Lithium battery online service life prediction method based on data fusion and ARIMA model
阅读说明：本技术 一种基于数据融合与arima模型的锂电池在线寿命预测方法 (Lithium battery online service life prediction method based on data fusion and ARIMA model ) 是由 赵洪博 王清 赵琦 庄忱 冯文全 于 2019-10-25 设计创作，主要内容包括：本发明公开一种基于数据融合与ARIMA模型的锂电池在线寿命预测方法,步骤如下：步骤一：采集锂电池的等电压放电、充电时间间隔与电池容量数据；步骤二：计算数据层融合的权重并对数据进行融合；步骤三：训练估计ARIMA模型参数并检验ARIMA模型；步骤四：将步骤二融合后的数据通过ARIMA模型预测RUL与下一周期的SOH；步骤五：输入实时在线获取的电池状态观测数据,重复步骤二～步骤四,更新ARIMA预测模型,实现在线预测。本发明实现了基于数据融合与ARIMA模型的锂电池在线寿命预测,提高了ARIMA模型的寿命预测精度,实现了锂电池的在线寿命预测功能,完成了航天器件锂电池的可靠性分析。(The invention discloses a lithium battery online service life prediction method based on data fusion and an ARIMA model, which comprises the following steps: the method comprises the following steps: collecting data of the equal-voltage discharge and charge time interval and the battery capacity of the lithium battery; step two: calculating the weight of data layer fusion and fusing data; step three: training and estimating ARIMA model parameters and checking the ARIMA model; step four: predicting RUL and SOH of the next period by the data fused in the step two through an ARIMA model; step five: and inputting the battery state observation data acquired online in real time, repeating the second step to the fourth step, and updating the ARIMA prediction model to realize online prediction. The invention realizes the on-line life prediction of the lithium battery based on the data fusion and the ARIMA model, improves the life prediction precision of the ARIMA model, realizes the on-line life prediction function of the lithium battery, and completes the reliability analysis of the lithium battery of the aerospace device.)
1. A lithium battery online service life prediction method based on data fusion and an ARIMA model is characterized in that: the method comprises the following steps:
the method comprises the following steps: collecting data of the equal-voltage discharge and charge time interval and the battery capacity of the lithium battery;
step two: calculating the weight of data layer fusion and fusing data;
step three: training and estimating ARIMA model parameters and checking the ARIMA model;
step four: predicting RUL and SOH of the next period by the data fused in the step two through an ARIMA model;
step five: and inputting the battery state observation data acquired online in real time, repeating the second step to the fourth step, and updating the ARIMA prediction model to realize online prediction.
2. The on-line lithium battery life prediction method based on data fusion and ARIMA model as claimed in claim 1, wherein: the specific process of the step one is as follows:
collecting voltage from V in K discharge cycles before lithium battery1Is put to V2Time value of (D) is recorded as
3. The on-line lithium battery life prediction method based on data fusion and ARIMA model as claimed in claim 1, wherein: the second step comprises the following specific processes:
s21, respectively calculating Pearson correlation coefficients of the equal-voltage charging and discharging time intervals as the online indirect health factors and the battery capacity as the offline direct health factors, which are obtained in the step one;
s22, carrying out normalization processing on the data of the interval between the charging time and the discharging time of the equivalent voltage;
and S23, carrying out weighted average on the equal-voltage charging and discharging time interval data subjected to the normalization processing in the step S22 by using the correlation coefficient calculated in the step S21 to obtain input training data of the ARIMA model.
4. The on-line lithium battery life prediction method based on data fusion and ARIMA model as claimed in claim 1, wherein: the third step comprises the following specific processes:
s31, performing stationarity judgment on the input training data after the data layer fusion in the step two, if the sequence is not stable, performing d-order difference on the training data, and confirming a parameter d of the ARIMA model;
s32, determining the optimizing ranges of the other two parameters p and q of the ARIMA model, including the autoregressive order p and the moving average order q, and training the ARIMA model under each parameter;
s33, estimating the optimal ARIMA model parameters by using a Bayesian information criterion;
s34, checking whether the ARIMA model parameters are significant by calculating the T statistic of the ARIMA model parameters.
5. The on-line lithium battery life prediction method based on data fusion and ARIMA model as claimed in claim 1, wherein: the specific process of the step four is as follows:
s41, obtaining a trained optimal ARIMA model through the model parameters estimated in the third step, performing autoregressive moving average operation on the fusion data in the second step, and predicting the remaining service life of the battery and the battery health state in the next test period;
s42, the battery capacity sequence { C without training in the step oneiNormalizing the i | (K +1, K + 2.., N }, representing the i | (K + 2) | into the health state of the battery, and using the health state as a verification data set of the model prediction effect;
and S43, calculating the absolute error between the estimated value of the residual service life prediction and the verification data value of S42 each time, and measuring the accuracy and the effect of online real-time prediction.
6. The on-line lithium battery life prediction method based on data fusion and ARIMA model as claimed in claim 1, wherein: the concrete process of the step five is as follows:
s51, sequentially inputting the ith (i ═ K +1, K +2, …, N) lithium battery state observation data of the new test cycle, including the equal voltage discharge, charge time interval and battery capacity, and forming a new training data set with the i-1 test cycle observation data input so far;
s52, repeating the second step to the fourth step every time data of one test period are input, continuously updating the ARIMA model parameters on line until the data processing of all the test periods is finished, and finally obtaining the predicted value of the battery health state of N-K test periods;
and S53, calculating the root mean square error of the prediction result and the verification data, and taking the root mean square error as an index for judging whether the prediction effect is good or bad.
Technical Field
The invention belongs to the field of life Prediction and Health Management (PHM), and particularly relates to an online life prediction method for a lithium battery based on data fusion and an ARIMA model.
Background
Compared with other energy storage devices, the lithium battery has the advantages of high energy density, light weight, stable discharge and the like. The lithium battery is used as energy storage equipment of the spacecraft, so that the storage efficiency and stability of an energy system of the spacecraft can be improved, the effective load is improved, and the emission cost is reduced. Due to the complexity, large-scale and intelligentization of spacecraft equipment and special working environment, the requirements of after-event maintenance and periodic maintenance cannot be met. At present, the technical means for maintaining the reliable operation of the spacecraft mainly depends on a health state evaluation technology and a prediction technology.
The traditional life prediction method utilizes the life data of product samples, and the life distribution estimation of the product is obtained by fitting the failure time of each sample through a life distribution model. However, for the spacecraft lithium battery with the characteristics of a degradation type failure mode and long service life, only a few failure data can be obtained in a life test, so that the traditional life prediction method is not effective for the health state management of the spacecraft lithium battery. However, spacecraft components have a large amount of available state monitoring data and sensor historical data, so that a data-driven life prediction method and a data-driven life prediction model in the technical field of spacecraft product health state management and prediction become research hotspots. The data-driven life prediction methods can be further classified into statistical regression analysis methods and artificial intelligence analysis methods.
An Autoregressive differential Moving Average model (ARIMA) is one of the commonly used statistical regression methods, and linear weighting is performed on observed values of time series historical moments and noise, and then parameter estimation is performed on the ARIMA model through a statistical information criterion to obtain an optimal estimation model. The ARIMA model has a simple prediction structure, only uses the observed value of the time sequence of the ARIMA model, and does not need additional auxiliary variables during prediction. The traditional ARIMA model cannot capture nonlinear change, has low prediction precision on nonlinear change data, and is used for online prediction of the later State of Health (SOH) and the residual service life (RUL) of a battery, wherein the later State of Health (SOH) and the residual service life (RUL) of the battery have larger prediction errors.
The information fusion technology is an information comprehensive processing technology, and obtains an information processing effect which is more reliable, more accurate and more efficient than a single sensor through an information fusion algorithm by utilizing the redundancy and complementarity of information of a plurality of sensors. The information fusion technology can appear on different information levels, and comprises data layer fusion, feature layer fusion and decision layer fusion.
Disclosure of Invention
The invention aims to provide a lithium battery online service life prediction method based on data fusion and an ARIMA model, which utilizes a data layer fusion method to improve the precision of lithium battery service life prediction by solely utilizing the ARIMA model, improves the online service life prediction precision of the ARIMA model by updating the ARIMA model online, completes the prediction of the health state and the residual service life of a lithium battery of a spacecraft, realizes the real-time tracking of the change trend of the battery, and enhances the reliability of the lithium battery service life prediction in the later period.
In order to achieve the purpose, the invention provides a lithium battery online service life prediction method based on data fusion and an ARIMA model. According to the method, the equal-voltage discharge time interval and the equal-voltage charge time interval of the lithium battery are collected and used as indirect health factors input by online prediction, then the correlation coefficient of the two data and the battery capacity is used as a weight, information fusion is carried out on a data layer, and finally the fused health factors are subjected to online prediction through an ARIMA model to obtain the predicted values of the remaining service life and the health state. The invention effectively improves the prediction precision of the ARIMA model and realizes the on-line life prediction of the spacecraft lithium battery.
The invention discloses a lithium battery online life prediction method based on data fusion and an ARIMA model, which comprises the following implementation steps of:
the method comprises the following steps: collecting data of the equal-voltage discharge and charge time interval and the battery capacity of the lithium battery;
step two: calculating the weight of data layer fusion and fusing data;
step three: training and estimating ARIMA model parameters and checking the ARIMA model;
step four: predicting RUL and SOH of the next period by the data fused in the step two through an ARIMA model;
step five: and inputting the battery state observation data acquired online in real time, repeating the second step to the fourth step, and updating the ARIMA prediction model to realize online prediction.
Wherein, in the step one, the method for collecting the equal voltage discharge, the charging time interval and the battery capacity data of the lithium battery comprises the following steps:
collecting voltage from V in K discharge cycles before lithium battery1Is put to V2Time value of (D) is recorded as
Wherein, in the step two, the weight of data layer fusion is calculated and data is fused, which includes the following steps:
s21, respectively calculating Pearson correlation coefficients (Pearson correlation coefficient) of the equal-voltage charging and discharging time intervals as the online indirect health factors and the battery capacity as the offline direct health factors, which are acquired in the step one;
s22, carrying out normalization processing on the data of the interval between the charging time and the discharging time of the equivalent voltage;
and S23, carrying out weighted average on the equal-voltage charging and discharging time interval data subjected to the normalization processing in the step S22 by using the correlation coefficient calculated in the step S21 to obtain input training data of the ARIMA model.
Wherein, the training of the ARIMA model parameters and the checking of the ARIMA model in the third step is as follows:
s31, performing stationarity judgment on the input training data after the data layer fusion in the step two, if the sequence is not stable, performing d-order difference on the training data, and confirming a parameter d of the ARIMA model;
s32, determining the optimizing ranges of the other two parameters p and q of the ARIMA model, including the autoregressive order p and the moving average order q, and training the ARIMA model under each parameter;
s33, estimating the optimal ARIMA model parameters by using Bayesian Information Criterion (BIC);
s34, checking whether the ARIMA model parameters are significant by calculating the T Statistic (T-statistical) of the ARIMA model parameters.
Wherein, in step four, the method for predicting the RUL and the SOH of the next period by the ARIMA model after the fusion is performed is as follows:
and S41, obtaining a trained optimal ARIMA model through the model parameters estimated in the third step, and performing autoregressive moving average operation on the fusion data in the second step to predict the remaining service life of the battery and the battery health state in the next test period.
S42, the battery capacity sequence { C without training in the step oneiAnd (3) carrying out normalization processing on | i ═ K +1, K +2,. and N }, representing the state of health of the battery, and using the state of health as a verification data set of the model prediction effect.
And S43, calculating the Absolute Error (Absolute Error) between the estimated value of the residual service life prediction and the verification data value of S42 each time, and using the Absolute Error to measure the accuracy and effect of online real-time prediction.
Wherein, in the fifth step, "inputting the battery state observation data of the new test period, repeating the second step to the fourth step, updating the ARIMA prediction model, and realizing online prediction", the method comprises the following steps:
and S51, sequentially inputting the ith (i ═ K +1, K +2, … and N) lithium battery state observation data (including the equal voltage discharge, the charging time interval and the battery capacity) of the new test period, and forming a new training data set with the previously input observation data of the i-1 test period.
And S52, repeating the second step to the fourth step every time data of one test period are input, continuously updating the ARIMA model parameters on line until the data processing of all the test periods is finished, and finally obtaining the predicted value of the battery health state of N-K test periods.
And S53, calculating the Root Mean Square Error (RMSE) of the prediction result and the verification data, and taking the Root Mean Square Error (RMSE) as an index for judging whether the prediction effect is good or bad.
Through the steps, the online service life prediction of the lithium battery based on the data fusion and the ARIMA model is realized, the service life prediction precision of the ARIMA model is improved, the online service life prediction function of the lithium battery is realized, and the reliability analysis of the lithium battery of the aerospace device is completed.
According to the design of the invention, the online life prediction of the lithium battery based on the data fusion and the ARIMA model is realized, the algorithm parameter identification method is simple and low in complexity, and the ARIMA model is easy to continuously update.
According to the design of the invention, the observation data in the battery degradation model is comprehensively utilized through a data layer fusion method, the accuracy of the ARIMA model for predicting the residual service life by using a single information source is effectively improved, and the accuracy and robustness of the reliability analysis of the spacecraft lithium battery are improved (the prediction result is shown in detail in figure 1).
According to the design of the invention, indirect health factors which can represent the health state of the battery in the lithium battery degradation model are extracted on line, the health state value of the next period is predicted each time, the ARIMA prediction model is continuously updated, the real-time observation of the change of the health state of the battery is realized, and the online life state prediction precision of the lithium battery is improved (the prediction result is shown in detail in figure 2).
Drawings
FIG. 1 is a comparison of RUL prediction curves at Cycle60 using different data information for the ARIMA model.
Fig. 2 is a graph comparing curves for predicting battery health at Cycle60 for a conventional ARIMA model and an improved ARIMA model, both using data layer fusion information.
FIG. 3 is a flow chart of the lithium battery online life prediction method based on data layer fusion and an ARIMA model.
FIG. 4 is a graph of the discharge voltage of the B0005
Fig. 5 is a plot of the B0005 battery isopotential discharge time intervals.
FIG. 6 is a plot of the charging voltage of the
Fig. 7 is a graph of voltage charge time intervals for a B0005 battery or the like.
Fig. 8 is a battery capacity time curve for B0005 batteries.
Detailed Description
So that the manner in which the features, objects, and functions of the invention are attained and can be understood in detail, a more particular description of the invention, briefly summarized above, may be had by reference to the embodiments thereof which are illustrated in the appended drawings.
As shown in fig. 3, the invention provides an online lithium battery life prediction method based on data fusion and ARIMA model, which comprises the following specific implementation steps:
the first step is as follows: and collecting the data of the equal-voltage discharge and charge time interval and the battery capacity of the lithium battery.
The data used in the invention is derived from a lithium battery life test data set disclosed by a NASA (national Center of Excellence, PCoE), and the lithium battery data with the number of B0005 is utilized to predict the life.
The voltage is plotted against time during the first discharge (Cycle 1) of the B0005 lithium cell, as shown in fig. 4. In fig. 4, the horizontal axis represents the discharge time of the lithium battery in seconds(s), and the vertical axis represents the corresponding discharge voltage in volts (V). The battery successively goes through a discharging stage and a standing stage, in the discharging stage, the voltage is reduced from 4.2V to 2.6V due to the continuous discharging of the battery, and the voltage value reduction speed is in a trend of slowing first and then accelerating. During the resting phase, the voltage gradually rises back to 3.3V due to the self-charging of the battery. Because the voltage rising amplitude in the standing stage is influenced by more factors and is difficult to determine, the time interval in the lithium battery discharging stage is used as one of the training data for online prediction.
During the first discharge cycle of the B0005 lithium battery, a voltage value V is first recorded1Time value when 3.8V
Similarly, the voltage of the B0005 lithium battery was plotted against time during the first charging process (Cycle 1), as shown in fig. 6. The abscissa and ordinate axes of fig. 6 represent the same physical meaning as the discharge process. In the charging process, the lithium battery is charged by 1.5A constant current and 4.2V constant voltage sequentially. The invention utilizes the time interval of the lithium battery in the constant current charging stage as one of the data for on-line analysis and prediction.
During the first charging cycle of the B0005 lithium battery, a voltage value V is recorded first3Time value when 3.8V
As shown in fig. 8, the present invention directly uses the battery capacity information provided in the data set.
The second step is that: and calculating the weight of data layer fusion and fusing the data.
If the sequence of the equal voltage discharge time intervals is {
wherein the content of the first and second substances,
The equal voltage discharge time interval, the equal voltage charge time interval and the battery capacity obtained from the first step are not within an order of magnitude range, so data normalization processing is performed before prediction.
The battery state of health calculation formula represented by the normalized equal-voltage discharge time interval is as follows:
the battery state of health calculation formula expressed by the normalized equal voltage charging time interval is as follows:
the battery state of health calculation formula represented by the battery capacity after normalization is as follows:
wherein td1、tc1And C1Respectively shows the observation condition of the lithium battery state at the initial moment of the test.
Using calculated Pearson correlation coefficient r1、r2And carrying out weighted fusion on the normalized equal-voltage charging and discharging time interval data, wherein the fused battery health state information is represented as:
the third step: training estimates ARIMA model parameters and tests ARIMA models.
The equal voltage discharging time interval and the equal voltage charging time interval are in a descending trend along with the increase of the charging and discharging use times, so that the fused data are non-stable time sequences. In the training process of the ARIMA model, d-order difference needs to be carried out on training data until the sequence after difference meets the stability.
After the difference order d is determined, the training range of the autoregressive order p and the moving average order q of the ARIMA (p, d, q) model is determined, the parameters of each ARIMA model are trained, and then the model parameters are estimated by utilizing the Bayesian information criterion BIC to obtain the optimal ARIMA prediction model.
Assume a sample time sequence of { X }t}，t＝0，±1，···N，Is an estimate of the residual variance of the model, the BIC function is calculated as follows:
the smaller the BIC calculation value is, the better the fitting effect of the relevant model is, and the smaller the prediction error is.
As shown in Table 1, after the ARIMA model is trained, whether the ARIMA model parameters are significant is checked by calculating the T Statistic (T-statistical) of the ARIMA model parameters, and if the parameters are significant, the overall fitting effect of the model is acceptable.
Table 1 is the static significance of ARIMA (1, 1, 1) model parameters.
The fourth step: and predicting the RUL and the SOH of the next period by the fused data through an ARIMA model.
And obtaining the optimal ARIMA model through the model parameters calculated in the third step. And predicting the remaining service life RUL of the battery and the SOH of the battery in the next test period by using the ARIMA model according to the fused data. Suppose a stationary time sequence of { X }tWhite noise sequence of { ε }t0, ± 1., the formula for ARIMA (p, d, q) is as follows:
Xt＝φ1Xt-1+φ2Xt-2+…+φpXt-p+εt-θ1εt-1-θ2εt-2-…-θqεt-q
wherein p is the order of the autoregressive model, q is the order of the moving average model, phi1，φ2，...φpBeing parameters of an autoregressive model, theta1，θ2，...，θqAre moving average model parameters. The parameters are obtained through model training in the third step.
The untrained battery capacity in the first step CiAnd (3) carrying out normalization processing on | i ═ K +1, K +2,. and N }, and representing the normalized | i ═ K +1, K +2,. and N }, wherein the normalized | i ═ K +2,. and N } is represented as the state of health (SOH) of the battery and is used as a verification data set of the prediction effect of the ARIMA.
After each prediction is finished, the Absolute Error (Absolute Error) between the estimated value and the true value of the verification set of each prediction of the residual service life RUL is calculated and used for measuring the accuracy and the effect of online real-time prediction. The absolute error is calculated as follows:
AEi＝|RULi-RULi_predict|
wherein AE isiFor absolute error of i-th prediction, RULi_predictResidual useful life estimate for the ith prediction, RULiThe actual value of the residual service life of the ith test is shown.
As shown in table 2, the ARIMA model uses the equal voltage discharge time interval and the equal voltage charge time interval to predict the absolute error of the remaining service life, which is generally larger than the prediction error of the ARIMA model using the data layer fusion information. Therefore, the method improves the accuracy and robustness of the prediction of the residual service life of the ARIMA model by a data layer fusion method.
Table 2 is the absolute error AE of the ARIMA model using different data sources to predict remaining useful life.
The fifth step: and inputting the battery state observation data acquired online in real time, repeating the second step to the fourth step, and updating the ARIMA prediction model to realize online prediction.
When the second on-line prediction is carried out after the first life prediction is finished, the real health state value of the battery in the K +1 th test and the observation data of the states of the previous K lithium batteries are used as a training data set, and then the AIRMA model is trained and predicted for a new round to obtain the predicted value and the prediction error of the K +2 th test period.
And by analogy, after data of a new test period is input every time, repeating the second step to the fourth step, continuously updating the ARIMA model parameters on line until the data processing of all the test periods is finished, and completing the on-line prediction of the service life of the lithium battery. Finally, through continuous model training, the predicted values of the battery health states in N-K test periods are obtained, the root mean square error RMSE of the predicted values and the verification data is calculated, and the prediction effect of the model is evaluated.
The root mean square error calculation formula for the lithium battery health state prediction is as follows:
wherein, SOHi_predictFor the ith predicted state of health estimate, SOHiThe actual observed value of the health state of the lithium battery in the ith test is shown, and n is the number of the predicted service lives.
In the traditional prediction method, K known data are used for training an ARIMA model, and the state of health of the battery in the remaining N-K periods is predicted after the optimal ARIMA model is solved. The traditional prediction method has low prediction precision on the later-stage battery health state, does not dynamically update an ARIMA training model, and cannot be used for a battery model with the later-stage health state accelerated degradation. The invention improves the ARIMA model prediction method, predicts the state value of the next period each time, and is beneficial to online real-time observation of the local change condition of the battery health state. As shown in Table 3, the prediction accuracy of the improved ARIMA model is far superior to that of the traditional ARIMA model, and when a large number of training samples are available in the later period, the prediction accuracy can be directly used for predicting the state change of the residual period and predicting the residual service life of the lithium battery.
Starting point of prediction
Traditional ARIMA prediction method
Improved
60
0.0474
0.0132
65
0.1025
0.0135
70
0.1430
0.0139
75
0.1942
0.0142
80
0.2122
0.0144
85
0.2218
0.0148
90
0.1000
0.0097
95
0.1628
0.0098
100
0.1736
0.0101
Table 3 is the root mean square error RMSE of the predicted battery capacity of both the traditional ARIMA model and the improved ARIMA model using the data layer fusion information.
Although the present invention has been described with reference to the above embodiments, the embodiments are merely exemplary, and not restrictive, and it should be understood that various changes and substitutions may be made herein without departing from the spirit and scope of the invention as defined by the appended claims.
- 上一篇：一种医用注射器针头装配设备
- 下一篇：预测锂离子电池循环寿命的方法