean of each element in the fuzzy battery matrix is calculated to obtain the average fuzziness of each standard battery vector. The average fuzziness values are then sorted from largest to smallest, and the top values are selected. The standard battery vectors corresponding to the average fuzziness are used as fuzzy centers. Average membership is calculated for each standard battery vector based on these fuzzy centers to obtain the average membership degree. Anomaly membership thresholds are then defined based on the average membership degree. ;in, Represents the abnormal membership threshold. The mean of the average membership degree. This represents the threshold adjustment parameter. The variance of the average membership degree is used to mark the battery state data corresponding to the average membership degree that is greater than or equal to the abnormal membership threshold as abnormal. All data marked as abnormal in the battery state data are removed from the battery state data to obtain the valid battery state data.
[0008] Furthermore, the formula for calculating the average membership is: ;in, Representing the The average membership degree of a standard battery vector. The number of fuzzy centers Representing the The standard battery vector and the first Similarity of fuzzy centers Representing the The fuzzy center and the first Similarity of fuzzy centers This represents the ambiguity parameter.
[0009] Furthermore, the adaptive clustering method for the valid battery data includes: Vectorize the effective battery state data to obtain effective battery vectors, and then randomly divide the effective battery vectors into... The process involves initializing the battery centers of each initial battery group, calculating the mean of each feature of the effective battery vectors in each initial battery group, and then using the mean of all features in each initial battery group to form the initial battery center. The battery weights of each initial battery group are initialized as the ratio of the number of effective battery vectors to the total number of effective battery vectors in that group. Based on the initial battery centers, the aggregation probability of each effective battery vector is calculated to obtain the aggregation probability. Based on the aggregation probability, the effective battery vectors are aggregated to the initial battery center with the highest aggregation probability to obtain new initial battery groups. The mean of each feature of the effective battery vectors in each new initial battery group is calculated to obtain new feature mean. All new feature mean values of each new initial battery group constitute a new initial battery center. Finally, the battery weights of each new initial battery group are updated using the following formula: ;in, The updated weight value representing the battery weight. The total amount of data representing the effective battery vector. The index represents the effective battery vector; the battery weights are updated with the updated weight value as the new value of the battery weights, and the new battery weights are obtained; this is repeated until the value of the battery weight of each initial battery group no longer changes, and the initial battery group at this time is the similar battery group.
[0010] Furthermore, the formula for performing the aggregation calculation is as follows: ;in, Representing the One effective battery vector, Representing the The effective battery vector is directed towards the first The aggregation probability of the initial battery center aggregation of the initial battery pack. The data dimension representing the effective battery vector, Representing the The covariance matrix of the initial battery pack Representing the The initial battery weights of the battery pack Representing the The inverse matrix of the covariance matrix of the initial battery pack Representing the The initial battery center of the initial battery pack. This represents the matrix transpose operation.
[0011] Furthermore, the battery fault prediction model is constructed in the following ways: collection The system uses historical battery fault data and historical battery state data. The historical battery fault data represents the fault types that occurred in the past, while the historical battery state data has the same data type and format. These two sets of data are used as the training sample set. A GRU model is used as the initial model, and the GRU model is trained using the training sample set. The historical battery state data is used as the input data for the battery fault prediction model, and the predicted battery fault categories are used as the output data. The training objective is to minimize the error between the actual historical battery fault data and the predicted battery fault categories. The recall function is used as the loss function for the battery fault prediction model. Training stops when the loss function converges, resulting in the battery fault prediction model.
[0012] Furthermore, the method for constructing the state reward table based on the faulty battery pack includes: Based on the trained battery fault prediction model, the initial battery center of each group of similar battery packs is used as the input of the battery fault prediction model for detection. Similar battery packs whose battery fault categories are predicted are identified as faulty battery packs and fault alarms are issued. Similar battery packs whose battery fault categories are not predicted are identified as normal battery packs. Define battery state parameters, including battery temperature, voltage, and current parameters. Battery temperature parameters include: low temperature, medium temperature, normal temperature, and high temperature; voltage parameters include: low current, medium current, normal current, and high current; current parameters include: low voltage, medium voltage, normal voltage, and high voltage. Use a permutation and combination algorithm to arrange and combine the battery temperature, voltage, and current parameters to obtain battery states. All battery states constitute the battery state space. Define battery actions in the battery action space, including: increasing charging current, decreasing charging current, increasing discharging current, and decreasing discharging current. Battery groups with the same fault category are grouped together. The number of fault data points for each battery fault category is counted. Based on the number of fault data points, the battery fault categories are sorted in ascending order. Then, based on the sorting order, initial weights are applied. The fault weight is assigned to each battery fault category in ascending order, with each increment being [value missing]. The system calculates the fault intervals of each dimension in the data group of the same battery fault category. Based on the comparison between the fault interval and the battery status parameters, it assigns penalty scores to battery status parameters whose fault intervals match the battery status parameters. For the remaining battery status parameters that are not assigned penalty scores, the maximum value of the fault weight is used as the reward score for the normal category in each battery status parameter category. The reward scores are then assigned to other parameters in descending order of average value to obtain the status reward and penalty table.
[0013] Furthermore, the reinforcement learning method for normal battery packs includes: Initialize the action learning table and the exploration probability. Based on the battery state space, each data in the normal battery pack is assigned to the corresponding battery state, and an initial state score is assigned to each data in the normal battery pack based on the state reward table. Starting with the battery state of each data point in a normal battery pack, a random selection number is generated using a random algorithm. When the random selection number is greater than or equal to the exploration probability, a battery action is selected from the battery action space using the random selection algorithm to update the battery state. The formula for state update is as follows. ;in Represents the final state score. Represents a new battery state. This represents a battery action that updates its status. Represents the initial state score. Represents the initial battery state. This represents the initial battery action. Represents the selection of battery action After the battery state becomes The subsequent choice of state score, Represents the discount factor. This represents the battery action space. When the updated state score is greater than the initial state score, the action update table records the state update process, which includes state changes and changes in the initial state score. When the number of random selections is less than the exploration probability, the state update process in the action learning table is matched with the battery state. The state update process with the same battery state as the starting point and the largest change in the initial state score is found to update the battery state. This process is repeated until the battery state of each data point in the normal battery pack no longer changes. The action learning table at this point is the charge and discharge optimization strategy.
[0014] This invention provides an intelligent operation and maintenance management method for battery packs of electric shovel loader, comprising: S1. Collect battery status data of the battery pack, preprocess the battery status data, and obtain valid battery status data. S2. Perform adaptive clustering on the effective battery data to obtain N similar battery groups. Based on the constructed battery fault prediction model, predict the similar battery groups to obtain faulty battery groups and normal battery groups. S3. Construct a state reward and punishment table based on the faulty battery pack, perform reinforcement learning on the normal battery pack based on the state reward and punishment table to obtain a charge and discharge optimization strategy, and optimize the charge and discharge of the battery based on the charge and discharge optimization strategy to improve work efficiency.
[0015] The technical effects and advantages of the intelligent operation and maintenance management platform and method for electric shovel loader battery packs of the present invention are as follows: This invention identifies abnormal battery state data through fuzzy matrix and membership degree calculation, making anomaly detection more flexible and accurate. By removing abnormal data, the system obtains more reliable battery state data, providing accurate input for subsequent analysis and improving data effectiveness. Adaptive clustering algorithms group valid battery data, allowing the system to adaptively adjust the grouping method according to changes in battery state, avoiding the limitations of fixed clustering algorithms on battery pack diversity. By dynamically adjusting the initial battery centers and battery weights, the system obtains a more accurate distribution of similar battery packs, effectively improving grouping accuracy. Reinforcement learning dynamically adjusts the charging and discharging behavior of normal battery packs; using exploration probabilities and action learning tables, the system finds the optimal charging and discharging scheme after multiple updates. Reinforcement learning makes the charging and discharging optimization strategy more adaptable and flexible. Through continuous learning and adjustment, the system can manage battery packs more efficiently, avoiding battery damage caused by frequent charging and discharging, further improving the overall working efficiency and economic benefits of the electric loader. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the process of an intelligent operation and maintenance management platform for electric shovel loader battery packs according to the present invention; Figure 2 This is a schematic diagram of an intelligent operation and maintenance management method for battery packs of electric loaders according to the present invention. Detailed Implementation
[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] Example 1 Please see Figure 1 As shown in this embodiment, an intelligent operation and maintenance management platform for electric shovel loader battery packs includes: a data acquisition and processing module: acquiring battery status data of the battery pack, preprocessing the battery status data, and obtaining valid battery status data; Anomaly detection module: Adaptively clusters valid battery data to obtain N similar battery groups, and predicts the similar battery groups based on the constructed battery fault prediction model to obtain faulty battery groups and normal battery groups; Strategy optimization module: Constructs a state reward table based on the faulty battery pack, performs reinforcement learning on the normal battery pack based on the state reward table to obtain a charge and discharge optimization strategy, and optimizes the charge and discharge of the battery based on the charge and discharge optimization strategy to improve work efficiency; Battery status data includes: battery temperature, voltage, current, impact intensity, and vibration frequency. Vibration frequency refers to the vibration caused by the electric mining conveyor during operation. Impact intensity refers to the impact force exerted on the battery pack during operation. Excessive impact force may damage the internal structure or connectors of the battery, and in severe cases, may even lead to battery pack failure. Battery temperature refers to the temperature of the battery in working or resting states. Battery temperature has a significant impact on its performance and lifespan. Excessive temperature may lead to decreased battery performance, shortened lifespan, and even safety hazards.
[0019] Methods for preprocessing battery status data include: The standard score algorithm is used to normalize the battery state data to obtain a standard battery vector, and a system is built based on the standard battery vector. Fuzzy battery matrix ,in The data volume representing the standard battery vector, each element in the fuzzy battery matrix is determined through fuzzy weighted calculation. The formula for fuzzy weighted calculation is as follows: ;in Representing the The standard battery vector and the first The ambiguity between the standard battery vectors, i.e., the 1st standard battery vector in the ambiguity battery matrix. Line number The values of the elements in the column. This represents the total number of data dimensions in a standard battery vector. Representing the In the nth standard battery vector Data in each dimension No. In the nth standard battery vector Data in each dimension This represents a smoothing parameter used to control the smoothness of data similarity; the mean of each row element in the fuzzy battery matrix is calculated to obtain the average fuzziness of each standard battery vector, and the values are sorted from largest to smallest based on the average fuzziness, selecting the top... The standard battery vector corresponding to each average fuzziness is used as the fuzziness center. An average membership calculation is performed on each standard battery vector based on the fuzziness center. The formula for the average membership calculation is: ;in, Representing the The average membership degree of a standard battery vector. The number of fuzzy centers Representing the The standard battery vector and the first Similarity of fuzzy centers Representing the The fuzzy center and the first The similarity between fuzzy centers can be calculated using common formulas such as the cosine similarity formula and the Euclidean distance formula. The ambiguity parameter controls the degree of ambiguity in the membership assignment of the standard battery vector to different ambiguity centers; an anomaly membership threshold is defined based on the average membership degree. ;in, Represents the abnormal membership threshold. The mean of the average membership degree. This represents the threshold adjustment parameter. The variance representing the average membership degree is used to mark battery state data with an average membership degree greater than or equal to the abnormal membership threshold as abnormal. All data marked as abnormal are removed from the battery state data to obtain valid battery state data. Abnormal battery state data is identified by using a fuzzy matrix and membership degree calculation, making anomaly detection more flexible and accurate. By removing abnormal data, the system can obtain more reliable battery state data, providing accurate input for subsequent analysis and improving data effectiveness. In this embodiment, the threshold adjustment parameter is preferably set to 0.05, and the fuzziness parameter is preferably set to 2.
[0020] Adaptive clustering methods for valid battery data include: Step 1: Vectorize the effective battery state data to obtain the effective battery vector, and randomly divide the effective battery vector into... Each initial battery group is initialized with an initial battery center. The battery weight of each initial battery group is the ratio of the effective battery vector data to the total effective battery vector data in each initial battery group. The initial battery center is initialized by calculating the mean of each feature of the effective battery vector in each initial battery group, and the mean of all features of each initial battery group constitutes the initial battery center. Step 2: Calculate the aggregation probability for each effective battery vector based on the initial battery center. The formula for aggregation calculation is: ;in, Representing the One effective battery vector, Representing the The effective battery vector is directed towards the first The aggregation probability of the initial battery center aggregation of the initial battery pack. The data dimension representing the effective battery vector, Representing the The covariance matrix of the initial battery pack Representing the The initial battery weights of the battery pack Representing the The inverse matrix of the covariance matrix of the initial battery pack Representing the The initial battery center of the initial battery pack. Represents the matrix transpose operation, which is used to convert a column vector into a row vector, such as a vector... After performing a matrix transpose operation, it will become ; Step 3: Based on the aggregation probability, aggregate the effective battery vectors towards the initial battery center with the highest aggregation probability to obtain new initial battery groups. Calculate the mean of each feature of the effective battery vectors in each new initial battery group to obtain new feature mean values. All new feature mean values of each new initial battery group constitute a new initial battery center. Update the battery weights of each new initial battery group using the following formula: ;in, The updated weight value representing the battery weight. The total amount of data representing the effective battery vector. The index represents the effective battery vector; the battery weight is updated with the updated weight value as the new value of the battery weight, and the new battery weight is obtained; Step 4: Repeat steps 2 and 3 until the battery weights of each initial battery pack no longer change. Output the initial battery packs at this point, which are the similar battery packs.
[0021] The methods for constructing battery failure prediction models include: collection The system uses historical battery fault data and historical battery state data. The historical battery fault data represents the fault types that occurred in the past, while the historical battery state data has the same data type and format. These two sets of data are used as the training sample set. A GRU model is used as the initial model, and the GRU model is trained using the training sample set. The historical battery state data is used as the input data for the battery fault prediction model, and the predicted battery fault categories are used as the output data. The training objective is to minimize the error between the actual historical battery fault data and the predicted battery fault categories. The recall function is used as the loss function for the battery fault prediction model. Training stops when the loss function converges, resulting in the battery fault prediction model. Based on the trained battery fault prediction model, the initial battery center of each group of similar battery packs is used as the input of the battery fault prediction model for detection. Similar battery packs whose battery fault categories are predicted are identified as faulty battery packs and a fault alarm is issued. Similar battery packs whose battery fault categories are not predicted are identified as normal battery packs.
[0022] The methods for constructing a status reward and punishment table include: Battery state parameters are defined, including battery temperature, voltage, and current parameters. Battery temperature parameters include: low temperature, medium temperature, normal temperature, and high temperature; voltage parameters include: low current, medium current, normal current, and high current; current parameters include: low voltage, medium voltage, normal voltage, and high voltage. A permutation and combination algorithm is used to arrange and combine the battery temperature, voltage, and current parameters to obtain battery states. All battery states constitute the battery state space. For example, low temperature, low voltage, and low current constitute a battery state, representing a low-power and low-load state. This battery state space... The system can adopt different charging and discharging methods under different states to achieve optimization. The definitions of various parameters in the battery state parameters are obtained through testing and verification by test engineers under different load conditions. For example, in actual testing, it was found that the battery temperature rises sharply after exceeding 40 degrees Celsius under normal voltage and normal current. The high temperature is defined as greater than or equal to 40 degrees Celsius. Each category in the battery state parameters represents a range. The battery actions in the battery action space are defined. Battery actions include: increasing charging current, decreasing charging current, increasing discharging current, and decreasing discharging current. Battery groups with the same fault category are grouped together. The number of fault data points for each battery fault category is counted. Based on the number of fault data points, the battery fault categories are sorted in ascending order. Then, based on the sorting order, initial weights are applied. The fault weight is assigned to each battery fault category in ascending order, with each increment being [value missing]. The system analyzes the fault intervals of each dimension in the data set of the same battery fault category. Based on a comparison between the fault intervals and battery state parameters, penalty scores are assigned to battery state parameters whose fault intervals match the battery state parameters. For the remaining battery state parameters without assigned penalty scores, the maximum fault weight is used as the reward score for the normal state within each battery state parameter category. This reward score is then distributed to other parameters in descending order of average value, resulting in a state reward / penalty table. The average descending order is calculated by taking the normal state within each battery state parameter category as the center and calculating the decrease value based on the interval size. For example, assuming the intervals without assigned reward scores are: normal temperature interval [30, 40], medium temperature interval [20, 30], low temperature interval [10, 20], and high temperature interval (40, 60], and the reward score for the normal state is 20, then the midpoint of the normal temperature interval is used as the starting point for allocation. That is, a temperature value of 35 has a reward score of 20, and the decrease value towards lower temperatures is... The decrease value in the direction of increasing temperature is For every degree the temperature changes, the reward points are distributed in a decreasing increment.
[0023] Methods for reinforcement learning on normal battery packs include: Initialize the action learning table and the exploration probability. Based on the battery state space, each data in the normal battery pack is assigned to the corresponding battery state, and an initial state score is assigned to each data in the normal battery pack based on the state reward table. Starting with the battery state of each data point in a normal battery pack, a random selection number is generated using a random algorithm to increase the diversity of battery action selection. When the random selection number is greater than or equal to the exploration probability, a battery action is selected from the battery action space using a random selection algorithm to update the battery state. The formula for state update is: ;in Represents the final state score. Represents a new battery state. This represents a battery action that updates its status. Represents the initial state score. Represents the initial battery state. This represents the initial battery action. Represents the selection of battery action After the battery state becomes The subsequent choice of state score, This represents the discount factor, used to balance changes in state components. This represents the battery action space. When the updated state score is greater than the initial state score, the action update table records the state update process, which includes state changes and changes in the initial state score. When the number of random selections is less than the exploration probability, the state update process in the action learning table is matched with the battery state. The state update process with the same battery state as the starting point and the largest change in the initial state score is found to update the battery state. This process is repeated until the battery state of each data point in the normal battery pack no longer changes. The action learning table at this point is the charge-discharge optimization strategy. Based on the charge-discharge optimization strategy, the battery is optimized for charge-discharge to improve work efficiency. In this embodiment, the preferred exploration rate is 0.1.
[0024] This embodiment identifies abnormal battery state data through fuzzy matrix and membership degree calculation, making anomaly detection more flexible and accurate. By removing abnormal data, the system obtains more reliable battery state data, providing accurate input for subsequent analysis and improving data effectiveness. Adaptive clustering algorithms group valid battery data, allowing the system to adaptively adjust the grouping method according to changes in battery state, avoiding the limitations of fixed clustering algorithms on battery pack diversity. By dynamically adjusting the initial battery centers and battery weights, the system obtains a more accurate distribution of similar battery packs, effectively improving grouping accuracy. Reinforcement learning dynamically adjusts the charging and discharging behavior of normal battery packs; using exploration probabilities and action learning tables, the system finds the optimal charging and discharging scheme after multiple updates. Reinforcement learning makes the charging and discharging optimization strategy more adaptable and flexible. Through continuous learning and adjustment, the system can manage battery packs more efficiently, avoiding battery damage caused by frequent charging and discharging, further improving the overall working efficiency and economic benefits of the electric loader.
[0025] Example 2 Please see Figure 2 As shown, for parts not described in detail in this embodiment, please refer to the description in Embodiment 1. A method for intelligent operation and maintenance management of battery packs for electric shovel loader is provided, including: S1. Collect battery status data of the battery pack, preprocess the battery status data, and obtain valid battery status data. S2. Perform adaptive clustering on the effective battery data to obtain N similar battery groups. Based on the constructed battery fault prediction model, predict the similar battery groups to obtain faulty battery groups and normal battery groups. S3. Construct a state reward and punishment table based on the faulty battery pack, perform reinforcement learning on the normal battery pack based on the state reward and punishment table to obtain a charge and discharge optimization strategy, and optimize the charge and discharge of the battery based on the charge and discharge optimization strategy to improve work efficiency.
[0026] Example 3 This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the operation mode of the intelligent operation and maintenance management platform for electric shovel loader battery packs described above.
[0027] Since the electronic device described in this embodiment is the electronic device used to implement the intelligent operation and maintenance management platform for electric shovel loader battery packs in this application embodiment, those skilled in the art can understand the specific implementation methods and various variations of the electronic device in this embodiment based on the intelligent operation and maintenance management platform for electric shovel loader battery packs described in this application embodiment. Therefore, how the electronic device implements the method in this application embodiment will not be described in detail here. As long as those skilled in the art implement the electronic device used in the intelligent operation and maintenance management platform for electric shovel loader battery packs in this application embodiment, it falls within the scope of protection of this application.
[0028] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.
[0029] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.
Claims
1. An intelligent operation and maintenance management platform for battery packs of electric loaders, characterized in that, include: Data acquisition and processing module: Collects battery status data from the battery pack, preprocesses the battery status data, and obtains valid battery status data; Anomaly detection module: Adaptively clusters valid battery data to obtain N similar battery groups, and predicts the similar battery groups based on the constructed battery fault prediction model to obtain faulty battery groups and normal battery groups; Strategy optimization module: Constructs a state reward and punishment table based on the faulty battery pack, performs reinforcement learning on the normal battery pack based on the state reward and punishment table to obtain a charge and discharge optimization strategy, and optimizes the charge and discharge of the battery based on the charge and discharge optimization strategy to improve work efficiency.
2. The intelligent operation and maintenance management platform for electric loader battery packs according to claim 1, characterized in that, The battery status data includes: battery temperature, voltage, current, impact intensity, and vibration frequency.
3. The intelligent operation and maintenance management platform for electric loader battery packs according to claim 2, characterized in that, The methods for preprocessing battery state data include: The standard score algorithm is used to normalize the battery state data to obtain a standard battery vector, and a system is built based on the standard battery vector. The fuzzy battery matrix, in which The data volume representing the standard battery vectors is used to perform fuzzy weighted calculations on different standard battery vectors to obtain fuzziness. This fuzziness is then mapped to a fuzzy battery matrix. The mean of each element in the fuzzy battery matrix is calculated to obtain the average fuzziness of each standard battery vector. The average fuzziness values are then sorted from largest to smallest, and the top values are selected. The standard battery vectors corresponding to the average fuzziness are used as fuzzy centers. Average membership is calculated for each standard battery vector based on these fuzzy centers to obtain the average membership degree. Anomaly membership thresholds are then defined based on the average membership degree. ;in, Represents the abnormal membership threshold. The mean of the average membership degree. This represents the threshold adjustment parameter. The variance of the average membership degree is used to mark the battery state data corresponding to the average membership degree that is greater than or equal to the abnormal membership threshold as abnormal. All data marked as abnormal in the battery state data are removed from the battery state data to obtain the valid battery state data.
4. The intelligent operation and maintenance management platform for electric loader battery packs according to claim 3, characterized in that, The formula for calculating the average membership is: ;in, Representing the The average membership degree of a standard battery vector. The number of fuzzy centers Representing the The standard battery vector and the first Similarity of fuzzy centers Representing the The fuzzy center and the first Similarity of fuzzy centers This represents the ambiguity parameter.
5. The intelligent operation and maintenance management platform for electric loader battery packs according to claim 4, characterized in that, The adaptive clustering method for valid battery data includes: Vectorize the effective battery state data to obtain effective battery vectors, and then randomly divide the effective battery vectors into... The process involves initializing the battery centers of each initial battery group, calculating the mean of each feature of the effective battery vectors in each initial battery group, and then using the mean of all features in each initial battery group to form the initial battery center. The battery weights of each initial battery group are initialized as the ratio of the number of effective battery vectors to the total number of effective battery vectors in that group. Based on the initial battery centers, the aggregation probability of each effective battery vector is calculated to obtain the aggregation probability. Based on the aggregation probability, the effective battery vectors are aggregated to the initial battery center with the highest aggregation probability to obtain new initial battery groups. The mean of each feature of the effective battery vectors in each new initial battery group is calculated to obtain new feature mean. All new feature mean values of each new initial battery group constitute a new initial battery center. Finally, the battery weights of each new initial battery group are updated using the following formula: ;in, The updated weight value representing the battery weight. The total amount of data representing the effective battery vector. The index represents the effective battery vector; the battery weights are updated with the updated weight value as the new value of the battery weights, and the new battery weights are obtained; this is repeated until the value of the battery weight of each initial battery group no longer changes, and the initial battery group at this time is the similar battery group.
6. The intelligent operation and maintenance management platform for electric loader battery packs according to claim 5, characterized in that, The formula for performing the aggregation calculation is: ;in, Representing the One effective battery vector, Representing the The effective battery vector is directed towards the first The aggregation probability of the initial battery center aggregation of the initial battery pack. The data dimension representing the effective battery vector, Representing the The covariance matrix of the initial battery pack Representing the The initial battery weights of the battery pack Representing the The inverse matrix of the covariance matrix of the initial battery pack Representing the The initial battery center of the initial battery pack. This represents the matrix transpose operation.
7. The intelligent operation and maintenance management platform for electric loader battery packs according to claim 6, characterized in that, The battery fault prediction model is constructed in the following ways: collection The system uses historical battery fault data and historical battery state data. The historical battery fault data represents the fault types that occurred in the past, while the historical battery state data has the same data type and format. These two sets of data are used as the training sample set. A GRU model is used as the initial model, and the GRU model is trained using the training sample set. The historical battery state data is used as the input data for the battery fault prediction model, and the predicted battery fault categories are used as the output data. The training objective is to minimize the error between the actual historical battery fault data and the predicted battery fault categories. The recall function is used as the loss function for the battery fault prediction model. Training stops when the loss function converges, resulting in the battery fault prediction model.
8. The intelligent operation and maintenance management platform for electric loader battery packs according to claim 7, characterized in that, The method for constructing a state reward table based on faulty battery packs includes: Based on the trained battery fault prediction model, the initial battery center of each group of similar battery packs is used as the input of the battery fault prediction model for detection. Similar battery packs whose battery fault categories are predicted are identified as faulty battery packs and fault alarms are issued. Similar battery packs whose battery fault categories are not predicted are identified as normal battery packs. Define battery state parameters, including battery temperature, voltage, and current parameters. Battery temperature parameters include: low temperature, medium temperature, normal temperature, and high temperature; voltage parameters include: low current, medium current, normal current, and high current; current parameters include: low voltage, medium voltage, normal voltage, and high voltage. Use a permutation and combination algorithm to arrange and combine the battery temperature, voltage, and current parameters to obtain battery states. All battery states constitute the battery state space. Define battery actions in the battery action space, including: increasing charging current, decreasing charging current, increasing discharging current, and decreasing discharging current. Battery groups with the same fault category are grouped together. The number of fault data points for each battery fault category is counted. Based on the number of fault data points, the battery fault categories are sorted in ascending order. Then, based on the sorting order, initial weights are applied. The fault weight is assigned to each battery fault category in ascending order, with each increment being [value missing]. The system calculates the fault intervals of each dimension in the data group of the same battery fault category. Based on the comparison between the fault interval and the battery status parameters, it assigns penalty scores to battery status parameters whose fault intervals match the battery status parameters. For the remaining battery status parameters that are not assigned penalty scores, the maximum value of the fault weight is used as the reward score for the normal category in each battery status parameter category. The reward scores are then assigned to other parameters in descending order of average value to obtain the status reward and penalty table.
9. The intelligent operation and maintenance management platform for electric loader battery packs according to claim 8, characterized in that, The methods for performing reinforcement learning on normal battery packs include: Initialize the action learning table and the exploration probability. Based on the battery state space, each data in the normal battery pack is assigned to the corresponding battery state, and an initial state score is assigned to each data in the normal battery pack based on the state reward table. Starting with the battery state of each data point in the normal battery pack, a random selection number is generated using a random algorithm. When the random selection number is greater than or equal to the exploration probability, a battery action is selected from the battery action space using the random selection algorithm to update the battery state. The formula for updating the state is as follows: ;in Represents the final state score. Represents a new battery state. This represents a battery action that updates its status. Represents the initial state score. Represents the initial battery state. This represents the initial battery action. Represents the selection of battery action After the battery state becomes The subsequent choice of state score, Represents the discount factor. This represents the battery action space. When the updated state score is greater than the initial state score, the action update table records the state update process, which includes state changes and changes in the initial state score. When the number of random selections is less than the exploration probability, the state update process in the action learning table is matched with the battery state. The state update process with the same battery state as the starting point and the largest change in the initial state score is found to update the battery state. This process is repeated until the battery state of each data point in the normal battery pack no longer changes. The action learning table at this point is the charge and discharge optimization strategy.
10. A method for intelligent operation and maintenance management of battery packs for electric loaders, implemented based on an intelligent operation and maintenance management platform for battery packs of electric loaders as described in any one of claims 1 to 9, characterized in that, include: S1. Collect battery status data of the battery pack, preprocess the battery status data, and obtain valid battery status data. S2. Perform adaptive clustering on the effective battery data to obtain N similar battery groups. Based on the constructed battery fault prediction model, predict the similar battery groups to obtain faulty battery groups and normal battery groups. S3, Strategy Optimization Module: Constructs a state reward and punishment table based on the faulty battery pack, performs reinforcement learning on the normal battery pack based on the state reward and punishment table to obtain a charge and discharge optimization strategy, and optimizes the charge and discharge of the battery based on the charge and discharge optimization strategy to improve work efficiency.