Slurry sensing and adaptive control method for lithium-ion battery homogenization process

Abstract

This invention discloses a slurry sensing and adaptive control method for the homogenization process of lithium-ion batteries, relating to the field of lithium-ion battery production technology. The method includes the following steps: obtaining the electrolyte concentration matrix and slurry flow velocity field of the target object; performing differential fusion on the electrolyte concentration matrix and the slurry flow velocity field to generate a local coupling index; performing weighted autoregression calculation on the electrolyte concentration matrix, and correcting the weighted autoregression calculation result based on the local coupling index and the slurry flow velocity field to generate an electrolyte concentration prediction result. This scheme extracts the electrolyte element set at the most recent A time points along the time axis for each grid point and performs weighted autoregression calculation. In the time dimension, it can differentiate the weighting of data according to the importance of different time points, further performing flow field diffusion calculation on the slurry flow velocity field and concentration gradient field, and using the local coupling index to spatially correct the weighted autoregression result.

Description

Technical Field

[0001] This invention relates to the field of lithium-ion battery production technology, specifically to a method for slurry sensing and adaptive control in the homogenization process of lithium-ion batteries. Background Technology

[0002] As a core energy storage device in new energy vehicles and portable electronic devices, the stability of the electrode quality in the production process of lithium-ion batteries directly affects the cycle life and safety performance of the battery. Electrodes are usually prepared by wet process, which involves mixing raw materials such as active materials, conductive agents, binders and solvents in proportion to form a slurry, mixing it evenly in a homogenization process, and then coating, drying and rolling.

[0003] In the aforementioned technological context, existing technologies often use conductivity sensors to monitor the local electrolyte concentration in the slurry, supplemented by flow velocity sensors to estimate the local flow state, and then combine this with a threshold judgment strategy to determine the non-uniformity of the slurry. These solutions have the advantages of simple structure and easy implementation, and can alarm or trigger control actions for obvious segregation phenomena. However, in actual production, the method of triggering control actions only when the electrolyte concentration exceeds a preset threshold makes it difficult to perform feedforward assessment of the concentration evolution trend of the slurry on a short timescale. This makes it difficult to identify potential abnormal areas that are "not yet exceeded but have been continuously rising," resulting in these areas exhibiting process defects such as locally high concentrations and uneven thickness in subsequent coating stages. The lack of concentration change trend risk assessment means there is a lack of ability to dynamically adjust zoning based on the degree of risk. Consequently, when high-risk areas migrate spatially due to changes in the flow field, the zoning results become disconnected from the actual risk distribution, resulting in some truly high-risk areas not being controlled in a timely and accurate manner. Summary of the Invention

[0004] The purpose of this invention is to provide a slurry sensing and adaptive control method for the homogenization process of lithium-ion batteries, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows: In a first aspect, the present invention discloses a slurry sensing and adaptive control method for the homogenization process of lithium-ion batteries, comprising the following steps: Obtain the electrolyte concentration matrix and slurry flow velocity field of the target object; The electrolyte concentration matrix and the slurry velocity field are differentially fused to generate a local coupling index; A weighted autoregression calculation is performed on the electrolyte concentration matrix, and the weighted autoregression calculation results are corrected according to the local coupling index and the slurry flow velocity field to generate electrolyte concentration prediction results; The difference between the predicted electrolyte concentration and the preset concentration threshold is calculated, and a risk heat map is generated according to the ratio of the difference to the preset concentration threshold. Then, based on the risk heat map, the predicted electrolyte concentration, the local coupling index, and the flow velocity modulus are summed in segments to generate a partition sensitivity matrix. The velocity modulus is obtained by calculating the vector modulus of the slurry velocity field. A constraint function with the objective of minimizing the concentration variance is constructed and solved for the partition sensitivity matrix, the electrolyte concentration matrix and the flow velocity modulus to generate a partition control vector. Then, control parameters are transformed and reflux paths are screened for the partition control vector and the partition sensitivity matrix to generate a control scheme.

[0006] In a second aspect, the present invention discloses a slurry sensing and adaptive control system for the homogenization process of lithium-ion batteries, comprising: The data acquisition module is used to acquire the electrolyte concentration matrix and slurry flow velocity field of the target object; The differential processing module is used to perform differential fusion of the electrolyte concentration matrix and the slurry flow velocity field to generate a local coupling index; The prediction module is used to perform weighted autoregression calculation on the electrolyte concentration matrix, and correct the weighted autoregression calculation results according to the local coupling index and the slurry flow velocity field to generate electrolyte concentration prediction results. The partition analysis module is used to calculate the difference between the electrolyte concentration prediction result and the preset concentration threshold, and generate a risk heat map according to the ratio of the difference to the preset concentration threshold. Then, based on the risk heat map, the electrolyte concentration prediction result, the local coupling index and the flow velocity modulus are segmented and weighted to generate a partition sensitivity matrix. The regulation scheme generation module is used to construct and solve a constraint function with the objective of minimizing the concentration variance for the partition sensitivity matrix, the electrolyte concentration matrix, and the flow velocity modulus, generate a partition control vector, and perform control parameter transformation and reflux path screening on the partition control vector and the partition sensitivity matrix to generate a regulation scheme.

[0007] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This scheme extracts the electrolyte element set at the most recent A time points along the time axis for each grid point and performs weighted autoregression calculation. In the time dimension, it can differentiate the weighting of data according to the importance of different time points, and further perform flow field diffusion calculation on the slurry velocity field and concentration gradient field. The local coupling index is used to spatially correct the weighted autoregression results, which can introduce the influence of convection-diffusion behavior on concentration evolution into the prediction process, making the prediction results closer to the actual flow state in spatial distribution, thereby improving the effectiveness of subsequent zonal control and anomaly repair.

[0008] 2. This scheme calculates the ratio of the difference to the preset concentration threshold and maps it to a risk heatmap, which can quantitatively highlight high-risk grid areas. By weighted normalization of the electrolyte element set and slurry flow velocity field, the preset concentration threshold is determined, allowing the concentration threshold to be adaptively adjusted according to the formula and operating conditions, reducing the probability of false alarms and false alarms. By clustering analysis of the risk heatmap and local coupling index to generate a partition identification matrix, and within each partition, the electrolyte concentration prediction result, local coupling index and flow velocity modulus are segmented and weighted to generate a partition sensitivity matrix according to the risk heatmap. This allows the control system to allocate stirring and reflux intensity according to the sensitivity differences of different partitions, thereby improving control accuracy and reducing overall energy consumption. Attached Figure Description

[0009] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein: Figure 1 This is a flowchart of the slurry sensing and adaptive control method for the lithium-ion battery homogenization process of the present invention. Figure 2 This is a schematic diagram of the process for generating electrolyte concentration prediction results provided by the present invention. Figure 3 This is a schematic diagram of the process for generating a partition sensitivity matrix provided by the present invention; Figure 4 A schematic diagram of the generation and control scheme provided by the present invention; Figure 5 This is a schematic diagram of the module functions of the slurry sensing and adaptive control system for the homogenization process of lithium-ion batteries provided by the present invention. Detailed Implementation

[0010] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0011] Application Overview: In the traditional lithium-ion battery homogenization process, the slurry monitoring system only triggers control actions based on whether the electrolyte concentration exceeds a preset threshold. It lacks a feedforward assessment mechanism for the short-term evolution trend of concentration, making it difficult to identify potential abnormal areas that have not yet exceeded the limit but are continuously rising. This problem stems from the fact that the existing technology does not incorporate the concentration change trend into the risk assessment framework, making it difficult for the zoning results to dynamically match the actual risk distribution. Consequently, when high-risk areas migrate spatially due to changes in the flow field, they cannot be controlled in a timely and accurate manner, ultimately affecting the key indicators of electrode quality stability and battery cycle life and safety performance.

[0012] For example, in the slurry mixing stage of a new energy vehicle battery production line, when the slurry delivery pump experiences sudden changes in local flow rate due to mechanical fluctuations, the electrolyte concentration, although not reaching the preset threshold, shows a continuous upward trend. The existing system only monitors the instantaneous concentration value and fails to generate a prediction result of the concentration evolution trend. This causes the potentially abnormal area to exhibit process defects such as localized high concentration and uneven thickness in the subsequent coating stage. At the same time, the fixed zoning strategy is difficult to dynamically adjust the control parameters based on the risk heat map, resulting in the high-risk area after flow field migration becoming disconnected from the zoning control vector, further aggravating the spread of process defects.

[0013] If the above problems are not addressed, potential abnormal areas will continue to accumulate process deviations due to the lack of trend warnings. Process defects may extend along the coating direction and affect the overall uniformity of the electrode. The resulting localized high concentration will lead to uneven stress distribution during battery charging and discharging, thereby reducing battery cycle life and increasing the risk of thermal runaway, ultimately damaging the reliability and safety of the battery system.

[0014] After introducing the basic concept of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0015] Example 1: Please see Figure 1 A slurry sensing and adaptive control method for the homogenization process of lithium-ion batteries includes the following steps: Obtain the electrolyte concentration matrix and slurry flow velocity field of the target object; The electrolyte concentration matrix and slurry velocity field are differentially fused to generate a local coupling index; Weighted autoregression calculations are performed on the electrolyte concentration matrix, and the weighted autoregression calculation results are corrected based on the local coupling index and slurry velocity field to generate electrolyte concentration prediction results. The difference between the predicted electrolyte concentration and the preset concentration threshold is calculated, and a risk heat map is generated according to the ratio of the difference to the preset concentration threshold. Then, based on the risk heat map, the predicted electrolyte concentration, local coupling index and flow velocity modulus are summed in segments to generate a regional sensitivity matrix. The velocity modulus is obtained by calculating the vector modulus of the slurry velocity field. A constraint function with the objective of minimizing the concentration variance is constructed and solved for the zonal sensitivity matrix, electrolyte concentration matrix, and flow velocity modulus to generate zonal control vectors. Then, control parameters are transformed and reflux paths are screened for the zonal control vectors and zonal sensitivity matrix to generate a control scheme.

[0016] Among them, the electrolyte concentration matrix refers to a digital data structure used to characterize the spatial distribution of electrolyte concentration of the target object on a preset computing grid in the lithium-ion battery homogenization process; Slurry velocity field refers to a set of spatial vector data constructed in the homogenization process of lithium-ion batteries, with the spatial position in the homogenization container or conveying pipeline as the independent variable and the instantaneous flow state of the slurry as the characterization object. Differential fusion refers to the process of uniformly quantifying the differences between electrolyte concentration matrix and slurry velocity field in time, space and trend dimensions based only on the obtained electrolyte concentration matrix and slurry velocity field, and generating a local coupling index that can characterize the interaction intensity between electrolyte distribution and slurry flow through regular combination. The local coupling index is a quantitative data index used to characterize the coupling strength between the electrolyte concentration change behavior and the slurry flow behavior at the same spatial location in the lithium-ion battery homogenization process. Weighted autoregression calculation refers to the process of using the acquired electrolyte concentration matrix as the only time series input data, and generating the predicted value of the basic electrolyte concentration at the current moment by performing regression modeling on the electrolyte concentration matrix in the time dimension without introducing external state variables. Electrolyte concentration prediction results refer to the set of predicted data on the distribution of electrolyte concentration at various spatial locations within a future preset time window; The preset concentration threshold is a numerical limit used in the lithium-ion battery homogenization process to determine whether there is a risk of abnormal segregation in the predicted electrolyte concentration. It is obtained by performing distribution statistics on the electrolyte element set and then weighting and normalizing the distribution statistics results in combination with the slurry flow velocity field. A risk heatmap is a data structure used to describe the relative risk level distribution of abnormal electrolyte concentration segregation at different spatial locations during the homogenization process. The velocity modulus refers to the non-directional scalar value calculated from the vector modulus based on the instantaneous velocity vector of the slurry at each spatial location in the slurry velocity field during the homogenization process of lithium-ion batteries. Segmented weighted summation refers to the process of dividing each grid point in the computational space into multiple risk level segments based on the risk heat map, and then performing weighted calculations on the electrolyte concentration prediction results, local coupling index, and flow velocity modulus for the same grid point within each risk level segment, according to the weight set matched to the corresponding risk level. The partition sensitivity matrix refers to a structured data matrix used to characterize the response intensity of different spatial partitions within a homogenizing tank to control actions; Vector modulus calculation refers to the process of converting each velocity vector into a corresponding non-negative real value through vector modulus operation without changing the original spatial distribution structure of the slurry velocity field, thereby forming a velocity modulus field that corresponds one-to-one with the slurry velocity field. The partition control vector refers to a set of multi-dimensional vectors used to characterize the control amplitude and control priority of each control execution unit; Control parameter transformation refers to the process of converting the numerical control results into a set of control command parameters with physically executable semantics after obtaining the partition control vector and partition sensitivity matrix, according to the mapping rule set pre-stored in the system. Backflow path screening refers to the process of traversing, screening, and selecting executable slurry flow paths based on the known connection relationships of pipelines, valves, pumping nodes, and backflow channels within a homogenization system, using the partition control vector and partition sensitivity matrix as search constraints, in order to determine the optimal or feasible combination of slurry control paths that meet the target control requirements. The control scheme refers to a set of structured control results generated by the system after solving the partition control vector, which can be directly parsed and executed by the execution layer.

[0017] This scheme achieves a quantitative characterization of the concentration distribution and flow state in the homogenization region by acquiring the electrolyte concentration matrix and slurry velocity field, providing a unified data foundation for all subsequent calculations. By differentially fusing the electrolyte concentration matrix and slurry velocity field to generate a local coupling index, the system can characterize the local instability sensitivity caused by the superposition of concentration gradient and flow intensity, in addition to single concentration or velocity indices. Through weighted autoregression of the electrolyte concentration matrix and correction using the local coupling index and slurry velocity field, electrolyte concentration prediction results are obtained, which can be used to characterize the concentration evolution trend of each grid point on a short timescale in advance. The difference between the predicted electrolyte concentration and the preset concentration threshold is calculated, and a risk heat map is generated according to the ratio of the difference to the concentration threshold. Based on this, the predicted electrolyte concentration, local coupling index, and flow velocity modulus are weighted and summed in segments to generate a zonal sensitivity matrix, so that high-risk zones are explicitly highlighted in subsequent control. Then, a constraint optimization function with the objective of minimizing the concentration variance is constructed and solved for the zonal sensitivity matrix, electrolyte concentration matrix, and flow velocity modulus to obtain the zonal control vector. After regularization mapping and pipeline topology search, a control scheme is formed to output stirring, conveying, and recirculation control commands that meet the actuator capabilities and safety constraints.

[0018] The above describes a complete scheme for slurry sensing and adaptive control in the homogenization process of lithium-ion batteries. The following section describes how to obtain the electrolyte concentration matrix and slurry flow velocity field of the target object, specifically including: The electrolyte concentration data and slurry flow rate data of the target object are obtained and then meshed using grid points as the basic unit. Coordinate interpolation transformation is performed on electrolyte concentration data and slurry flow rate data to generate a spatial grid mapping; Based on the spatial grid mapping, spatiotemporal interpolation is performed on the electrolyte concentration data and slurry flow velocity data to generate the electrolyte concentration matrix and slurry flow velocity field.

[0019] Among them, electrolyte concentration data refers to the set of raw sensor data used to characterize the instantaneous spatial distribution of electrolytes in slurry; Slurry flow rate data refers to the basic set of raw data that reflects the flow state of slurry in the homogenizing tank over time during the homogenizing process; Coordinate interpolation transformation refers to the process of calculating the spatial relationship between the original discrete sampling coordinates of the sensor and the preset calculation grid coordinates, based on the acquired electrolyte concentration data and slurry flow rate data, with the sensor installation coordinates as the input. Spatial grid mapping refers to establishing a one-to-one or one-to-many mapping relationship between the sensor spatial locations corresponding to the electrolyte concentration data and the preset computational grid nodes, based on only acquiring electrolyte concentration data and slurry flow rate data. Spatiotemporal interpolation refers to a data processing procedure that uses three-dimensional grid points determined by spatial grid mapping as a unified calculation benchmark, and performs interpolation reconstruction processing on the two types of data in the spatial and temporal dimensions, respectively, using only the acquired electrolyte concentration data and slurry flow rate data.

[0020] The above content will be described in detail below: Electrolyte concentration data of the target object is obtained by pre-deploying electrolyte concentration sensors; the electrolyte concentration data includes, but is not limited to, instantaneous electrolyte concentration values, sampling timestamp information, and spatial coordinate information of the installation location of the electrolyte concentration sensors; The slurry flow rate data of the target object is obtained by pre-deployed slurry flow rate sensors; the slurry flow rate data includes, but is not limited to, the instantaneous flow rate value at each sampling location, the corresponding flow rate direction information, the flow rate time series data, and the spatial coordinate information of the slurry flow rate sensor installation location, etc. According to the preset spatial division rules, the homogenization area corresponding to the target object is divided into several grid points, and each grid point is defined as a basic data processing unit. Then, according to the spatial relationship between the sensor installation position coordinates and each grid point, the acquired electrolyte concentration data and slurry flow rate data are respectively processed into grids. Using the sensor's installation location coordinates and its corresponding electrolyte concentration and slurry flow rate data as known sampling point data, at least one interpolation operator among distance-weighted interpolation, linear interpolation, or spline interpolation is used to perform numerical interpolation calculations on the known sampling point data at grid points. This yields the corresponding interpolated electrolyte concentration value and interpolated slurry flow rate vector value at each grid point. The spatial coordinates of each grid point are associated with and stored with their corresponding interpolated electrolyte concentration value and interpolated slurry flow rate vector value to form a spatial grid mapping. The spatiotemporal interpolation process for electrolyte concentration data is as follows: According to the coordinate relationship between the installation location coordinates and the target grid point recorded in the spatial grid mapping, the interpolated electrolyte concentration value in the neighborhood of each grid point is read, and the interpolated electrolyte concentration value is spatially interpolated based on the preset interpolation operator. The preset interpolation operator is preferably inverse distance weighted interpolation or linear interpolation, which is used to calculate the electrolyte concentration estimate of the grid point at the current time. Then, combined with the adjacent sampling time in the time dimension, the electrolyte concentration estimate is subjected to temporal interpolation or resampling processing to obtain the grid-level electrolyte concentration value under a unified time reference, thereby constructing an electrolyte concentration matrix node by node. The spatiotemporal interpolation process for slurry velocity data is as follows: the installation position coordinates of the slurry velocity sensor are aligned according to the spatial grid mapping, the interpolated slurry velocity vector value in the neighborhood of each grid point is read, and spatial interpolation is performed on each component of the interpolated slurry velocity vector value to obtain the velocity vector of the grid point. Then, the velocity vector is subjected to temporal interpolation or smoothing processing in combination with the adjacent sampling time after time synchronization to form a grid-level slurry velocity vector under a unified spatiotemporal reference, thereby constructing a slurry velocity field. The electrolyte concentration matrix and slurry velocity field are differentially fused to generate a local coupling index, the specific calculation formula of which is as follows: ; In the formula, Represents grid points The local coupling index, Represents grid points The electrolyte concentration spatial gradient vector, Represents grid points The velocity vector in the slurry flow field Indicates modulo operation. Represents grid points The concentration value in the electrolyte concentration matrix corresponding to the current moment. Represents grid points The concentration value in the electrolyte concentration matrix corresponding to the previous sampling time. This indicates the time interval between two consecutive electrolyte concentration data samplings. and These represent the corresponding weighting coefficients. All the data above have been normalized during the calculation.

[0021] This scheme acquires electrolyte concentration data and slurry flow rate data and performs gridding processing based on grid points, unifying discrete sampling into a regular grid. This facilitates calculation and comparison within the same spatial unit. Coordinate interpolation transformation is performed on the electrolyte concentration data and slurry flow rate data to generate a spatial grid mapping, which is used to correct spatial deviations caused by irregular sensor positions and improve data spatial alignment accuracy. Then, spatiotemporal interpolation is performed on the electrolyte concentration data and slurry flow rate data based on the spatial grid mapping to generate an electrolyte concentration matrix and a slurry flow rate field. This structures the input under a unified coordinate and time reference, facilitating subsequent coupled modeling and control calculations.

[0022] The above describes how to obtain the electrolyte concentration matrix and slurry velocity field of the target object. The following describes the weighted autoregression calculation of the electrolyte concentration matrix, and the correction of the weighted autoregression calculation results based on the local coupling index and slurry velocity field to generate the electrolyte concentration prediction results. Please refer to [link / reference]. Figure 2 , Figure 2This is a schematic diagram of the process for generating electrolyte concentration prediction results provided in this application embodiment. Generating electrolyte concentration prediction results specifically includes: For each grid point, extract the electrolyte element set of the most recent A time points according to the time axis of the electrolyte concentration matrix, and perform weighted autoregression calculation on the electrolyte element set; The flow field diffusion of the slurry velocity field and concentration gradient field is calculated, and the weighted autoregression calculation results are corrected based on the local coupling index and the flow field diffusion calculation results to generate the electrolyte concentration prediction results for each grid point. The concentration gradient field is obtained by performing a central difference operation on the electrolyte concentration matrix within the grid neighborhood.

[0023] Among them, the electrolyte element set refers to the ordered data set composed of multiple electrolyte concentration values ​​extracted from the electrolyte concentration matrix in chronological order along the time axis for the same spatial grid point when performing time series analysis on the electrolyte concentration matrix. Concentration gradient field refers to vector field data used to characterize the spatial direction and magnitude of changes in electrolyte concentration; Flow field diffusion calculation refers to the process by which the system numerically calculates the convective diffusion trend of electrolyte concentration during slurry flow on a grid scale, based on the obtained slurry velocity field and concentration gradient field. Central difference operation refers to a numerical discretization method used to calculate the spatial rate of change of electrolyte concentration on a regular grid.

[0024] The above content will be described in detail below: Extract the set of electrolyte elements for the most recent A time points from the electrolyte concentration matrix point by point along the time axis; where A is obtained as follows: For each grid point, the concentration sequence of the most recent B consecutive time steps (B is a fixed upper limit constant) is read from the electrolyte concentration matrix along the time axis. The adjacent sampled differential amplitude sequence is calculated by interpolation on this concentration sequence, and the moving statistics (such as moving mean and moving variance) of the adjacent sampled differential amplitude sequences are calculated. Then, the stability index and abrupt change index of this grid point are calculated. The stability index is obtained by adding the moving mean and moving variance, and then dividing the moving mean by the sum. The abrupt change index is obtained by dividing the maximum value of the adjacent sampled differential amplitude sequences by the moving mean. Based on the stability index of this grid point With mutation index The calculation yields A, and the specific formula for its calculation is as follows: ; In the formula, Represents grid points The value of A, This indicates the minimum number of allowed replay times. Indicates the maximum number of allowed replay times. and These represent the corresponding weight coefficients. This indicates the floor operator. This represents the truncation function; all the data above have been normalized during the calculation. The weighted autoregression calculation for the electrolyte element set is calculated using the following formula: ; In the formula, Represents grid points In time The weighted autoregression calculation results, Represents grid points The Middle Autoregressive weights for each historical moment Represents grid points In time The electrolyte concentration values ​​are shown above, and all data have been normalized during calculation. Perform a dot product operation on the slurry velocity field and concentration gradient field to generate a local convection flux term; The gradient difference results of the concentration gradient field at adjacent grid points are calculated by second-order spatial difference to generate diffusion driving terms. The local convection flux term and diffusion driving term are linearly combined according to preset physical weight coefficients (e.g., the physical weight coefficients of both types of data are 0.5) to generate the flow field diffusion calculation results. Based on the calculation results of local coupling index and flow field diffusion, the weighted autoregression calculation results are corrected by multiplication operation to generate the electrolyte concentration prediction results for each grid point; The concentration gradient field is obtained by performing a central difference operation on the electrolyte concentration matrix within the grid neighborhood: Specifically, in the X direction, the concentration change rate is obtained by reading the concentration values ​​of the adjacent grid points to the right and the adjacent grid points to the left of the current grid point, calculating the difference between the two and dividing by the corresponding spatial step size. The same data reading and calculation method is used in the Y and Z directions to obtain the concentration change rate in the corresponding directions. The concentration change rates of each grid point in the X, Y and Z directions are combined to generate a concentration gradient field. Calculate the difference between the predicted electrolyte concentration and the preset concentration threshold.

[0025] This scheme extracts the electrolyte element set at the most recent A time points along the time axis for each grid point and performs weighted autoregression calculations. This filters out random measurement noise while preserving the concentration evolution trend, improving the stability and accuracy of time prediction. By performing flow field diffusion calculations on the slurry velocity field and the concentration gradient field obtained by central difference of the electrolyte concentration matrix, and correcting the autoregression results based on the local coupling index, the convection-diffusion coupling effect is explicitly introduced. This finely characterizes the impact of mass transport between grid points on local concentration, improving the accuracy of spatial distribution prediction and the sensitivity to segregation risk.

[0026] The above describes the weighted autoregression calculation of the electrolyte concentration matrix, followed by correction of the weighted autoregression results based on the local coupling index and slurry velocity field to generate electrolyte concentration prediction results. The following describes the generation of risk heatmaps based on the ratio of the difference to a preset concentration threshold. Then, based on the risk heatmaps, the electrolyte concentration prediction results, local coupling index, and velocity modulus are summed in segments with weights to generate a regional sensitivity matrix. Please refer to [reference needed]. Figure 3 , Figure 3 This is a schematic diagram of the process for generating a partition sensitivity matrix provided in an embodiment of this application. Generating the partition sensitivity matrix specifically includes: Calculate the ratio of the difference to the preset concentration threshold, perform statistical mapping on the comparison values, and generate a risk heatmap; The preset concentration threshold is obtained by statistically analyzing the distribution of electrolyte elements and then weighting and normalizing the statistical results in combination with the slurry flow velocity field. The risk heatmap and local coupling index are weighted and clustered within the grid neighborhood to generate a partition identification matrix. Based on the partition identification matrix, the electrolyte concentration prediction results, local coupling index and flow velocity modulus are piecewise weighted and summed in each partition according to the risk heatmap to generate a partition sensitivity matrix.

[0027] Statistical mapping refers to the calculation process of mapping the ratio of the difference to a preset concentration threshold from a continuous numerical space to a discrete risk level space after obtaining the ratio. Cluster analysis refers to the calculation process of automatically dividing grid points with similar risk characteristics and coupling behavior into the same control partition by using risk heatmap and local coupling index as the only numerical features to measure similarity and group them under spatial constraints. The partition identifier matrix is ​​a data structure used to represent the control partition number to which each grid point belongs.

[0028] The above content will be described in detail below: Calculate the ratio of the difference to the preset concentration threshold; Comparison values ​​are statistically mapped to generate a risk heatmap: Feature statistics are performed using comparative values, including calculating the mean and standard deviation of the ratios for each grid point, and then a threshold set containing three layers of thresholds is constructed: The first threshold is obtained by adding the standard deviation of the ratio to the mean of the ratios by a first factor (e.g., 0.65). The second threshold is obtained by adding a second multiple (e.g., 0.35 times) of the standard deviation of the ratio to the mean of the ratios; The third threshold is obtained by subtracting the standard deviation of the ratio by a third multiple (e.g., 0.45 times) from the mean of the ratios; The ratio is compared with a set of thresholds to generate a risk heatmap: If the ratio is greater than the first threshold, the grid point corresponding to the ratio is marked as high-risk. If the ratio is between the second threshold and the first threshold, the grid point corresponding to the ratio is marked as medium-risk. If the ratio is between the third threshold and the second threshold, the grid point corresponding to the ratio is marked as low-risk. A risk heat map is generated by combining the risk level information of each grid point. The preset concentration threshold is obtained by statistically analyzing the distribution of electrolyte elements and then weighting and normalizing the statistical results in conjunction with the slurry flow velocity field. The specific calculation formula is as follows: ; In the formula, Represents grid points The corresponding preset concentration threshold, This indicates the upper limit of the maximum allowable concentration threshold. Represents grid points The mean of the corresponding electrolyte element set, Represents grid points The standard deviation of the corresponding electrolyte element set, Represents grid points The average slurry flow rate over the most recent A time intervals. Represents positive numbers. This represents the concentration fluctuation adjustment coefficient. This represents the velocity weighting factor. This indicates that the weighted statistical result of all grid points is taken as the maximum value. All the above data have been normalized during the calculation. The risk heatmap and local coupling index are weighted, fused, and clustered within the grid neighborhood to generate a partition identification matrix. The risk level information and local coupling index of the corresponding grid point in the risk heatmap are read point by point. Then, taking the current grid point as the center, a set of grid neighborhoods is determined according to a preset spatial neighborhood radius (e.g., a six-neighborhood or twenty-six-neighborhood structure). Within this grid neighborhood, the risk heatmap and local coupling index are weighted and fused separately. The specific data processing method is as follows: First, the risk heatmap and local coupling index of each grid point in the grid neighborhood are normalized by applying a decay weight representing the reciprocal of the spatial distance to the central grid point, so as to obtain the neighborhood risk weight value and the neighborhood coupling weight value. Then, according to the preset fusion rule, the neighborhood risk weight value and the neighborhood coupling weight value are combined with linear weights (e.g., the weighting coefficient of the neighborhood risk weight value is 0.3 and the weighting coefficient of the neighborhood coupling weight value is 0.7) to generate the neighborhood fusion feature. After obtaining the neighborhood fusion features corresponding to each grid point, the neighborhood fusion features of all grid points are constructed into a fusion feature matrix, and clustering analysis is performed on the fusion feature matrix. The clustering analysis can be distance-based clustering or similarity-based clustering. The similarity between neighborhood fusion features is calculated (e.g., Euclidean distance similarity or cosine similarity). Based on the similarity between neighborhood fusion features, the grid points are automatically divided into several categories, each category corresponds to a partition, and a partition label matrix is ​​generated. Based on the partition identifier matrix, the predicted electrolyte concentration, local coupling index, and flow velocity modulus within each partition are summed in segments according to the risk heatmap to generate a partition sensitivity matrix: Based on the risk heatmap, different adjustment coefficients are assigned to grid points of different risk levels. For example, the adjustment coefficient for grid points of high risk level is 0.5, the adjustment coefficient for grid points of medium risk level is 0.3, and the adjustment coefficient for grid points of low risk level is 0.2. Within each partition, the predicted electrolyte concentration, local coupling index, and flow velocity modulus are weighted and normalized according to the adjustment coefficient of the grid point to obtain the sub-sensitivity value of the corresponding grid point. The sub-sensitivity values ​​of all grid points in the partition are then weighted and averaged to obtain the sensitivity of the partition. The sensitivity of each partition is combined to generate a partition sensitivity matrix. The velocity modulus is obtained by calculating the vector modulus of the slurry velocity field, that is, by summing the squares of each component of the slurry velocity field at the grid points and performing square root calculations. A constraint function with the objective of minimizing the concentration variance is constructed and solved for the zoning sensitivity matrix, electrolyte concentration matrix, and flow velocity modulus to generate the zoning control vector. The specific calculation formula is as follows: ; In the formula, Indicates partition The corresponding partition control vector, Indicates partition The vector of control variables to be determined. Indicates partition The set of grid points contained therein Represents grid points In the elements of the electrolyte concentration matrix, Indicates partition The sensitivity coefficients are calculated using normalized data.

[0029] This scheme generates a risk heatmap by calculating the difference between the predicted electrolyte concentration and the preset concentration threshold and performing statistical mapping. This quantifies and classifies the deviation of each grid point, facilitating the priority identification of high-risk areas. The preset concentration threshold is determined by statistically analyzing the distribution of electrolyte elements and combining it with weighted normalization of the slurry velocity field. This allows the concentration threshold to be updated adaptively with the formulation and flow field conditions, thereby improving the accuracy and robustness of risk identification. By weightedly fusing the risk heatmap and local coupling index within the grid neighborhood and performing cluster analysis, a partition identification matrix is ​​generated. This allows the control partition to simultaneously characterize the segregation risk level and fluid coupling characteristics, ensuring that the partition boundaries more closely match actual physical differences. Within each partition, the predicted electrolyte concentration, local coupling index, and velocity modulus are segmented and weighted according to risk level to generate a partition sensitivity matrix. This allows the comprehensive response capability of the partition to the intensity of regulation and the priority of execution to be converged into a single quantitative index, which is convenient for subsequent control algorithms to directly call and compare.

[0030] The above describes the generation of risk heatmaps based on the ratio of the difference to a preset concentration threshold. Then, based on the risk heatmap, the electrolyte concentration prediction results, local coupling index, and flow velocity modulus are weighted and summed piecewise to generate a regional sensitivity matrix. The following describes the generation of abnormal grid points after generating the regional control vector, specifically including: The difference between the electrolyte concentration matrix and the electrolyte concentration prediction result is calculated for each grid point. It is then determined whether the difference calculation result is greater than the preset discrimination threshold. If so, the grid point corresponding to the difference calculation result is determined to be an abnormal grid point. The preset discrimination threshold is dynamically set based on the risk information of the grid points within the risk heatmap.

[0031] The preset discrimination threshold is a numerical limit used to measure whether the current concentration deviation of a single grid point constitutes an anomaly. It is obtained by mapping the risk information of the grid point in the risk heatmap to a risk adjustment factor and multiplying the risk adjustment factor with the element of the grid point in the electrolyte concentration matrix. Abnormal grid points are those that are determined to have significant deviations and require adjustment.

[0032] The above content will be described in detail below: The difference between the electrolyte concentration matrix and the electrolyte concentration prediction result is calculated for each grid point. It is then determined whether the difference calculation result is greater than the preset discrimination threshold. If so, the grid point corresponding to the difference calculation result is determined to be an abnormal grid point. The preset discrimination threshold is dynamically set based on the risk information of the grid points within the risk heatmap: Extract the risk level information of grid points within the risk heatmap, and extract the basic threshold corresponding to the risk level information from the system's built-in parameter library. For example, the basic threshold corresponding to the high risk level is 0.65, the basic threshold corresponding to the medium risk level is 0.35, and the basic threshold corresponding to the low risk level is 0.15. The grid points included in each risk level are sorted in ascending order of their ratios to obtain the relative position information of each grid point within its corresponding risk level. Then, the preset discrimination threshold for each grid point is calculated. The specific calculation formula is as follows: ; In the formula, Represents grid points Risk level information Represents grid points The smoothing weight is determined by dividing the relative position information of a grid point in its corresponding risk level by the total number of grid points included in the corresponding risk level. This represents the basic threshold corresponding to the risk level. All the above data have been normalized during the calculation.

[0033] This scheme calculates the difference between the electrolyte concentration matrix and the predicted electrolyte concentration at each grid point and compares it with the discrimination threshold. This enables the system to accurately locate areas of concentration deviation at the grid scale, avoiding missed anomalies or coarse localization caused by relying solely on overall statistical indicators. The discrimination threshold is dynamically set based on the risk level of each grid point in the risk heatmap, allowing the anomaly detection sensitivity to be adaptively adjusted according to different operating conditions and risk distributions, reducing false alarms and mis-controls, and improving the stability and reliability of anomaly detection.

[0034] The above describes the process after generating the partition control vector, which also includes generating abnormal grid points. The following section describes the control parameter transformation and backflow path filtering performed on the partition control vector and partition sensitivity matrix to generate a control scheme. Please refer to [link / reference]. Figure 4 , Figure 4 This is a flowchart illustrating the generation of a control scheme provided in an embodiment of this application. The generation of the control scheme specifically includes: For each anomaly grid point, the partition control vector, partition sensitivity matrix, and local coupling index are solved through mathematical modeling and real-time constraint optimization to generate a set of repair actions. Based on the spatial grid mapping and partition identification matrix, a recirculation scheme is generated for the repair action set through pipeline topology modeling and constraint timing scheduling; A candidate action list is constructed based on the zoning control vector and the reflux scheme. The feasibility and conflict of the candidate action list are then verified based on the slurry velocity field and the zoning sensitivity matrix to generate a control scheme.

[0035] Among them, mathematical modeling refers to the constraint optimization modeling process oriented towards abnormal grid points and using actuator actions as solution variables; Real-time constraint optimization solution refers to the data processing process that uses abnormal grid points as the smallest decision unit, and constructs and solves a constraint optimization model with the goal of maximizing the repair effect and minimizing the regulation cost based on the obtained partition control vector, partition sensitivity matrix, and local coupling index, without introducing new data types. The set of repair actions refers to the minimum set of intervention actions used to restore the electrolyte balance in an abnormal region. Pipeline topology modeling refers to the data processing process that, based on the acquired spatial grid mapping, partition identification matrix, and slurry velocity field, structurally abstracts the physical pipelines, valves, pumps, and their connection relationships related to reflux, transportation, and diversion in the homogenization system, forming a directed constrained topology model that can be used for algorithm search and time-series scheduling. Constrained timing scheduling refers to the data processing process that uses the set of repair actions as the set of execution objects, without introducing new sensor data, and calculates and arranges the execution order, execution start time, execution duration, and concurrency / mutual exclusion relationship of each repair action in the time dimension based on the pipeline topology relationship constructed by spatial grid mapping and partition identification matrix, combined with the partition control vector, partition sensitivity matrix, local coupling index, and dynamic constraints reflected by the slurry flow velocity location. The backflow scheme refers to the backflow control plan data that can be directly sent to the execution layer after path search and timing constraint scheduling in a given backflow pipeline topology based on spatial grid mapping and partition identification matrix for abnormal grid points corresponding to backflow-type actions in the repair action set. The candidate action list refers to the set of actions obtained by mapping, merging, deduplicating, and sorting the partition control vector and the executable actions in the backflow scheme according to a unified data structure, and is used for subsequent feasibility and conflict verification. Feasibility and conflict verification refers to the data-driven verification of candidate actions in three aspects: physical executability, timing consistency, and rationality of cross-zone impact, based on the slurry velocity field and the zonal sensitivity matrix, and combined with the actuator mapping relationship corresponding to the zonal control vector, in order to screen out actions that are not executable or may cause control conflicts.

[0036] The above content will be described in detail below: For each anomalous grid point, the partition control vector is used as the decision variable, and the partition sensitivity matrix and local coupling index are used as constraints to construct a linear programming function and solve it to generate the control solution vector for that anomalous grid point. The specific calculation formula is as follows: ; In the formula, Represents grid points Partition control vector, Represents grid points Abnormal deviation amount, This represents the coupling penalty coefficient. Represents grid points The transpose of the local coupling index; all the above data have been normalized during the calculation. The control solution vectors are mapped to repair action entries according to predefined action encoding rules. The repair action entries for all abnormal grid points are then integrated to obtain the repair action set. The predefined action coding rules are invoked to parse the control solution vector, converting it into a repair action entry that includes at least the actuator type, actuator identifier, action type, action amplitude, and action timing parameters. The predefined action coding rules predefine the one-to-one correspondence between each component of the control solution vector and the action type, including but not limited to the coding methods for stirring power adjustment, micro-pump flow forward or reverse adjustment, reflux valve opening or closing, and action duration. After generating a single repair action entry, the repair action entry is further bound to the grid identifier and the partition identifier of the corresponding abnormal grid point. The above process is repeated for all abnormal grid points to obtain multiple repair action entries corresponding to different abnormal grid points. Then, the repair action entries are summarized and processed to generate a repair action set. The spatial grid mapping and the partition identifier matrix are combined with the grid point index as the common key to form a correspondence between the pipeline topology nodes and the partition number. Then, the repair action records in the repair action set are read one by one, and the repair action records with the action type field as backflow or containing the backflow valve actuator ID are filtered to form a backflow candidate list. Using pipeline topology nodes as vertices, perform connectivity analysis between adjacent grid points to generate a pipeline topology adjacency matrix: Read the spatial grid mapping to determine the spatial coordinates of each grid point and its six-neighbor or twenty-six-neighbor adjacency relationship, and read the slurry velocity vector of each grid point at the current time from the slurry velocity field; A candidate node set is constructed using the pipeline topology nodes as the vertex set. For any pair of adjacent grid points (node ​​m and node n) in the candidate node set, a connectivity determination is performed. The data processing methods for connectivity determination include: calculating the relative displacement direction vector from node m to node n using the grid point coordinates, and calculating the projection value of the slurry velocity vector in this direction. Only when both the projection value from node m to node n and the projection value from node n to node m are greater than a preset connectivity threshold, it is determined that there is a bidirectional effective flow connection between the two nodes. Alternatively, only when the projection value from node m to node n is greater than the preset connectivity threshold, it is considered that there is a directed connection from node m to node n. After completing the connectivity determination, a pipeline topology adjacency matrix is ​​generated. The elements of the pipeline topology adjacency matrix represent the connectivity determination results. When there is a bidirectional effective flow connection between two nodes, the corresponding element of the two nodes is set to 1. If there is a directed connection between two nodes, the corresponding element of the two nodes is set to 0.5. In other cases, the corresponding element of the two nodes is set to 0. Map the reflow candidate list to start node and target node pairs: Based on the spatial grid mapping, the grid index of each return candidate item is used for position lookup. The grid index is mapped to the node number in the three-dimensional computational grid as the starting node, and the return valve actuator ID is mapped to its inlet node number in the pipeline topology as the target node, thus forming a one-to-one pair of starting nodes and target nodes corresponding to each return candidate item. Using the pipeline topology adjacency matrix as path constraints, a time-windowed path search is performed on the starting and target node pairs to generate a backflow scheme: Read the corresponding executable time windows for the starting node and the target node respectively. The executable time windows are generated based on the executor load status and safety rules. The system reads the load status data of each actuator, verifies and classifies the load status according to preset safety rules, and then calculates the corresponding allowable start-up time for each actuator. Allowed duration With necessary safety distance This forms an executable time window with start and end boundaries, and its specific calculation formula is as follows: ; ; ; In the formula, Indicates the current time. Indicates actuator The end time of the last executed action, Indicates actuator Minimum safe recovery time, This indicates taking the maximum value. Indicates actuator Maximum continuous operating time under no-load or safe conditions Indicates actuator Current load status data, Indicates actuator Maximum allowed load, Indicates actuator The basic safety interval, This represents the safety amplification factor. All the above data have been normalized during the calculation. Based on this, a restricted search algorithm with time window constraints is performed on the pipeline topology adjacency matrix, including: reading the set of reachable neighboring nodes corresponding to the current node in the pipeline topology adjacency matrix one by one, determining whether the estimated travel time from the current node to the neighboring node falls within the executable time window of the neighboring node, and adding the neighboring node to the candidate expansion queue only when the executable time window constraint is satisfied. A heuristic search algorithm is used to search and compare all feasible paths in the candidate expansion queue, and finally select the path sequence with the minimum comprehensive cost corresponding to the current node. The path sequence is represented in the form of node sequence and edge sequence, and further mapped to the corresponding return valve opening sequence, pipeline section passage sequence and the entry and exit time points of each node, thereby generating a structured return scheme. The partition control vectors and reflow schemes are merged and rearranged in chronological order to generate a candidate action list: The partition control vector and backflow scheme are globally sorted based on the start timestamp. During the sorting process, multiple actions corresponding to the same actuator ID are conflict detected. If time overlap or insufficient safety interval is detected, the conflicting actions are postponed or inserted for adjustment. The adjustment process involves incrementing the start time field of the conflicting actions and rewriting it into the timeline. After sorting and conflict resolution are completed, actions that are consecutive on the timeline and have the same actuator ID are merged. The merging process involves performing weighted summation or taking the maximum value operation on the action quantity fields of adjacent actions and merging their time intervals to finally form a candidate action list. Using the slurry velocity field as input, the velocity change is calculated for each candidate action in the candidate action list. The specific calculation formula is as follows: ; In the formula, Indicates candidate actions The calculation results of the flow velocity change Indicates the candidate action The corresponding set of affected grid points, Represents the grid points in the affected grid point set The corresponding value in the slurry flow velocity field Represents grid points With candidate actions The spatial distance between the corresponding actuators Indicates candidate actions The action vector, The matrix represents the effect of the action on the kernel function. All the above data have been normalized during the calculation. The flow velocity change calculation results are weighted and corrected using multiplication based on the partition sensitivity matrix to generate a conflict relationship matrix: The partition identifier matrix is ​​read to map the velocity change calculation results of each grid point corresponding to the candidate action to the corresponding partition, and candidate conflict pairs are constructed based on the actuator's coverage relationship with the grid: When the coverage mesh sets of two actuators x and y intersect, (x,y) is marked as a candidate conflict pair. For each candidate conflict pair (x,y), the flow rate change calculation results are summarized in the coverage mesh set of the actuators and weighted by multiplication according to the elements of the corresponding partition in the partition sensitivity matrix to obtain the weighted conflict intensity. Finally, using the actuator as the row and column index, the weighted conflict intensity of all candidate conflict pairs is filled into the corresponding position in the matrix to form a conflict relationship matrix. The elements in the conflict relationship matrix represent the potential concurrent conflict degree of actuators x and y in the current time window. The candidate action list is adjusted based on the conflict relationship matrix to generate a control plan: If an element in the conflict relationship matrix is ​​greater than a preset conflict threshold, then two candidate actions corresponding to the corresponding element are retrieved, and conflict resolution decisions are made for the two candidate actions, including but not limited to delaying the execution time of the candidate actions to form a time-series stagger, or proportionally attenuating the action amount of the candidate actions corresponding to the low-sensitivity partitions according to the partition sensitivity matrix, and generating the final control scheme after completing the conflict traversal and adjustment of all candidate actions.

[0037] This scheme performs mathematical modeling and constraint optimization of the partition control vector, partition sensitivity matrix, and local coupling index on anomaly grid points. It can output a set of repair actions that match the segregation according to the grid, avoiding ineffective or over-adjustment. Then, based on the spatial grid mapping and partition identification matrix, it performs pipeline topology modeling and timing scheduling of the repair action set to ensure that the return path and opening sequence meet the pipeline and flow direction constraints and reduce the return dead zone. Finally, it constructs a candidate action list using the partition control vector and return scheme, and verifies the feasibility and conflicts by combining the slurry velocity field and partition sensitivity matrix, and outputs the control scheme within the actuator safety boundary.

[0038] The above describes the process of transforming control parameters and filtering backflow paths for the partition control vector and partition sensitivity matrix to generate a control scheme. The following describes how to solve the partition control vector, partition sensitivity matrix, and local coupling index point-by-point anomaly using mathematical modeling and real-time constraint optimization to generate a set of repair actions, specifically including: For each anomalous grid point, the partition control vector is used as the decision variable, and the partition sensitivity matrix and local coupling index are used as constraints to construct a linear programming function and solve it to generate the control solution vector for that anomalous grid point. The control solution vector is mapped to repair action entries according to a predefined action encoding rule. The repair action entries of all abnormal grid points are integrated to obtain the repair action set.

[0039] Here, the control solution vector refers to the set of control variable values ​​that minimize the intervention cost; The predefined action coding rule refers to the data processing process of converting the control solution vector obtained for each abnormal grid point into a structured repair action item according to a unified set of fields and coding mapping function; A repair action item refers to a structured control data unit that can be directly parsed and executed by the execution layer after the system completes the linear programming solution for a single abnormal grid point. It is used to describe the minimum intervention repair control behavior corresponding to the abnormal grid point.

[0040] This part has already been described in detail above, so I will not repeat it here.

[0041] This scheme constructs a linear programming function by using the partition control vector as the decision variable and the partition sensitivity matrix and local coupling index as constraints for each abnormal grid point, and then solves it to generate the control solution vector for that abnormal grid point. This achieves the effect of refining, quantifying, and optimizing the repair parameters of abnormal grid points at the data level, thereby reducing the disturbance of ineffective or over-repair to neighboring grid points and the overall homogenization state, and improving the pertinence and stability of abnormal repair. On this basis, the control solution vector is mapped to repair action items according to predefined action coding rules, and the repair action items of all abnormal grid points are integrated to form a repair action set. This realizes the processing effect of transforming the abstract numerical optimization results into structured control instructions that correspond one-to-one with specific actuators and can be directly called by the control system. This allows the repair strategies of multiple abnormal grid points to be uniformly arranged and conflict resolved at the action level, thereby ensuring the coordination of repair actions in terms of time sequence, spatial scope, and execution resource consumption, and improving the feasibility of overall repair execution and real-time control efficiency.

[0042] The above describes how, for each anomaly grid point, the partition control vector, partition sensitivity matrix, and local coupling index are solved through mathematical modeling and real-time constraint optimization to generate a set of repair actions. The following describes how, based on the spatial grid mapping and partition identifier matrix, the repair action set is used to generate a recirculation scheme through pipeline topology modeling and constraint timing scheduling, specifically including: The spatial grid mapping and the partition identifier matrix are combined to form a correspondence between pipeline topology nodes and partition numbers, and loop actions are extracted from the repair action set to generate a backflow candidate list. Connectivity analysis is performed between adjacent grid points using pipeline topology nodes as vertices to generate a pipeline topology adjacency matrix. The backflow candidate list is then mapped to a pair of starting and target nodes. Finally, using the pipeline topology adjacency matrix as path constraints, a path search with a time window is performed on the starting and target node pairs to generate a backflow scheme.

[0043] Among them, the backflow candidate list refers to the set of all candidate backflow action entries that are related to backflow and meet the routable conditions, obtained by loop action extraction and filtering from a given set of repair actions. Connectivity analysis refers to the process of calculating the reachability between pipeline topology nodes based on the obtained spatial grid mapping and the set of pipeline topology nodes derived from it, using the physical adjacency relationship of the grid and the actual connectivity relationship of the pipeline as the criteria, and identifying the set of nodes that can reach each other under a given connectivity rule as the same connected domain. The pipeline topology adjacency matrix refers to structured matrix data used to characterize the physical connectivity relationships and constraint attributes between pipeline nodes in a homogenization system. In pipeline topology modeling, a starting node and a target node pair are ordered nodes used to represent the starting position of the backflow action and the target position of the backflow action.

[0044] This part has already been described in detail above, so I will not repeat it here.

[0045] This solution extracts loop actions from the repair action set based on spatial grid mapping and partition identifier matrix, generating a backflow candidate list. This establishes a precise one-to-one mapping between each repair action and its corresponding physical pipeline node location and functional partition. This eliminates the reliance on manual experience in pipeline segment selection, which can lead to positioning errors and false triggering across partitions, a problem common in traditional solutions. Furthermore, using pipeline topology nodes as vertices, it performs connectivity analysis between adjacent grid points to generate a pipeline topology adjacency matrix. The backflow candidate list is then mapped to start and target node pairs, and further, the pipeline topology... The adjacency matrix serves as a path constraint. For the starting node and target node pair, a path search with a time window is performed to generate a return flow scheme. This ensures that the selection of the return flow path not only satisfies physical connectivity constraints but also incorporates scheduling constraints related to time windows and resource occupancy within intervals. It can automatically select an optimized return flow route from multiple candidate return flow paths that satisfies reasonable flow direction, avoids high-load areas, and avoids flow conflicts or frequent valve opening and closing within the same time window. This significantly improves the stability and energy efficiency of the return flow execution, reduces the risk of secondary disturbance to the homogenized flow field, and improves the speed and consistency of overall slurry concentration recovery.

[0046] The above describes how, based on spatial grid mapping and partition identification matrices, a set of repair actions is used to generate a recirculation scheme through pipeline topology modeling and constraint timing scheduling. The following describes how a candidate action list is constructed based on partition control vectors and the recirculation scheme, and how the feasibility and conflict of this candidate action list are verified based on the slurry velocity field and partition sensitivity matrix to generate a control scheme. Specifically, this includes: The partition control vectors and backflow schemes are merged and rearranged in chronological order to generate a candidate action list; Using the slurry velocity field as input, the velocity change is calculated for each candidate action in the candidate action list, and the velocity change calculation results are weighted and corrected according to the partition sensitivity matrix to generate a conflict relationship matrix. The candidate action list is adjusted based on the conflict relationship matrix to generate a control plan.

[0047] Among them, the velocity change calculation result refers to the predictive data result obtained by simulating and deriving the local and adjacent slurry velocity changes that the candidate action may cause based on the slurry velocity field at the current moment. The conflict relationship matrix is ​​a data structure used to characterize the degree of mutual influence between any two candidate actions in the candidate action list on the slurry flow state within the same or adjacent time windows.

[0048] This part has already been described in detail above, so I will not repeat it here.

[0049] This scheme generates a candidate action list by merging and rearranging the zoned control vectors and return schemes in chronological order. This structures the control commands, originally scattered across different zones and control logics, into serialized data on a unified time axis, providing a complete data foundation for subsequent conflict detection and optimization. Furthermore, using the slurry velocity field as input, velocity change calculations are performed for each candidate action in the candidate action list. The velocity change calculation results are then weighted and corrected according to the zone sensitivity matrix. This allows the impact of each candidate action on the local and overall flow fields under different zones and operating conditions to be quantified and prioritized. This combines physical-level velocity disturbances with the zoned response capabilities at the control level, enabling early identification of potential velocity abrupt changes and localized disturbances. By eliminating high-risk combinations of actions that involve stagnation or excessive shearing, the homogeneity of the slurry is reduced, and the adaptability of the control strategy to complex flow states is improved. Furthermore, the candidate action list is adjusted based on the conflict relationship matrix to generate a control scheme. The mutual constraints between candidate actions in the time and spatial partition dimensions are made explicit, and the execution order, concurrent combination, and start and stop times of the actions are optimized accordingly. This enables automatic avoidance and rearrangement of actions that involve resource competition, flow field interference, or conflicting control objectives. This significantly reduces execution conflicts between different control actions and the negative superposition effects on the flow field, improves the execution stability and effectiveness after the control command is issued, and enables the slurry velocity field to maintain flow continuity and controllable energy consumption while achieving target viscosity control and concentration homogenization.

[0050] Example 2: Please see Figure 5 A slurry sensing and adaptive control system for the homogenization process of lithium-ion batteries, including: The data acquisition module is used to acquire the electrolyte concentration matrix and slurry flow velocity field of the target object; The differential processing module is used to perform differential fusion of the electrolyte concentration matrix and the slurry velocity field to generate a local coupling index. The prediction module is used to perform weighted autoregression calculations on the electrolyte concentration matrix, and correct the weighted autoregression calculation results based on the local coupling index and slurry velocity field to generate electrolyte concentration prediction results. The partition analysis module is used to calculate the difference between the electrolyte concentration prediction result and the preset concentration threshold, and generate a risk heat map according to the ratio of the difference to the preset concentration threshold. Then, based on the risk heat map, the electrolyte concentration prediction result, local coupling index and flow velocity modulus are piecewise weighted and summed to generate a partition sensitivity matrix. The regulation scheme generation module is used to construct and solve constraint functions with the objective of minimizing concentration variance from the zonal sensitivity matrix, electrolyte concentration matrix, and flow velocity modulus, generate zonal control vectors, and perform control parameter transformation and reflux path screening on the zonal control vectors and zonal sensitivity matrix to generate regulation schemes.

[0051] This embodiment has the same technical effects as Embodiment 1.

[0052] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The data mentioned in this application have undergone normalization and other preprocessing to unify dimensions during formula calculations.

[0053] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for slurry sensing and adaptive control in the homogenization process of lithium-ion batteries, characterized in that, Includes the following steps: Obtain the electrolyte concentration matrix and slurry flow velocity field of the target object; The electrolyte concentration matrix and the slurry velocity field are differentially fused to generate a local coupling index; A weighted autoregression calculation is performed on the electrolyte concentration matrix, and the weighted autoregression calculation results are corrected according to the local coupling index and the slurry flow velocity field to generate electrolyte concentration prediction results; The difference between the predicted electrolyte concentration and the preset concentration threshold is calculated, and a risk heat map is generated according to the ratio of the difference to the preset concentration threshold. Then, based on the risk heat map, the predicted electrolyte concentration, the local coupling index, and the flow velocity modulus are summed in segments to generate a partition sensitivity matrix. The velocity modulus is obtained by calculating the vector modulus of the slurry velocity field. A constraint function with the objective of minimizing the concentration variance is constructed and solved for the partition sensitivity matrix, the electrolyte concentration matrix and the flow velocity modulus to generate a partition control vector. Then, control parameters are transformed and reflux paths are screened for the partition control vector and the partition sensitivity matrix to generate a control scheme.

2. The slurry sensing and adaptive control method for the lithium-ion battery homogenization process according to claim 1, characterized in that: Obtaining the electrolyte concentration matrix and slurry flow field of the target object specifically includes: The electrolyte concentration data and slurry flow rate data of the target object are obtained and then meshed using grid points as the basic unit. The electrolyte concentration data and the slurry flow rate data are subjected to coordinate interpolation transformation to generate a spatial grid mapping; Based on the spatial grid mapping, spatiotemporal interpolation is performed on the electrolyte concentration data and the slurry flow rate data to generate an electrolyte concentration matrix and a slurry flow rate field.

3. The slurry sensing and adaptive control method for the lithium-ion battery homogenization process according to claim 2, characterized in that: The electrolyte concentration matrix is ​​subjected to weighted autoregression calculation, and the weighted autoregression calculation results are corrected according to the local coupling index and the slurry velocity field to generate electrolyte concentration prediction results, specifically including: For each grid point, extract the electrolyte element set of the most recent A time points according to the time axis of the electrolyte concentration matrix, and perform weighted autoregression calculation on the electrolyte element set; The flow field diffusion calculation is performed on the slurry velocity field and concentration gradient field, and the weighted autoregression calculation results are corrected based on the local coupling index and the flow field diffusion calculation results to generate the electrolyte concentration prediction results for each grid point. The concentration gradient field is obtained by performing a central difference operation on the electrolyte concentration matrix within the grid neighborhood.

4. The slurry sensing and adaptive control method for the lithium-ion battery homogenization process according to claim 3, characterized in that: Risk heatmaps are generated based on the ratio of the difference to a preset concentration threshold. Then, based on these risk heatmaps, the electrolyte concentration prediction results, the local coupling index, and the flow velocity modulus are summed in segments with weights to generate a regional sensitivity matrix. Specifically, this includes: Calculate the ratio of the difference to a preset concentration threshold, and perform statistical mapping on the ratio to generate a risk heatmap; The preset concentration threshold is obtained by performing distribution statistics on the electrolyte element set and then weighting and normalizing the distribution statistics results in combination with the slurry flow velocity field. The risk heatmap and the local coupling index are weighted and fused and clustered within the grid neighborhood to generate a partition identification matrix. Based on the partition identification matrix, the electrolyte concentration prediction result, the local coupling index and the flow velocity modulus are piecewise weighted and summed in each partition according to the risk heatmap to generate a partition sensitivity matrix.

5. The slurry sensing and adaptive control method for the lithium-ion battery homogenization process according to claim 4, characterized in that: After generating the partition control vector, the process also includes generating abnormal grid points, specifically including: The difference between the electrolyte concentration matrix and the electrolyte concentration prediction result is calculated for each grid point. It is then determined whether the difference calculation result is greater than a preset discrimination threshold. If so, the grid point corresponding to the difference calculation result is determined to be an abnormal grid point. The preset discrimination threshold is dynamically set based on the risk information of the grid points within the risk heatmap.

6. The slurry sensing and adaptive control method for the lithium-ion battery homogenization process according to claim 5, characterized in that: The control parameter transformation and backflow path screening of the partition control vector and the partition sensitivity matrix to generate the control scheme specifically include: For each abnormal grid point, the partition control vector, the partition sensitivity matrix, and the local coupling index are solved through mathematical modeling and real-time constraint optimization to generate a set of repair actions. Based on the spatial grid mapping and the partition identifier matrix, a recirculation scheme is generated for the repair action set through pipeline topology modeling and constraint timing scheduling; A candidate action list is constructed based on the partition control vector and the recirculation scheme. The feasibility and conflict of the candidate action list are verified based on the slurry velocity field and the partition sensitivity matrix to generate a control scheme.

7. The slurry sensing and adaptive control method for the lithium-ion battery homogenization process according to claim 6, characterized in that: The specific steps for generating a set of repair actions include solving the partition control vector, the partition sensitivity matrix, and the local coupling index for each anomaly grid point through mathematical modeling and real-time constraint optimization: For each abnormal grid point, the partition control vector is used as the decision variable, and the partition sensitivity matrix and the local coupling index are used as constraints to construct a linear programming function and solve it to generate the control solution vector of the abnormal grid point. The control solution vector is mapped to repair action entries according to a predefined action encoding rule. The repair action entries of all abnormal grid points are integrated to obtain a repair action set.

8. The slurry sensing and adaptive control method for the lithium-ion battery homogenization process according to claim 6, characterized in that: Based on the spatial grid mapping and the partition identifier matrix, the reflow scheme is generated for the repair action set through pipeline topology modeling and constraint timing scheduling, specifically including: The spatial grid mapping is combined with the partition identifier matrix to form a correspondence between pipeline topology nodes and partition numbers, and loop actions are extracted from the repair action set to generate a backflow candidate list. Connectivity analysis is performed between adjacent grid points using pipeline topology nodes as vertices to generate a pipeline topology adjacency matrix. The backflow candidate list is then mapped to a pair of starting nodes and target nodes. Using the pipeline topology adjacency matrix as path constraints, a path search with a time window is performed on the pair of starting nodes and target nodes to generate a backflow scheme.

9. The slurry sensing and adaptive control method for the homogenization process of lithium-ion batteries according to claim 6, characterized in that: A candidate action list is constructed based on the partition control vector and the recirculation scheme. The feasibility and conflict checks of the candidate action list are then performed based on the slurry velocity field and the partition sensitivity matrix to generate a control scheme, specifically including: The partition control vector and the backflow scheme are merged and rearranged in chronological order to generate a candidate action list; Using the slurry velocity field as input, the velocity change is calculated for each candidate action in the candidate action list, and the velocity change calculation results are weighted and corrected according to the partition sensitivity matrix to generate a conflict relationship matrix. The candidate action list is adjusted based on the conflict relationship matrix to generate a control scheme.

10. A slurry sensing and adaptive control system for the homogenization process of lithium-ion batteries, characterized in that, include: The data acquisition module is used to acquire the electrolyte concentration matrix and slurry flow velocity field of the target object; The differential processing module is used to perform differential fusion of the electrolyte concentration matrix and the slurry flow velocity field to generate a local coupling index; The prediction module is used to perform weighted autoregression calculation on the electrolyte concentration matrix, and correct the weighted autoregression calculation results according to the local coupling index and the slurry flow velocity field to generate electrolyte concentration prediction results. The partition analysis module is used to calculate the difference between the electrolyte concentration prediction result and the preset concentration threshold, and generate a risk heat map according to the ratio of the difference to the preset concentration threshold. Then, based on the risk heat map, the electrolyte concentration prediction result, the local coupling index and the flow velocity modulus are segmented and weighted to generate a partition sensitivity matrix. The regulation scheme generation module is used to construct and solve a constraint function with the objective of minimizing the concentration variance for the partition sensitivity matrix, the electrolyte concentration matrix, and the flow velocity modulus, generate a partition control vector, and perform control parameter transformation and reflux path screening on the partition control vector and the partition sensitivity matrix to generate a regulation scheme.

Images