A new energy vehicle lithium battery state prediction method and system

Abstract

The present application relates to the technical field of state prediction, and particularly relates to a new energy vehicle lithium battery state prediction method and system. The method comprises the following steps: firstly, acquiring vehicle driving climbing working conditions and heavy load information, and analyzing corresponding lithium battery discharge state to extract high-power discharge timing change characteristics; based on the working condition information, deeply analyzing the incremental correlation of the high-power discharge timing to obtain discharge power incremental data under the heavy load climbing working condition, and combining the data to evaluate the power consumption estimation deviation degree of the lithium battery; according to the estimation deviation data, the residual capacity state of the lithium battery is predicted to generate accurate residual capacity prediction data. The present application optimizes the lithium battery state prediction technology to make the lithium battery state prediction technology more accurate.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of state prediction, in particular to a new energy vehicle lithium battery state prediction method and system. BACKGROUND

[0002] As the core energy storage component of new energy vehicles, the performance of lithium batteries directly affects the vehicle's endurance, safety and service life. Therefore, accurately predicting the remaining capacity, health status and other key performance indicators of lithium batteries is crucial for improving the energy efficiency of new energy vehicles, extending the service life of batteries, and ensuring driving safety. The state of lithium batteries, especially under high-power discharge conditions such as climbing and heavy loading, the discharge characteristics of the battery are more complex, and the change of discharge power is affected by the interaction of multiple factors. However, the traditional new energy vehicle lithium battery state prediction method cannot accurately couple the power consumption deviation state under climbing and heavy loading conditions, resulting in low accuracy of lithium battery remaining capacity prediction. SUMMARY

[0003] Therefore, it is necessary to provide a new energy vehicle lithium battery state prediction method and system to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a new energy vehicle lithium battery state prediction method, the method comprising the following steps:

[0005] Step S1: Obtain the climbing and heavy loading information of the new energy vehicle, and obtain the corresponding lithium battery discharge state; perform high-power discharge characteristic analysis on the lithium battery discharge state to output high-power discharge timing change characteristics;

[0006] Step S2: According to the climbing and heavy loading information, the high-power discharge timing change characteristics are analyzed to obtain the discharge power increment correlation data of the heavy loading and climbing working condition; according to the discharge power increment correlation data, the lithium battery power consumption estimation deviation degree coupling is carried out, and the power consumption estimation deviation degree coupling data is obtained;

[0007] Step S3: According to the power consumption estimation deviation degree coupling data, the state of the remaining capacity of the lithium battery is predicted to obtain the remaining capacity state prediction data.

[0008] The present application also provides a new energy vehicle lithium battery state prediction system for executing the new energy vehicle lithium battery state prediction method as described above, which comprises:

[0009] The feature analysis module is configured to obtain the driving uphill working condition and heavy load information of the new energy vehicle, and obtain the corresponding lithium battery discharge state; the lithium battery discharge state is subjected to high-power discharge feature analysis to output the high-power discharge timing change feature;

[0010] The power consumption estimation deviation analysis module is configured to perform discharge power increment correlation analysis of the high-power discharge timing change feature according to the driving uphill working condition and heavy load information, to obtain discharge power increment correlation data of the heavy load uphill working condition; and perform lithium battery power consumption estimation deviation degree coupling according to the discharge power increment correlation data, to obtain power consumption estimation deviation degree coupling data.

[0011] The state prediction module is configured to perform lithium battery residual capacity state prediction according to the power consumption estimation deviation degree coupling data, to obtain residual capacity state prediction data.

[0012] The present application has the advantages that by obtaining the uphill working condition and heavy load information of the new energy vehicle during driving, and combining the actual discharge state of the battery, detailed analysis is performed. By monitoring the battery discharge characteristics under high power demand, the change feature of the high-power discharge timing can be extracted, which lays a foundation for subsequent power increment correlation analysis. Through analysis of these discharge characteristics, the performance of the battery under different working conditions can be accurately identified, providing a scientific basis for accurate prediction of battery performance. By combining the uphill working condition and heavy load information, the high-power discharge characteristics of the battery are analyzed in depth, and the discharge power increment correlation data is determined. This analysis reveals the power change law of the battery under high load working conditions, which can more accurately evaluate the power consumption of the battery under specific working conditions. At the same time, by coupling the data with the power consumption estimation deviation of the battery, the accuracy of the battery discharge process is further improved, providing a more accurate estimation basis for subsequent residual capacity prediction. Through comprehensive analysis of the data, the residual capacity state of the battery can be accurately predicted under complex driving conditions. This prediction not only helps to improve the efficiency of the battery, but also provides real-time power information for the driver, avoiding safety hazards caused by insufficient power. Through accurate residual capacity prediction, the intelligent level of the battery management system (BMS) is improved, thereby prolonging the service life of the battery and optimizing the performance of the vehicle. Therefore, the present application is an improvement on the traditional new energy vehicle lithium battery state prediction method, which solves the problem of low accuracy of residual capacity prediction of the lithium battery under the conditions of uphill and heavy load in the traditional new energy vehicle lithium battery state prediction method, improves the coupling degree of the power consumption deviation state, and improves the accuracy of the residual capacity prediction of the lithium battery. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1It is a new energy vehicle lithium battery state prediction method step flow diagram;

[0014] Figure 2 It is Figure 1 The detailed implementation step flow diagram of step S2 in the embodiment. DETAILED DESCRIPTION

[0015] Please refer to Figure 1 A new energy vehicle lithium battery state prediction method, the method comprises the following steps:

[0016] Step S1: obtaining the driving climbing working condition and heavy load information of the new energy vehicle, and obtaining the corresponding lithium battery discharge state; performing high-power discharge characteristic analysis on the lithium battery discharge state to output high-power discharge time sequence change characteristics;

[0017] In the embodiment of the application, the state monitoring device in the new energy vehicle power system performs real-time data acquisition on the vehicle during driving, including data such as longitudinal slope of the vehicle in different road sections, total mass of the vehicle, front and rear axle load ratio, wheel end torque, vehicle speed, driving motor output torque, acceleration and environmental temperature, etc. The data of the vehicle in the scene where the slope is greater than 8% and the load reaches the designed maximum load is screened to form a heavy load climbing working condition sample set. The current, voltage, instantaneous power and temperature rise signals recorded in the battery management system are read at every 0.05 seconds as the time step, the collected high-power discharge data is uniformly time-synchronized and filtered, and the signal peak and random noise are eliminated by using the sliding average and second-order median filtering joint algorithm. Then, a power change array based on time dimension is constructed, the continuous high-power interval is taken as the segmentation node, the instantaneous power increment of the power rising section and the power falling section is calculated respectively, and the power transition characteristic value is extracted by using the discrete difference calculation method. The time, power, current and internal resistance change are normalized as one-dimensional sequence, the discharge amplitude change rate curve is derived by time step statistics, and the power slope extreme section is derived by using the gradient ascent method. Finally, the high-power discharge time sequence change characteristic data is output, including the peak time distribution of the discharge power, the continuous interval length, the power transition gradient and the internal resistance response delay time sequence, which provides the time sequence characteristic basis for the subsequent steps.

[0018] Step S2: according to the driving climbing working condition and heavy load information, the high-power discharge time sequence change characteristics are analyzed in association with the discharge power increment of the heavy load climbing working condition, and the discharge power increment association data of the heavy load climbing working condition is obtained; according to the discharge power increment association data, the consumption estimation deviation degree coupling is carried out, and the consumption estimation deviation degree coupling data is obtained;

[0019] In the embodiment of the present application, after obtaining the high-power discharge time sequence variation characteristic data, the power output parameters of the new energy vehicle under different load and slope conditions are taken as input, and a power variation and force correlation matrix is constructed based on the real-time running data of the vehicle on the climbing road section. First, the values of traction force, rolling resistance, air resistance and slope resistance are extracted through the vehicle longitudinal dynamics equation, the demand change of the vehicle for motor output power in the dynamic climbing process is calculated from the vector superposition result of these component forces, and the discharge power variation speed is calculated through time sequence differential calculation. Then, the power incremental rate curve is calculated by using the increment ratio of first-order derivative and second-order derivative, and the nonlinear response function based on energy transmission efficiency is established according to the evolution result of power incremental rate with time, and the discharge power incremental correlation data is formed after the response function is aligned on the time axis. Next, the discrete clustering algorithm based on the power variation rate distribution characteristics is used to identify the concentrated section of the power cross-interval increment, and the progressive ratio of discharge power in different time periods and the dynamic adjustment trend of internal resistance are calculated, and these change trends are converted into data sequences corresponding to energy output and load response. By comparing the deviation between the theoretical discharge energy consumption and the measured energy output, the energy attenuation bias formed in different power incremental stages is calculated. According to these bias data, the fitting algorithm is used to generate the deviation coupling data set to reflect the degree of deviation between the energy output and the energy consumption estimation of the lithium battery under strong load conditions. The final discharge estimation deviation degree coupling data contains time variation rate, power increment rate, internal resistance dynamic deviation parameters and discharge energy deviation curve.

[0020] Step S3: predicting the residual capacity state of the lithium battery according to the discharge estimation deviation degree coupling data to obtain residual capacity state prediction data;

[0021] In the embodiment of the present application, after obtaining the consumption estimation deviation degree coupling data, first, the data is horizontally time-sliced, the deviation data under different road conditions is grouped according to the working conditions to establish a time trend matrix of the consumption change; the deviation growth rate of the consumption with time is extracted through difference analysis to form a consumption estimation trend feature, and then the feature is input into a reinforcement learning framework constructed based on a Q-learning algorithm, a discrete state space is set as a two-dimensional combination of the SOC interval and the power change rate interval, an action space is set as the SOC prediction output adjustment amplitude, the Q value table is updated in real time according to the change direction of the consumption deviation trend, and the value function iteration is performed on each state-action pair according to the Bellman update logic. Through multiple policy cycles, the optimal decision from the deviation trend to the remaining capacity estimation is formed in the prediction process, when the learning converges, the output consumption estimation deviation trend learning data is extracted, and the voltage response delay term and the capacity attenuation term in the traditional SOC estimation formula are corrected by using the data, so that the lithium battery remaining capacity state prediction data is calculated, and the prediction result contains the real-time remaining capacity time series distribution and the state change curve of the lithium battery under high-power discharge conditions such as heavy load climbing, which provides quantitative basis for subsequent intelligent energy management and charging scheduling.

[0022] Specifically, in the whole lithium battery state prediction process, the driving climbing working condition and the heavy load information are first used as an external load source to describe the discharge behavior of the battery under high stress conditions, so that the high-power transient characteristics in the discharge state can be completely extracted and form continuous time series change characteristics; then, by associating and analyzing these time series characteristics with the climbing load intensity, the incremental mode of the discharge power under the heavy load condition is obtained, and the deviation amount between the incremental mode and the actual consumption path of the battery is coupled, so that the estimation deviation caused by the power change is quantitatively characterized; finally, on the basis of the deviation characterization, the deviation amount is used as a correction factor to adjust the remaining capacity estimation, so that the prediction result can closely fit the real consumption behavior of the battery under the heavy load climbing condition, thereby realizing high-accuracy prediction of the remaining capacity of the lithium battery under complex working conditions.

[0023] Step S1 comprises the following steps:

[0024] Step S11: obtaining the background monitoring permission of the new energy vehicle lithium battery and the driving climbing working condition and the heavy load information of the new energy vehicle;

[0025] Step S12: obtaining the lithium battery discharge state corresponding to the driving climbing working condition and the heavy load information based on the background monitoring permission;

[0026] Step S13: performing high-power discharge characteristic analysis on the lithium battery discharge state to obtain high-power discharge characteristics;

[0027] Step S14: Perform time-series variation analysis on the high-power discharge characteristics to output the high-power discharge time-series variation characteristics.

[0028] In this embodiment of the invention, a communication link is established with the vehicle control unit through the data acquisition system of the new energy vehicle, and authorized access to vehicle driving data and battery management system data is granted based on controlled permissions. Vehicle driving records, including vehicle speed, torque, RPM, pedal opening, braking signal, ambient temperature, and road slope information, are obtained through authenticated background monitoring permissions, and the lithium battery pack information configured in the vehicle is indexed based on the vehicle identification number. Driving and climbing condition data are filtered based on the following criteria: road slope greater than 5%, vehicle longitudinal acceleration continuously greater than 0.5 m / s², and load exceeding 80% of the design load. Driving condition data and battery data are time-series aligned using timestamp synchronization to ensure that the vehicle's physical state and battery state are synchronized at each point in time. After acquiring driving and climbing condition and heavy load information, the above data is stored in a time-series format in a data processing structure, providing a complete basic dataset for subsequent discharge state extraction.

[0029] Based on background monitoring permissions, discharge-related physical quantity data, including individual battery voltage, overall battery voltage, output current, temperature, total power, and charge output rate, are retrieved from the battery management system. The battery data is synchronized with the driving condition data in step S11 using a time index field to ensure data consistency across sample points. A signal filtering method is used to correct abrupt changes and outliers in the original data, and a sliding window filter is used to smooth the voltage signal and remove transient interference. Subsequently, the instantaneous discharge state intensity is calculated based on the correspondence between vehicle power demand and battery output current, and the relationship between load and power output is compared to detect whether the battery is operating in a high-power range. When the instantaneous output power exceeds 70% of the rated power, it is marked as a high-power discharge segment. The entire operating dataset is divided into high-power and low-power regions using a segmentation algorithm, and the time-varying sequences of current, voltage, temperature, and power within the high-power segment are extracted to form lithium battery discharge state data.

[0030] After obtaining the state of discharge (SOD) of the lithium battery, high-power discharge characteristic analysis is performed on the SOD data. First, a power change array of the high-power discharge time series is constructed, with the array dimensions being time and power values. The power increment is calculated using a first-order difference algorithm on the power change array, and the continuous rate of change of power increase is calculated to form the basic data of discharge dynamic changes. Then, second-derivative progressive curvature analysis is used to quantitatively extract the acceleration of the power increase segment, thereby determining the characteristics of power peak formation. The trends of time, power, current, internal resistance, and temperature are fused to generate a feature field matrix. Principal component decomposition (PCD) is used to extract feature components with high correlation coefficients, and the main discharge factors are determined by sorting them by feature contribution rate. These processes output high-power discharge characteristic data, which includes the power peak distribution, the proportion of power increase segments, the current and power synchronization offset rate, and the temperature rise trend curve, used to characterize the dynamic energy release characteristics of the lithium battery under continuous high load.

[0031] After obtaining the high-power discharge characteristics, a time-series variation analysis was performed. First, the characteristic data was divided into continuous equal-length intervals according to time periods, and missing values ​​were filled in using a linear interpolation algorithm to ensure temporal integrity. Then, a time series model was established for the power peak and time interval data. The discharge power oscillation period was extracted using an autoregressive moving average method, and the recursive pattern between high-power segments was analyzed. The correlation between the synchronization rate of power and internal resistance fluctuations within each time window was calculated, obtaining the difference sequence between discharge response delay time and temperature rise delay time. Based on this, the temporal gradient of power change, i.e., the continuous increment of discharge rate change, was calculated, and a high-power discharge time-series variation curve was plotted. A clustering inflection point identification algorithm was used to mark key inflection points of power transitions and calculate the inflection point intervals, forming time-series variation characteristic data. Finally, the high-power discharge time-series variation characteristics were output, including peak interval distribution, power jump rate, internal resistance change delay parameters, and temperature rise-related delay sequence. This characteristic data lays the data foundation for subsequent discharge power analysis and energy estimation deviation coupling.

[0032] refer to Figure 2 The aforementioned step S2 includes the following steps:

[0033] Step S21: Perform step-jump intensity analysis on the timing variation characteristics of high-power discharge to obtain the timing discharge power intensity;

[0034] In this embodiment of the invention, the power time-series variation characteristics are discretized in chronological order, and the power difference between adjacent moments is taken as the power change. The average gradient of the power change over several consecutive moments is calculated using a sliding time window to characterize the instantaneous rate of power increase. Then, the differential mean and variance analysis method is used to detect abrupt changes in the power change sequence; when the gradient exceeds a set threshold, it is identified as a jump point. The power increase between adjacent jump points is statistically analyzed and mapped to power step intervals. Subsequently, step function superposition analysis is used to correlate the power increase values ​​of multiple step intervals with their durations, calculating the power increment rate matrix. By calculating the standard deviation and skewness of the power increase distribution, the concentrated areas of jump intensity are determined, and finally, time-series discharge power intensity data is generated based on the energy change rate of these concentrated intervals. This data records the power transition intensity, continuous increase, and energy release rate in the time dimension, serving as an important intermediate quantity for subsequent power increment correlation analysis.

[0035] Step S22: Based on the driving and climbing conditions and heavy load information, perform incremental correlation analysis of the discharge power intensity under heavy load climbing conditions to obtain incremental correlation data of discharge power under heavy load climbing conditions.

[0036] In this embodiment of the invention, after obtaining the time-series discharge power intensity, the power intensity is correlated and analyzed based on the vehicle's driving and climbing conditions and heavy load information. First, the slope angle and road resistance parameters are extracted from the driving condition data to calculate the additional traction resistance caused by the slope. From the heavy load information, the vehicle's total weight, center of gravity height, and drive wheel radius are extracted, and the additional work input requirement per unit time is calculated based on mechanical projection. These resistance parameters are converted into corresponding equivalent power requirements and matched with the time-series discharge power intensity according to the time coordinate to obtain a time-synchronized structure between slope load and power output. Next, nonlinear fitting is performed based on the power synchronization structure, and curve fitting is used to calculate the power output increment ratio curve with slope in the power increment interval, forming power increment correlation data. To ensure data continuity, a smoothing curve is calculated for the offset between power and load, and abnormal fluctuation points are corrected to stable trend values. Finally, the discharge power increment correlation data for heavy load climbing conditions is obtained, which includes the mapping relationship between different slope intervals, load proportions, and power increment rates.

[0037] Step S23: Based on the discharge power increment correlation data, deduce the change intensity of the circuit internal resistance;

[0038] In this embodiment of the invention, after obtaining the discharge power increase correlation data, the intensity of the line internal resistance mutation is derived based on this data. First, the instantaneous output voltage is calculated through the correspondence between power and current, and the voltage drop curves at each time step are recorded. Then, the ratio of the voltage difference to the current change rate is calculated to obtain the trend of line internal resistance change. To reduce the interference of high-frequency noise on mutation judgment, the internal resistance change sequence is low-pass filtered and a first-order difference is calculated to detect local mutation points. Then, statistical analysis is performed on the time interval and amplitude changes between mutation points, and mutation intensity segments are defined based on the concentrated interval of the sum of squares of mutation amplitudes. The weighted average of the internal resistance increase amplitude within each segment is calculated to form a continuous mutation intensity sequence. This sequence reflects the distribution and rate of change of internal resistance mutations in the battery at different power increase stages, constituting the data source for subsequent cluster fitting analysis.

[0039] Step S24: Perform nonlinear clustering fitting on the intensity of sudden changes in line resistance to obtain clustering fitting data for sudden changes in resistance.

[0040] In this embodiment of the invention, after obtaining the sequence of changes in line resistance intensity, nonlinear clustering fitting is performed on it. First, the sequence of changes in intensity is standardized to a uniform scale range, and the similarity between adjacent change points is calculated using a distance metric based on the squared difference between adjacent points. Hierarchical aggregation analysis is performed based on the similarity matrix, grouping data with high similarity in resistance changes into the same cluster unit. Then, a nonlinear growth function is fitted to the data point sequence within each cluster unit, and the fitting residuals are obtained. The cluster fitting convergence is determined based on the principle of minimizing the sum of squared residuals, and finally, the cluster fitting data for resistance changes is output. This data includes the average increase in resistance corresponding to the cluster center, the duration of the change, and the change trend curve, providing a quantitative basis for the thermal effect and attenuation characteristics in the energy consumption deviation assessment stage.

[0041] Step S25: Obtain the estimation logic of the remaining SOC of the lithium battery through the background monitoring authority of the new energy vehicle lithium battery;

[0042] In this embodiment of the invention, the estimation logic for SOC calculation in the battery management system is retrieved through the background monitoring permissions of the new energy vehicle. This logic takes real-time voltage, current, temperature, and capacity estimates as input, calculates the cumulative discharge amount through ampere-hour integration, and inversely derives the current SOC value based on the open-circuit voltage-SOC curve. Conventional techniques use the ampere-hour integration method as the algorithm for calculating battery SOC.

[0043] Step S26: Based on the discharge power increment correlation data and the internal resistance mutation cluster fitting data, the estimation logic is coupled with the degree of lithium battery power consumption estimation deviation to obtain power consumption estimation deviation coupling data.

[0044] In this embodiment of the invention, the peak sequence of power increments is extracted from the discharge power increment correlation data and used as the instantaneous discharge current load index. Based on this index, a three-dimensional lookup table (input: current load index, battery temperature, current SOC; output: polarization overpotential) calibrated beforehand through electrochemical experiments is consulted to derive the additional energy loss caused by concentration polarization and electrochemical polarization, which is quantified as polarization charge loss data. Next, a thermal effect diffusion gradient is constructed based on the centroid points in the internal resistance abrupt cluster fitting data. Specifically, the current battery temperature is compared with the temperature components of each centroid, and an inverse distance weighted interpolation method is used, combined with the internal resistance abrupt intensity component of the centroid, to calculate a gradient value characterizing the current thermal stress state of the battery. Then, the power consumption time-series divergence accumulation coupling is performed: the polarization charge loss data is multiplied by the thermal effect diffusion gradient to obtain an instantaneous deviation dissipation power sequence, and then the sequence is integrated over time to obtain an additional power consumption caused by polarization and thermal effects that accumulates over time, i.e., power consumption divergence accumulation data. Finally, based on the estimation logic obtained in step S25, the theoretical power consumption is calculated using the ampere-hour integration method. This theoretical power consumption is added to the cumulative power consumption divergence data to obtain a corrected total power consumption. The difference between the corrected total power consumption and the theoretical power consumption is used as the final output, which is the power consumption estimation deviation coupling data.

[0045] Specifically, starting with the time-series jump characteristics during high-power discharge, the actual discharge power performance of the battery under high load conditions is fully extracted by quantifying the jump amplitude and duration. Then, this discharge power intensity is correlated with load changes under ramp and heavy load conditions, forming continuous correlation data of the discharge power variation with increasing load. Based on this, by analyzing anomalous segments of power change in this increasing pattern, the intensity of line internal resistance mutations caused by increasing load is derived, further revealing the battery's internal load behavior. After obtaining the internal resistance mutation intensity, different mutation segments are classified and fitted using nonlinear clustering, forming structured data of the overall trend of internal resistance change. Subsequently, the SOC estimation logic of the battery's backend maintenance is obtained and used as the inherent basic calculation chain. Finally, using the discharge power increasing correlation data and internal resistance mutation clustering fitting data as correction basis, the SOC estimation logic is subjected to deviation coupling processing, enabling it to dynamically correct according to the actual consumption characteristics under different load conditions, forming deviation-coupled data that reflects the actual power consumption path under high load conditions.

[0046] Step S21 includes the following steps:

[0047] Step S211: Plot the high-power discharge time-series variation curve based on the high-power discharge time-series variation characteristics; mark the convex points in the high-power discharge time-series variation curve;

[0048] Step S212: Calculate the slope of the high-power discharge timing curve near the peak based on the convexity point to obtain the slope of the curve near the peak.

[0049] Step S213: Measure the variance of adjacent peak quantile widths of the high-power discharge timing curve based on the slope of the curve, so as to generate the variance of adjacent peak quantile widths.

[0050] Step S214: Based on the variance of the width of the adjacent peak quantile, perform peak jump increment order analysis on the high-power discharge time-series change curve to obtain the peak jump increment order;

[0051] Step S215: Perform step transition intensity analysis based on the peak transition increment order to obtain the time-series discharge power intensity.

[0052] In this embodiment of the invention, based on the obtained high-power discharge time-series variation characteristic data, a time-series array is constructed with time as the horizontal axis and discharge power as the vertical axis. The power values ​​at all sampling time points are sequentially arranged, and a high-power discharge time-series variation curve is plotted on the coordinate plane. To eliminate signal noise interference, a five-point smoothing-based moving average method is used to smooth the curve data. The first-order difference of power change between adjacent time points is calculated, and the sign of the change is determined. The turning points from positive to negative or vice versa are recorded. Local extreme points in the curve are detected based on the positive and negative changes of the second-order numerical difference of the curve, and points where the change changes from negative to positive are identified as convex points. Based on this, a local peak detection algorithm is used to filter out low-amplitude pseudo-peaks, retaining only obvious convex points with amplitudes exceeding twice the average power change amplitude. The marked convex points are superimposed on the original time-series variation curve in a dot matrix format. The convex point sequence completely describes the peak distribution of the power signal during high discharge, providing a location basis for subsequent slope calculation and jump intensity analysis.

[0053] After marking the convex points of the high-power discharge curve, multiple sets of sampling points adjacent to each convex point are extracted, using each convex point as a peak reference point. To accurately depict the peak change trend, the power and time values ​​within the interval of each convex point are standardized by class interval. The slope of the power change line segment before and after the peak is calculated using the least squares linear fitting method, and the slope approaching the peak is taken as the slope of the curve approaching the peak at that convex point. By averaging multiple slope samples within the peak interval, the average local rise rate and fall rate are obtained, thus characterizing the rise and decay rate of the power curve at the peak position. The slopes approaching the peak of all convex points are arranged in chronological order to form a slope sequence, which is used to represent the dynamic change rate distribution of the power signal over time, laying the foundation for the next step of peak width variance analysis.

[0054] Based on the obtained slope sequence of the near-peak curve, the variance of the adjacent peak quantile width is measured for the high-power discharge time-series variation curve. First, the time interval between every two adjacent peak points is determined, and the power value variation range within each interval is recorded; these time intervals are used as the adjacent peak quantile width. Then, the squared difference of each adjacent width is calculated, and the root mean square value is taken to measure the fluctuation degree of adjacent peak widths. A variance-stable sequence is generated through time-mean normalization, transforming asynchronous peak width variations into a standardized width variance index. Based on this index, the peak fluctuation characteristics are statistically analyzed, ultimately generating adjacent peak quantile width variance data. This variance data reflects the overall uneven distribution of peak intervals in the power time series and reveals the temporal expansion law of power energy changes, providing a quantitative basis for analyzing the order of peak jumps.

[0055] Based on the variance data of adjacent peak quantile widths, a peak jump increment order analysis was performed on the time-series variation curve of high-power discharge. First, the variance data was used as input, and the variance change was calculated by differentiating it according to time sequence, determining the direction and rate of increase of variance over time. Second, a progressive distribution map was established with the variance change rate as the x-axis and the time period as the y-axis to observe the fluctuation range. Clustering was used to stratify the increment of the variance change rate according to its order, defining two adjacent increments changing in the same direction as increment orders. Through time-series analysis, the number of adjacent variance increment orders within each time window was calculated, and these orders were accumulated to form a peak jump increment order sequence. This sequence reflects the number and frequency of power peak recursion in the time dimension, and is an important factor for measuring the rate of change of discharge power in subsequent jump intensity analysis.

[0056] The peak jump increment order is arranged sequentially according to the time index and multiplied by the power increase of the corresponding time period to obtain the instantaneous jump accumulation value. Then, the jump accumulation value is smoothed using a sliding time window averaging method, and a continuous change curve of jump intensity is calculated. The high point of this curve corresponds to the stage of rapid power increase. To eliminate abnormal peak errors, the curve is resampled using the median removal method, correcting excessively high peak values ​​to the average adjacent value range. The energy area of ​​the corrected curve is calculated, i.e., the total integral of the jump intensity, which is used to characterize the overall intensity of the power jump. Finally, time-series discharge power intensity data is output, including jump amplitude, increment frequency, power accumulation rate, and the change law of intensity distribution over time, establishing a data framework for subsequent discharge power increment correlation analysis.

[0057] Step S22 includes the following steps:

[0058] Step S221: Perform slope profile analysis on the driving and climbing conditions to obtain a slope profile dataset; perform vectorization decomposition of vehicle total mass, wheel end load and inertial parameters based on the slope profile dataset and heavy load information to obtain the load coupling vector.

[0059] Step S222: Derive the increasing series of climbing load resistance based on the load coupling vector;

[0060] Step S223: Correlate the time-series discharge power intensity with the increasing series of climbing load resistance to obtain the discharge power intensity correlated with the resistance series.

[0061] Step S224: Perform nonlinear incremental evolution induction on the discharge power intensity associated with the resistance levels to obtain power increment evolution data between resistance levels; derive the power law relationship between the resistance levels and the power increment based on the power increment evolution data;

[0062] Step S225: Based on the power law relationship, perform incremental correlation analysis of discharge power under heavy load climbing conditions to obtain incremental correlation data of discharge power under heavy load climbing conditions.

[0063] In this embodiment of the invention, the collected data on the driving and climbing of new energy vehicles is first processed by slope profile analysis. Using actual road distance measurement data as the horizontal reference and the vertical elevation difference of road segments as the vertical variable, a distance-slope correlation table is established for each driving path segment. The slope change rate is calculated for continuous sampling points at 5m intervals. Then, spline interpolation is used to refit the irregular slope curve, generating a continuous and smooth slope profile curve. The slope value and slope change rate at each moment are extracted from this curve, and a slope profile dataset is constructed according to the time series. Subsequently, the slope profile data is matched with the vehicle's heavy load information to obtain the vehicle's total mass, front and rear axle load distribution, vehicle center of gravity height, wheel hub radius, rolling radius ratio, and moment of inertia parameters. Based on the mechanical decomposition method, the vehicle's total weight is decomposed into components along the slope direction and ground support reactions, and the action decomposition relationship is expressed in vector form. The wheel-end forces are calculated through vectorized projection to obtain the contribution ratio of each wheel-end load to the vehicle's traction requirements during the climbing process. By further integrating the inertia factor of the critical slope section with the changes in gravity components under acceleration, a three-dimensional load coupling vector is formed that includes the total mass of the vehicle, wheel-end load, and inertial parameters, providing a structural basis for subsequent calculations of the load-power relationship.

[0064] Based on the generated load coupling vector, a progressive series of climbing load resistance is derived. First, the load coupling vector is decomposed into a traction force component along the slope and a vertical support component. For the traction force component, the difference between the total driving force and the slope resistance is calculated based on the vehicle's dynamic equilibrium conditions, yielding the net traction demand of the vehicle at each profile position. Then, the traction demand values ​​are arranged in ascending order of slope gradient, differentially processed, and the resistance increment caused by a unit slope change is taken to form a slope-resistance increment comparison sequence. This sequence reflects the nonlinear trend of resistance increasing with slope gradient. The sequence is then segmented, and the average resistance increment for each slope interval is statistically analyzed, arranged in ascending numerical order to form a progressive series of climbing load resistance. This series provides the resistance growth amplitude and distribution structure between different slope segments, offering quantitative input for analyzing the relationship between discharge power increase characteristics and climbing load.

[0065] Based on the derived increasing series of slope load resistance, a correlation mapping is performed on the time-series discharge power intensity. First, the time-series discharge power intensity sequence is standardized to a data length consistent with the time resolution of the slope profile. Then, a mapping matrix between the slope resistance series and the discharge power intensity is established under the same time index, where each matrix element corresponds to a match between a slope resistance series value and a power intensity value. To eliminate errors caused by sampling intervals, an interpolation approximation method is used to correct low-sampling intervals. Next, the response gradient of power intensity to the resistance series is calculated, i.e., the ratio of the power change rate to the resistance change rate. By analyzing the distribution of this ratio on the time axis, the response characteristics of power changes under different slope segments and loads are obtained. Finally, discharge power intensity data correlated with the resistance series is generated, reflecting the dynamic response law of battery discharge capacity affected by increasing slope load.

[0066] Nonlinear incremental evolution induction was performed on the discharge power intensity data associated with drag levels. First, the average increment of power intensity within each adjacent drag level interval was calculated, forming a power increment sequence. Then, this sequence was smoothed and second-order differences were calculated to identify the evolution direction of power increments in the high-drag region. Time series aggregation transformed the power increments from a discrete distribution to a continuous function form, forming a power increment curve. A fitting algorithm was used to obtain the growth trend line of the curve. The residual sequence of the measured power increments and the trend line fitting was statistically summarized to obtain the power evolution offset rate. Based on the trajectory of the offset rate, the power increment evolution trajectory between each drag level was derived, obtaining the power increment evolution data between drag levels. Then, the logarithmic correspondence point distribution of drag increments and power increments was calculated. Linear regression fitting was used to obtain the power-law relationship exponent between the two, determining the power-law relationship between drag levels and power increments, establishing a quantitative mapping basis for subsequent correlation analysis.

[0067] Based on the derived power-law relationship between resistance levels and power increments, this study analyzes the correlation between discharge power increase under heavy-load climbing conditions. According to the power-law exponent, the power increment and resistance increase trend are combined over time, and a cumulative effect sequence of power response with resistance growth is established through numerical mapping. Subsequently, power response zones are divided according to slope intervals, and the power increase rate within each zone is integrated step-by-step to obtain the cumulative change in power over time. Then, the coupling curve between the power change rate and the load change rate is calculated using differential calculations to describe their dynamic relationship during the climbing process. Through complete mapping calculations, discharge power increase correlation data under heavy-load climbing conditions is generated. This data includes the power response gradient to load levels, the rate of change ratio, and the energy output growth law under different slope sections, providing data support for the accurate characterization and energy estimation of lithium battery discharge behavior.

[0068] Step S26 includes the following steps:

[0069] Step S261: Calculate the rate ratio of the increasing discharge power correlation data, and then calculate the instantaneous discharge current load index;

[0070] Step S262: Based on the instantaneous discharge current load index, derive the polarization effect charge loss process to obtain polarization charge loss data;

[0071] Step S263: Based on the internal resistance mutation clustering fitting data, perform thermal effect diffusion gradient coupling to obtain the thermal effect diffusion gradient;

[0072] Step S264: Based on the polarization power loss data and the thermal effect diffusion gradient, perform time-series divergence-cumulative coupling of power consumption to obtain power consumption divergence-cumulative data;

[0073] Step S265: Evaluate the theoretical power consumption based on the estimation logic; perform lithium battery power consumption estimation deviation coupling on the theoretical power consumption according to the power consumption divergence cumulative data to obtain power consumption estimation deviation coupling data.

[0074] In this embodiment of the invention, the power increment data sequence is arranged chronologically, and the ratio of power change to time change is calculated within a continuous time window to form a power increment rate sequence. Then, the power increment rate is standardized to remove the influence of time resolution differences, and the change ratio of adjacent segments in the continuous rate sequence is calculated as an increment rate ratio. After extracting the increment rate ratio, the power increment rate is synchronously mapped to the lithium battery current output value. Based on the functional relationship between power and current, the instantaneous discharge current value is calculated. The current load change ratio is multiplied by the real-time current output to form a current load change ratio sequence. Then, the instantaneous discharge current load index is calculated based on the integral of this ratio sequence. This index reflects the intensity of the impact of load change on the battery discharge current per unit time, and its numerical distribution is used to characterize the dynamic changes in battery load response, providing an accurate input basis for further analysis of the power loss process.

[0075] Using the calculated instantaneous discharge current load index, a mapping relationship between the current load and the battery electrode polarization behavior is established using the nonlinear electrochemical kinetic discharge response theory. The current load index is mapped to the calibration sample set stored in the historical polarization response database using a spectral iterative comparison method. The sample interval closest to the load index is matched in the database using the Euclidean minimum distance prediction model. The corresponding instantaneous polarization voltage mean and duration are extracted in the target segment. Based on the energy consumption conversion coefficient composed of polarization voltage and discharge current, the energy loss caused by polarization process per unit time is estimated. Then, under a given statistical duration, the energy consumption generated by all current regions above the polarization initiation threshold is accumulated to form polarization energy loss data for further energy consumption error analysis.

[0076] Temperature dimension identification was performed on the internal resistance abrupt cluster fitting data. The average internal resistance corresponding to different cluster intervals was correlated with the temperature rise data to calculate the internal resistance temperature rise change rate per unit time. Then, based on the continuous change trend of the temperature rise change rate corresponding to the time series, the temperature gradient sequence was obtained using the difference method. Next, the discharge current change curve was retrieved at the same time scale, and the Joule heat accumulation value was derived by multiplying the square of the current by the time, and then superimposed on the corresponding temperature gradient sequence. Weighted superposition processing was used to comprehensively calculate the heat diffusion intensity of different cluster segments, resulting in the heat gradient change curve over time. This curve is the thermal effect diffusion gradient, recording the conduction expansion rate and spatial distribution trend of heat at different discharge stages, providing a change path reference for coupled calculation of energy loss.

[0077] This paper proposes a time-series divergence-cumulative coupling method for power consumption based on polarization power loss data and thermal diffusion gradient. First, the polarization power loss data and thermal diffusion gradient are synchronized over time using isochronous index matching to establish a one-to-one correspondence between the polarization loss value and the corresponding thermal diffusion rate at each moment. Then, for each pair of matched data, an energy decay weight is calculated, and a local power consumption divergence factor is formed by calculating the time accumulation of the product of polarization loss and thermal gradient. These divergence factors are integrated in time sequence to obtain the power consumption divergence-cumulative curve. To reflect the overall distribution trend of energy loss, the curve is interpolated and smoothed, and the energy divergence growth rate at different time points is calculated. Finally, the power consumption divergence-cumulative data is obtained, including the contribution ratio of polarization decay, the thermal diffusion superposition rate, and the time distribution of energy divergence, comprehensively describing the cumulative diffusion characteristics of energy loss during battery discharge.

[0078] After obtaining the cumulative power consumption data, the theoretical power consumption is assessed based on the SOC estimation logic. First, the SOC estimation function is invoked through the background monitoring logic, using real-time discharge current, voltage, time, and capacity parameters to calculate the theoretical discharge power consumption sequence. Then, the theoretical power consumption and the cumulative power consumption data are registered and aligned in the time dimension, comparing the difference between the theoretical value and the cumulative power consumption value at the same moment to calculate the estimation deviation per time unit. Next, the mean deviation and the deviation increase rate are calculated within a sliding time window, and the deviation degree curve is formed by superimposing the average value within the continuous deviation interval. To prevent noise influence, the deviation curve is smoothed and the deviation gradient is calculated, thus forming a power consumption estimation deviation time series. The overall deviation degree quantification value is obtained by integrating the deviation time series with the time correlation function, ultimately generating coupled data on the power consumption estimation deviation degree. This data includes the power consumption deviation rate, deviation time distribution, and energy dissipation fitting trend, providing accurate deviation correction parameters for subsequent SOC prediction.

[0079] In another embodiment of step S262, after obtaining the instantaneous discharge current load index after completing step S261, based on the mechanism of current-induced polarization during cell operation and its causal relationship with energy loss, a non-steady-state polarization loss derivation method based on electrochemical dynamic response is used to quantify the energy loss caused by polarization. The specific processing flow is as follows: First, the instantaneous discharge current load index data is constructed into a continuous function sequence in chronological order, local extreme points of current load change are extracted, and the data segments of one time window before and after these extreme points are used as the polarization reaction analysis interval; second, the polarization initiation threshold is determined based on the battery physical characteristics, and when the current load index exceeds this threshold, the battery is considered to be in a polarized operating state. Under polarized operation, the standard polarization response curve is referenced in the battery production test parameter library. The amplitude of the standard curve is adjusted using the actual current value within the interval. This adjustment is further refined by considering factors such as battery operating temperature, state of charge, and cycle count, thereby accurately fitting the local polarization voltage increment under actual operating conditions. Subsequently, the fitted polarization voltage increment value is used to construct the output power offset with the average discharge current over the same period. The additional energy loss caused by polarization per unit time is calculated and numerically superimposed, ultimately outputting the polarization charge loss for each interval in the form of data frames. To improve the timing continuity processing capability, a timing smoothing interpolation method is used to fill the value range of discontinuous data segments caused by sampling frequency differences, ensuring gradient continuity at the connection points. Then, the charge loss data for all polarized segments is accumulated using a recursive integration method to form complete polarization charge loss data.

[0080] Step S264 includes the following steps:

[0081] The loss segment order is calculated based on the polarization charge loss data; the polarization loss divergence approach variance is obtained by performing a divergence series approach variance evolution on the polarization charge loss data according to the loss segment order.

[0082] The heat loss due to electrical energy conduction is estimated based on the thermal effect diffusion gradient, and the data of heat loss due to electrical energy conduction are obtained.

[0083] The skewed series of heat loss during electrical energy conduction is inferred from the heat loss data to obtain the skewed series of heat loss.

[0084] Based on the polarization loss divergence approaching variance and the skewed series of electrical energy heat loss, the time-series divergence accumulation coupling of power consumption is performed to obtain the power consumption divergence accumulation data.

[0085] In this embodiment of the invention, after acquiring polarized power loss data, the data is first segmented into equal-step segments according to the discharge time sequence interval. The segment order of the loss is calculated based on the change in the power loss rate within each time segment. The segment order is defined as the number of progressions in the loss trend between adjacent time segments, and its calculation is achieved by detecting the number of changes in the sign of the loss increase in adjacent intervals. After completing the segment order statistics, the polarized power loss data is divided into multiple divergence layers based on these orders. Variance is calculated for the loss value sequence of each layer, and the continuous sequence of variance changes over time is recorded. Subsequently, extreme value detection and normalization are performed on these variance values. The rate of change between variance extreme points is iteratively calculated using a nonlinear recursive method to derive the evolution law of the variance convergence interval. This process generates polarized loss divergence convergence variance data, which reflects the convergence evolution characteristics of the variance changing with the segment order during the accumulation of polarized power loss over time, providing a time series basis for the quantification of energy loss divergence behavior.

[0086] The thermal diffusion gradient is used to estimate the heat loss during electrical energy conduction. First, the thermal diffusion gradient is sorted by time, and power output and temperature rise data are extracted at the same time points to calculate the rate of temperature rise corresponding to power per unit time. The Joule heat release rate within that time period is determined by the ratio of power to temperature rise. Then, the ratio of the temperature rise rate to the heat transfer path length is calculated as the thermal diffusion efficiency index. This index is multiplied by the power input to obtain the heat loss value during electrical energy conduction for the corresponding time period. The heat loss values ​​for each time period are summarized to form a time series. This series is then smoothed and differentially analyzed to analyze the increasing trend of heat conduction over time. Finally, electrical energy conduction heat loss data is generated, recording the intensity and temporal distribution of energy loss due to thermal diffusion at different discharge stages, providing a basic data series for thermal correlation skewness analysis.

[0087] This paper proposes a time-series divergence-cumulative coupling method for power consumption based on the polarization loss divergence variance and the skewed series of electrical energy heat loss data. First, the two sets of data are matched and synchronized along the time axis to ensure a one-to-one correspondence between the polarization variance value and the skewed series in each time period. Then, at each synchronization time point, the variance value and the skewed series are weighted and superimposed. The weighting factor depends on the proportion of power increment per unit time during the discharge phase to reflect the strength of energy transfer. An initial divergence weight sequence is obtained through superposition calculation, and this sequence is then integrated and accumulated to obtain the initial value of power consumption divergence accumulation. Next, the initial value sequence is normalized and differencing to analyze the changing trend of the accumulated value over time and calculate the growth rate of energy divergence. Finally, the power consumption divergence accumulation data is output, which includes the combined effect of the variance increase and the change in skewed series within each time step. This data reflects the nonlinear divergence accumulation law of lithium battery energy over time under different operating conditions, enabling a time-series quantitative description of the evolution behavior of discharge energy loss.

[0088] Step S3 includes the following steps:

[0089] Step S31: Perform deviation trend characteristic analysis on the coupled data of power consumption estimation deviation to obtain power consumption estimation trend characteristics;

[0090] Step S32: Perform reinforcement learning on the power consumption estimation trend features based on the Q-learning algorithm to output power consumption estimation deviation trend learning data;

[0091] Step S33: Based on the power consumption estimation deviation trend learning data, predict the remaining power state of the lithium battery to obtain the remaining power state prediction data.

[0092] In this embodiment of the invention, the power consumption estimation deviation data is coupled and sorted according to time index order to eliminate the temporal non-stationarity caused by uneven sampling intervals. Then, the deviation data is smoothed using a moving average, and the average deviation value and deviation increase rate within a continuous time period are calculated using a time window statistical method. A difference operation is performed on the obtained mean sequence and increase sequence to obtain the deviation change rate sequence and deviation direction change sequence, which are used to represent the direction of deviation trend change. Then, the variance analysis method is used to calculate the dispersion of the deviation rate distribution within different time segments, and the deviation concentration interval is obtained by combining the trend direction information. For the fluctuation characteristics of high deviation segments, the curvature detection method is used to extract the extreme points of deviation change, and the time interval between adjacent extreme values ​​is calculated to represent the trend periodicity. The direction change, increase rate, and periodic time interval are used as feature dimensions to construct a deviation trend feature matrix, and the power consumption estimation trend features are obtained through feature sequence normalization. This feature data reflects the trend direction, change rate, and periodic fluctuation law of power consumption estimation deviation under heavy load ramp discharge scenarios, providing stable input for the reinforcement learning stage.

[0093] Reinforcement learning using the Q-learning algorithm is performed based on power consumption estimation trend features. First, the state space is defined as the combination of deviation magnitude and direction levels in the deviation trend feature matrix, and the action space is defined as the adjustment step size of the SOC evaluation value. The reward function is set as the absolute difference of the negative values ​​of the reduction in SOC prediction deviation in a single iteration, constraining prediction stability. After initializing the Q-value table, initial states are selected from historical deviation trend features and sequentially input into the Q-learning process. Action selection is performed according to an ε-greedy strategy, recording the state transition and corresponding immediate reward after each round of selection. The Q-value table is iteratively corrected using the Bellman update equation, gradually converging the Q-value to a stable optimal solution. To avoid local optima traps, the Q-value table is randomly resampled and updated after multiple iterations to improve global convergence performance. After sufficient training, the action strategy corresponding to the highest Q-value is the optimal correction path for the deviation trend. Finally, power consumption estimation deviation trend learning data is output, containing the optimal action value, average reward, and deviation correction rate sequence for each state during the reinforcement learning process, serving as input for remaining power state prediction.

[0094] The remaining state of charge (SOC) of lithium batteries is predicted based on the power consumption estimation deviation trend learning data. First, the optimal action sequence from the deviation trend learning data is mapped to the SOC update rule, and the SOC change at each moment is inferred from the state change sequence. This process uses historical current, voltage, and power data as constraints, iteratively calculating the predicted SOC value. Within each calculation cycle, the actual observed SOC value is differiated from the predicted SOC value to obtain the instantaneous prediction deviation. Cumulative weighted smoothing is performed on the deviation sequence to obtain the overall stable prediction value, suppressing the impact of local fluctuations. Subsequently, a time series backtracking method is used to correct the SOC error of the previous cycle, ensuring that each prediction is synchronized with the updated energy conversion efficiency parameters. Through multiple iterations, the SOC state prediction trajectory achieves stable convergence. The final output remaining SOC state prediction data includes the SOC prediction time series, the rate of change curve, and the distribution of remaining available capacity, accurately reflecting the entire process of energy state changes of new energy vehicle lithium batteries under high-power discharge and heavy-load ramping conditions, ensuring the continuity and consistency of the prediction results in terms of time-series response.

[0095] Specifically, based on the power consumption data after deviation coupling, the deviation trend, deviation gradient inflection point, and deviation accumulation rate over time are continuously analyzed, transforming the deviation change pattern in different operating segments into describable trend features. These trend features are then input into the Q-learning algorithm. Through iterative processes of state mapping, action selection, and reward update, the variation law of power consumption deviation under different operating conditions is continuously converged by the reinforcement learning process, thus forming deviation trend learning data that reflects the deviation evolution. Finally, using the learned trend data as the decision-making basis, by judging the correlation between the deviation accumulation structure and the discharge consumption path, the prediction of remaining power no longer relies on the static discharge curve but is based on dynamic inference of the deviation change direction and rate, enabling the output remaining power status result to accurately reflect the battery's true remaining level under complex loads and variable environments.

[0096] The present invention also provides a new energy vehicle lithium battery state prediction system for performing the new energy vehicle lithium battery state prediction method as described above. The new energy vehicle lithium battery state prediction system includes:

[0097] The feature analysis module is used to obtain the driving and climbing conditions and heavy load information of new energy vehicles, and to obtain the corresponding lithium battery discharge state; to perform high-power discharge feature analysis on the lithium battery discharge state, so as to output the high-power discharge time sequence change characteristics.

[0098] The power consumption estimation deviation analysis module is used to perform incremental correlation analysis of discharge power under heavy load climbing conditions based on the characteristics of high power discharge time sequence changes under driving and climbing conditions and heavy load information, and obtain incremental correlation data of discharge power under heavy load climbing conditions; and to perform lithium battery power consumption estimation deviation degree coupling based on the incremental correlation data of discharge power, and obtain power consumption estimation deviation degree coupling data.

[0099] The state prediction module is used to predict the remaining power state of the lithium battery by coupling data with the degree of deviation in power consumption estimation, so as to obtain the remaining power state prediction data.

[0100] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for predicting the state of lithium batteries in new energy vehicles, characterized in that, Includes the following steps: Step S1: Obtain the driving and hill-climbing conditions and heavy load information of the new energy vehicle, and obtain the corresponding lithium battery discharge status. High-power discharge characteristics are analyzed to determine the state of discharge of lithium batteries and output the time-series variation characteristics of high-power discharge. Step S2: Based on the driving and climbing conditions and heavy load information, analyze the incremental correlation of discharge power under heavy load climbing conditions to obtain the incremental correlation data of discharge power under heavy load climbing conditions; based on the incremental correlation data of discharge power, perform lithium battery power consumption estimation deviation coupling to obtain power consumption estimation deviation coupling data. Step S3: Based on the deviation of power consumption estimation, couple the data to predict the remaining state of power of the lithium battery to obtain the remaining state of power prediction data.

2. The method for predicting the state of lithium batteries in new energy vehicles according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Obtain background monitoring permissions for the lithium battery of the new energy vehicle and information on the driving and climbing conditions and heavy load of the new energy vehicle. Step S12: Based on the background monitoring permissions, obtain the lithium battery discharge state corresponding to the driving hill climbing condition and heavy load information; Step S13: Perform high-power discharge characteristic analysis on the discharge state of the lithium battery to obtain high-power discharge characteristics; Step S14: Perform time-series variation analysis on the high-power discharge characteristics to output the high-power discharge time-series variation characteristics.

3. The method for predicting the state of lithium batteries in new energy vehicles according to claim 2, characterized in that, Step S2 includes the following steps: Step S21: Perform step-jump intensity analysis on the timing variation characteristics of high-power discharge to obtain the timing discharge power intensity; Step S22: Based on the driving and climbing conditions and heavy load information, perform incremental correlation analysis of the discharge power intensity under heavy load climbing conditions to obtain incremental correlation data of discharge power under heavy load climbing conditions. Step S23: Based on the discharge power increment correlation data, deduce the change intensity of the circuit internal resistance; Step S24: Perform nonlinear clustering fitting on the intensity of sudden changes in line resistance to obtain clustering fitting data for sudden changes in resistance. Step S25: Obtain the estimation logic of the remaining SOC of the lithium battery through the background monitoring authority of the new energy vehicle lithium battery; Step S26: Based on the discharge power increment correlation data and the internal resistance mutation cluster fitting data, the estimation logic is coupled with the degree of lithium battery power consumption estimation deviation to obtain power consumption estimation deviation coupling data.

4. The method for predicting the state of lithium batteries in new energy vehicles according to claim 3, characterized in that, Step S21 includes the following steps: Step S211: Plot the high-power discharge time-series variation curve based on the high-power discharge time-series variation characteristics; mark the convex points in the high-power discharge time-series variation curve; Step S212: Calculate the slope of the high-power discharge timing curve near the peak based on the convexity point to obtain the slope of the curve near the peak. Step S213: Measure the variance of adjacent peak quantile widths of the high-power discharge timing curve based on the slope of the curve, so as to generate the variance of adjacent peak quantile widths. Step S214: Based on the variance of the width of the adjacent peak quantile, perform peak jump increment order analysis on the high-power discharge time-series change curve to obtain the peak jump increment order; Step S215: Perform step transition intensity analysis based on the peak transition increment order to obtain the time-series discharge power intensity.

5. The method for predicting the state of lithium batteries in new energy vehicles according to claim 3, characterized in that, Step S22 includes the following steps: Step S221: Perform slope profile analysis on the driving and climbing conditions to obtain a slope profile dataset; perform vectorization decomposition of vehicle total mass, wheel end load and inertial parameters based on the slope profile dataset and heavy load information to obtain the load coupling vector. Step S222: Derive the increasing series of climbing load resistance based on the load coupling vector; Step S223: Correlate the time-series discharge power intensity with the increasing series of climbing load resistance to obtain the discharge power intensity correlated with the resistance series. Step S224: Perform nonlinear incremental evolution induction on the discharge power intensity associated with the resistance levels to obtain power increment evolution data between resistance levels; derive the power law relationship between the resistance levels and the power increment based on the power increment evolution data; Step S225: Based on the power law relationship, perform incremental correlation analysis of discharge power under heavy load climbing conditions to obtain incremental correlation data of discharge power under heavy load climbing conditions.

6. The method for predicting the state of lithium batteries in new energy vehicles according to claim 3, characterized in that, Step S26 includes the following steps: Step S261: Calculate the rate ratio of the increasing discharge power correlation data, and then calculate the instantaneous discharge current load index; Step S262: Based on the instantaneous discharge current load index, derive the polarization effect charge loss process to obtain polarization charge loss data; Step S263: Based on the internal resistance mutation clustering fitting data, perform thermal effect diffusion gradient coupling to obtain the thermal effect diffusion gradient; Step S264: Based on the polarization power loss data and the thermal effect diffusion gradient, perform time-series divergence-cumulative coupling of power consumption to obtain power consumption divergence-cumulative data; Step S265: Evaluate the theoretical power consumption based on the estimation logic; perform lithium battery power consumption estimation deviation coupling on the theoretical power consumption according to the power consumption divergence cumulative data to obtain power consumption estimation deviation coupling data.

7. The method for predicting the state of lithium batteries in new energy vehicles according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Perform deviation trend characteristic analysis on the coupled data of power consumption estimation deviation to obtain power consumption estimation trend characteristics; Step S32: Perform reinforcement learning on the power consumption estimation trend features based on the Q-learning algorithm to output power consumption estimation deviation trend learning data; Step S33: Based on the power consumption estimation deviation trend learning data, predict the remaining power state of the lithium battery to obtain the remaining power state prediction data.

8. A lithium battery state prediction system for new energy vehicles, characterized in that, For executing the new energy vehicle lithium battery state prediction method as described in claim 1, the new energy vehicle lithium battery state prediction system includes: The feature analysis module is used to obtain the driving and climbing conditions and heavy load information of new energy vehicles, and to obtain the corresponding lithium battery discharge state; to perform high-power discharge feature analysis on the lithium battery discharge state, so as to output the high-power discharge time sequence change characteristics. The power consumption estimation deviation analysis module is used to perform incremental correlation analysis of discharge power under heavy load climbing conditions based on the characteristics of high power discharge time sequence changes under driving and climbing conditions and heavy load information, and obtain incremental correlation data of discharge power under heavy load climbing conditions; and to perform lithium battery power consumption estimation deviation degree coupling based on the incremental correlation data of discharge power, and obtain power consumption estimation deviation degree coupling data. The state prediction module is used to predict the remaining power state of the lithium battery by coupling data with the degree of deviation in power consumption estimation, so as to obtain the remaining power state prediction data.

Images