Differential pressure time series data prediction method and device based on improved ARIMA model

Abstract

The invention discloses a differential pressure time series data prediction method and device based on an improved ARIMA model, and the method comprises the steps: collecting and packaging voltage data of a single battery of energy storage equipment, obtaining time series data, carrying out the cleaning processing, and obtaining the cleaned time series data; carrying out ADF stability verification on the cleaned time sequence data; after the time sequence data passes the cleaning, carrying out Ljung-Box white noise detection on the cleaned time sequence data; generating prediction data after checking; constructing an ARIMA model and determining parameters; training the ARIMA model by setting a training data set; performing performance evaluation on the trained ARIMA model by setting a verification data set to generate an evaluation index; when the evaluation index reaches a set standard, training is completed, and a trained ARIMA model is obtained; and inputting the prediction data into the trained ARIMA model for prediction processing to obtain differential pressure prediction data. The accuracy of predicting the voltage difference of the single battery is improved, the voltage abnormity can be monitored, and then the stability of energy storage equipment is maintained.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage equipment operation, inspection and maintenance, and in particular to a pressure difference time series data prediction method and device based on an improved ARIMA model. Background Art

[0002] In today's society, with the rapid development of the economy and the increasing advancement of science and technology, electricity has become an indispensable energy foundation for modern society. Whether it is daily life or industrial production, the demand for electricity is growing continuously. With the advancement of Industry 4.0, the demand for electricity in the manufacturing industry is also rising, especially in the high-tech and equipment manufacturing industries, such as photovoltaic equipment and electrical machinery, whose electricity consumption growth rate is significantly higher than that of traditional high-energy-consuming industries. The rapid development of these industries has not only promoted the growth of electricity consumption, but also put forward higher requirements for electricity quality and stability. In order to meet this challenge, power grid companies need to strengthen the construction of power grid infrastructure and improve the load capacity of the power grid. At the same time, by optimizing power dispatching and strengthening equipment maintenance, the safe and stable operation of the power grid during peak hours can be ensured. With the continuous improvement of electric energy storage technology, energy storage systems play an important role in controlling electricity demand. As an important part of the energy dispatching and stabilization of the power grid load of the power system, the stability of energy storage equipment is crucial. Therefore, monitoring, status prediction and maintenance and repair of energy storage equipment can effectively improve the stability of the energy storage system. The voltage difference of the battery pack's single cell voltage is an important indicator for determining the state of the energy storage device. When the voltage difference between the single cells in the battery pack is too large, "breakdown" is likely to occur, causing serious accidents such as fire or explosion. In addition, the voltage difference may also cause the temperature of the battery itself to rise too quickly, leading to battery failure or accelerated aging. Excessive voltage difference will affect the battery's voltage, capacity, internal resistance and other performance parameters, thereby affecting the battery's safety, reliability and service life, and further affecting the stability of the entire energy storage device, bringing huge safety hazards to the energy storage power station. Therefore, the prediction of battery voltage difference can effectively monitor the real-time status of the energy storage device and eliminate all unstable factors in the bud.

[0003] At present, the complexity and uncertainty of battery voltage difference prediction make voltage difference prediction a very challenging task. There are many reasons for the abnormal voltage difference of battery cells, such as uneven battery charging, long-term battery disuse, battery quality problems, inconsistent cell capacity decay rate, inconsistent DC internal resistance growth and even temperature difference. In addition, the daily change of voltage difference is closely related to the charge and discharge state of the battery. Therefore, traditional statistical methods and linear models have limitations in voltage difference prediction.

[0004] Therefore, how to invent a method for predicting the voltage differential data of energy storage equipment on a daily basis, improve the accuracy of battery cell voltage differential prediction, monitor voltage anomalies, and maintain the stability of energy storage equipment has become an urgent problem to be solved. Summary of the invention

[0005] To this end, the present invention provides a method and device for predicting pressure difference time series data based on an improved ARIMA model. The method gives full play to the characteristics of the ARIMA model that is suitable for processing time series data and can well capture the trend and seasonal changes of time series data, solves the problem that traditional statistical methods and linear models are difficult to accurately capture the laws of time series data, and better simulates the complexity of battery pressure difference fluctuations, thereby improving the accuracy of pressure difference prediction, and further improving the initiative of monitoring and maintenance of energy storage equipment.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a method for predicting pressure difference time series data based on an improved ARIMA model, comprising: The highest and lowest voltage data of the single battery of the energy storage device are collected to obtain voltage data; the voltage data are packaged to obtain time series data; the time series data are cleaned by setting a processing strategy to obtain cleaned time series data; Perform ADF stationarity verification on the cleaned time series data. If the cleaned time series data is stable, proceed to the next step. If the cleaned time series data is unstable, perform differential calculation and re-perform ADF stationarity verification until the cleaned time series data becomes stable. Performing a Ljung-Box white noise test on the cleaned time series data, if the cleaned time series data is a white noise sequence, taking the cleaned time series data as prediction data; if the cleaned time series data is not a white noise sequence, taking the cleaned time series data as unpredictable data, and re-collecting the voltage data; Constructing an ARIMA model and determining the parameters of the ARIMA model by setting a confirmation strategy; The ARIMA model is trained by setting a training data set; the performance of the trained ARIMA model is evaluated by setting a validation data set to generate an evaluation index; the training is completed when the evaluation index reaches the set standard to obtain a trained ARIMA model; The predicted data is input into the trained ARIMA model for prediction processing to obtain pressure difference prediction data.

[0007] As a preferred solution of the pressure difference time series data prediction method based on the improved ARIMA model, in the process of cleaning the time series data through the set processing strategy, the missing voltage data is supplemented by taking the nearest voltage value through the upward filling strategy; the data with missing timestamps is supplemented with timestamps; outliers are identified through the data screening threshold strategy, and the outliers are processed by the strategy of deleting and then filling upwards.

[0008] As a preferred solution of a pressure difference time series data prediction method based on an improved ARIMA model, when the time series data after cleaning is unstable, when the difference calculation is performed, the formula for the difference calculation is: ; ; In the formula, is the observed value; is the initial observation value of the time series; is the counting sequence number; is the time interval; It is the forward difference calculation formula, and the differential data is obtained by subtracting the function value at the previous moment from the function value at the next moment.

[0009] As a preferred solution of a pressure difference time series data prediction method based on an improved ARIMA model, in the process of determining the parameters of the ARIMA model through the setting confirmation strategy, the expression of the ARIMA model is: ARIMA (p, d, q); parameter p is the number of autoregressive terms; parameter q is the number of sliding average terms; parameter d is the difference order; parameter d is determined through an ADF test process; and parameters p and q are determined by drawing autocorrelation and partial autocorrelation graphs and according to the truncation and tailing forms of the autocorrelation and partial autocorrelation graphs.

[0010] As a preferred solution of the pressure difference time series data prediction method based on the improved ARIMA model, in the process of performing performance evaluation on the trained ARIMA model by setting the verification data set and generating the evaluation index, the evaluation index is the root mean square error and the mean absolute error; the mathematical expression of the root mean square error is: ; In the formula, is the root mean square error; is the sample size; is the predicted value of the current sample; is the actual value of the current sample; The mathematical expression of the mean absolute error is: ; In the formula, is the mean absolute error.

[0011] The present invention also provides a pressure difference time series data prediction device based on an improved ARIMA model, based on the above pressure difference time series data prediction method based on an improved ARIMA model, comprising: The data acquisition and cleaning processing module is used to collect the highest and lowest voltage data of the single battery of the energy storage device to obtain voltage data; encapsulate the voltage data to obtain time series data; and clean the time series data by setting a processing strategy to obtain cleaned time series data; An ADF stationarity verification module is used to perform ADF stationarity verification on the cleaned time series data. If the cleaned time series data is stable, the next step of processing is performed; if the cleaned time series data is unstable, differential calculation is performed and ADF stationarity verification is performed again until the cleaned time series data becomes stable; A Ljung-Box white noise test module is used to perform a Ljung-Box white noise test on the cleaned time series data. If the cleaned time series data is a white noise sequence, the cleaned time series data is used as prediction data; if the cleaned time series data is not a white noise sequence, the cleaned time series data is used as unpredictable data and the voltage data is re-collected; An ARIMA model building module is used to build an ARIMA model and determine the parameters of the ARIMA model by setting a confirmation strategy; An ARIMA model training module is used to train the ARIMA model by setting a training data set; to perform performance evaluation on the trained ARIMA model by setting a validation data set to generate an evaluation index; and to complete the training when the evaluation index reaches a set standard to obtain a trained ARIMA model; The pressure difference prediction data acquisition module is used to input the prediction data into the trained ARIMA model for prediction processing to obtain the pressure difference prediction data.

[0012] As a preferred solution of a pressure difference time series data prediction device based on an improved ARIMA model, in the data acquisition and cleaning processing module, in the process of cleaning the time series data by using the set processing strategy, the missing voltage data is supplemented by taking the nearest voltage value through the upward filling strategy; the data with missing timestamps is supplemented with timestamps; outliers are identified by screening data threshold strategy, and the outliers are processed by adopting the strategy of deleting and then filling upwards.

[0013] As a preferred solution of a pressure difference time series data prediction device based on an improved ARIMA model, in the ADF stationarity verification module, when the cleaned time series data is unstable, when the differential calculation is performed, the differential calculation formula is: ; ; In the formula, is the observed value; is the initial observation value of the time series; is the counting sequence number; is the time interval; It is the forward difference calculation formula, and the differential data is obtained by subtracting the function value at the previous moment from the function value at the next moment.

[0014] As a preferred solution of a pressure difference time series data prediction device based on an improved ARIMA model, in the ARIMA model construction module, in the process of determining the parameters of the ARIMA model through the setting confirmation strategy, the expression of the ARIMA model is: ARIMA (p, d, q); parameter p is the number of autoregressive terms; parameter q is the number of sliding average terms; parameter d is the difference order; parameter d is determined through an ADF test process; and parameter p and parameter q are determined by drawing autocorrelation and partial autocorrelation graphs and according to the truncation and tailing forms of the autocorrelation and partial autocorrelation graphs.

[0015] As a preferred solution of a pressure difference time series data prediction device based on an improved ARIMA model, in the ARIMA model training module, in the process of performing performance evaluation on the trained ARIMA model through the set verification data set and generating the evaluation index, the evaluation index is the root mean square error and the mean absolute error; the mathematical expression of the root mean square error is: ; In the formula, is the root mean square error; is the sample size; is the predicted value of the current sample; is the actual value of the current sample; The mathematical expression of the mean absolute error is:

[0016] In the formula, is the mean absolute error.

[0017] The present invention has the following advantages: the present invention acquires voltage data by collecting the highest and lowest voltage data of the single battery of the energy storage device; encapsulates the voltage data to acquire time series data; cleans the time series data by setting a processing strategy to acquire cleaned time series data; performs ADF stationarity verification on the cleaned time series data, and if the cleaned time series data is stable, performs the next step of processing; if the cleaned time series data is unstable, performs differential calculation, and performs ADF stationarity verification again until the cleaned time series data becomes stable; performs Ljung-Box white noise test on the cleaned time series data, and .... If the sequence data is a white noise sequence, the cleaned time series data is used as the prediction data; if the cleaned time series data is not a white noise sequence, the cleaned time series data is used as unpredictable data, and the voltage data is recollected; an ARIMA model is constructed, and the parameters of the ARIMA model are determined by setting a confirmation strategy; the ARIMA model is trained by setting a training data set; the performance of the trained ARIMA model is evaluated by setting a verification data set, and an evaluation index is generated; when the evaluation index reaches the set standard, the training is completed and the trained ARIMA model is obtained; the prediction data is input into the trained ARIMA model for prediction processing, and the pressure difference prediction data is obtained. The model provided by the present invention is simple and easy to understand, easy to understand and apply, and the ARIMA model prediction belongs to the time series prediction of a single variable, only considering the lag and difference of the data, and can well capture the trend and seasonal changes of the data, and can perform data prediction without huge data, which greatly reduces the resource consumption of the platform. According to the characteristics of ARIMA model prediction, a single prediction of a small amount of data is more accurate, and the amount of data required for prediction is small. The sliding window method is used to realize the training of long-distance prediction of data sets, and a longer and more accurate prediction data set can be generated. The ARIMA model predicts pressure difference data through the sliding window method to solve the problem that the model cannot establish long-term dependencies between time series, so as to provide more accurate pressure difference prediction data. The present invention has strong portability. After training according to the actual situation of the location where the model is deployed, it supports migrating the pressure difference prediction model to different environments and platforms, and can flexibly adapt to different field environments. The model provided by the present invention requires a small amount of data for prediction. Only pressure difference time series data is required, which is a univariate prediction model only for time series data, so the performance requirements of the machine are very small, which greatly reduces resource consumption. The model provided by the present invention is simple and easy to understand, easy to explain and apply. The model only needs to adjust 3 parameters, among which the d difference is determined in the ADF verification, and the determination of p and q can also be directly determined using BIC, which greatly reduces the process of manual parameter adjustment. The ARIMA model can handle nonlinear relationships and unconventional data.Taking into account the lag and difference of data, the present invention can effectively capture the trend and seasonal changes of data. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0019] The structures, proportions, sizes, etc. illustrated in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with the technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantial technical significance. Any structural modification, change in proportion or adjustment of size shall still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0020] Figure 1 A schematic flow chart of a method for predicting pressure difference time series data based on an improved ARIMA model provided in Example 1 of the present invention; Figure 2 A schematic diagram of a specific implementation process of a pressure difference time series data prediction method based on an improved ARIMA model provided in Example 1 of the present invention; Figure 3 This is a schematic diagram of the architecture of a pressure difference time series data prediction device based on an improved ARIMA model provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0021] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Example 1

[0023] See also Figure 1 and Figure 2 Embodiment 1 of the present invention provides a method for predicting pressure difference time series data based on an improved ARIMA model, comprising the following steps: S1. Collect the highest and lowest voltage data of the single battery of the energy storage device to obtain voltage data; encapsulate the voltage data to obtain time series data; clean the time series data by setting a processing strategy to obtain cleaned time series data; S2, performing ADF stationarity verification on the cleaned time series data. If the cleaned time series data is stable, proceed to the next step of processing; if the cleaned time series data is unstable, perform differential calculation and re-perform ADF stationarity verification until the cleaned time series data becomes stable; S3, performing Ljung-Box white noise test on the cleaned time series data. If the cleaned time series data is a white noise sequence, the cleaned time series data is used as prediction data; if the cleaned time series data is not a white noise sequence, the cleaned time series data is used as unpredictable data, and the voltage data is re-collected; S4, constructing an ARIMA model, and determining the parameters of the ARIMA model by setting a confirmation strategy; S5, training the ARIMA model by setting a training data set; performing performance evaluation on the trained ARIMA model by setting a validation data set to generate an evaluation index; completing the training when the evaluation index reaches the set standard to obtain a trained ARIMA model; S6. Input the predicted data into the trained ARIMA model for prediction processing to obtain pressure difference prediction data.

[0024] In this embodiment, in step S1, the highest and lowest voltage data of the single battery of the energy storage device are collected to obtain voltage data; the voltage data are packaged to obtain time series data; the time series data are cleaned by setting a processing strategy to obtain cleaned time series data; Specifically, in the data collection and data set construction stage, the highest and lowest voltage data of the energy storage device are obtained from the time series database, and the timestamps and values ​​of the data are encapsulated into a time series. The voltage difference data obtained from the time series database has the following characteristics: the timestamp interval is not uniform, the time in minutes is not fully filled, and the amount of data per minute is different, so the data needs to be cleaned and formatted.

[0025] Data cleaning is mainly to ensure the quality and consistency of data. First, the timestamp density is processed to keep only one data per minute per day, a total of 1440 data. If the timestamp data is insufficient, the missing timestamp is supplemented, and the missing voltage data value is supplemented by the method of upward filling to obtain the nearest voltage value to complete the formatting of all data.

[0026] The improved ARIMA model inputs very little forecast data each time, so abnormal values ​​or null values ​​will have a great impact on the forecast results. Therefore, it is necessary to process the abnormal values ​​and null values ​​of voltage. The abnormal values ​​are identified by screening the data threshold, and the abnormal values ​​are processed by deleting them and then filling them up.

[0027] In this embodiment, in step S2, the cleaned time series data is subjected to ADF stationarity verification. If the cleaned time series data is stable, the next step of processing is performed; if the cleaned time series data is unstable, differential calculation is performed, and ADF stationarity verification is performed again until the cleaned time series data becomes stable; Specifically, the ARIMA model requires that the time series data is stable, or that the time series after differentiation is stable. The present invention uses the ADF test (unit root test statistic) to test the data stationarity, and then uses Ljung-Box to verify whether the sequence is a white noise sequence. The ADF test is a method for testing whether a time series is stationary, and its null hypothesis is that there is a unit root, that is, non-stationary. According to the results of the ADF test, it is possible to determine whether to reject the null hypothesis, that is, to prove that the time series is stationary, by comparing the test statistic and the critical value, as well as the p-value and the significance level. If the test time series is non-stationary, the ADF verification is continued after the differentiation, and the differential formula is expressed as: ; ; In the formula, is the observed value; is the initial observation value of the time series; is the counting sequence number; is the time interval; It is the forward difference calculation formula, and the differential data is obtained by subtracting the function value at the previous moment from the function value at the next moment.

[0028] function The increment in each interval for The first-order forward difference of is the finite difference in calculus. The forward difference is usually the equivalent operation of the differential in a discrete function. The ADF test result is mainly determined by five parameters, which have different meanings: The first parameter, the result of the adt test, referred to as T value, represents the t statistic.

[0029] The second parameter, p-value, represents the probability value corresponding to the t statistic. The p-value represents the probability of a sample occurrence or observation under the condition of the null hypothesis. If the p-value is less than the threshold of the probability event (0.05 or 0.01), then the null hypothesis is rejected (that is, the data is stationary), otherwise the null hypothesis is accepted.

[0030] The third parameter is the lag order. The lag order in the time series refers to the number of event intervals between the current data point and the previous data points. If the lag order is too low, the model may not be able to capture all the important features of the time series, resulting in insufficient prediction accuracy; if it is too high, the model is too complex and overfits the training data, thus failing to generalize to new data.

[0031] The fourth parameter, AIC (Akaike Information Criterion), is a criterion for measuring the goodness of fit of a statistical model. The model with the smallest AIC is selected.

[0032] The fifth parameter, compared with the first parameter, is the critical ADF test value at 99%, 95%, and 90% confidence intervals.

[0033] In this embodiment, in step S3, the cleaned time series data is subjected to a Ljung-Box white noise test. If the cleaned time series data is a white noise sequence, the cleaned time series data is used as prediction data; if the cleaned time series data is not a white noise sequence, the cleaned time series data is used as unpredictable data, and the voltage data is re-collected. Specifically, after completing the ADF stationarity verification, the time series data is subjected to the Ljung-Box white noise test, whose main parameters are the lbbalue test statistic, and pvalue is based on the p statistic of the chi-square distribution. If the pvalue is greater than 0.05, it means that the null hypothesis (the sequence is a white noise sequence) cannot be rejected, and the cleaned time series data is used as the predicted data; otherwise, the null hypothesis is rejected and the sequence is not a white noise sequence, and the cleaned time series data is used as unpredictable data, and the voltage data is recollected.

[0034] In this embodiment, in step S4, an ARIMA model is constructed, and the parameters of the ARIMA model are determined by setting a confirmation strategy; Specifically, the construction of the ARIMA model requires the determination of three important parameters, p, d, and q. d is the order of difference, which can be determined through the ADF test process. p and q are determined by drawing autocorrelation and partial autocorrelation graphs, and then the values ​​of p and q are determined based on the truncation and tailing of the image. Finally, the final values ​​of p and q are determined by continuously fine-tuning the comparison results to determine the ARIMA model.

[0035] The ARIMA model consists of three parts, namely autoregression (AR), differencing (I), and moving average (MA). The model is denoted as ARIMA (p, d, q), where p is the number of autoregression terms, q is the number of moving average terms, and d is the order of differences made to make it a stationary series. Difference is a key step. The time series data predicted by the ARIMA model must be stationary. Unstable data cannot capture the time series regularity.

[0036] (1) AR Model Describe the relationship between the current value and the historical value, and use the historical time data of the variable itself to predict itself. The autoregressive model must meet the requirements of stability. In general, the formula definition of the p-order autoregressive process is: ; In the formula, is the current value; is a constant term; p is the order; is the autocorrelation coefficient; is the error, i.e. white noise.

[0037] Formula expansion: ; In the formula, is the observed value at time t; are parameters of the AR model, used to describe the relationship between the current value and the past p time points; is the error term at time t.

[0038] From the above formula, we can see that the current value is predicted by the historical value. p is an order in the autoregressive model, which means that the historical values ​​of the previous periods are used to predict the current value. Limitations of the autoregressive model: The autoregressive model uses its own data for prediction; the time series data must be stable; the time series data must be correlated. If the autocorrelation coefficient is less than 0.5, it should not be used; autoregression is only suitable for predicting phenomena related to its own previous period.

[0039] (2) MA model The moving average model focuses on the accumulation of error terms in the autoregressive model. In the AR model, if It is not a white noise, it is usually considered to be a moving average of order q.

[0040] ; In the formula, is a white noise sequence; is the relationship between the error between the current time point and the past q time points.

[0041] when , that is, the current value of the time series has nothing to do with the historical value, but only depends on the linear combination of historical white noise, and the MA model is obtained: ; The impact of historical white noise in the AR model directly affects the current forecast value (by affecting the historical time series value). The moving average method can effectively eliminate random fluctuations in the forecast. The formula of the q-order MA model is defined as: ; In the formula, is the relationship between the value at the current time point and the observed value.

[0042] (3) ARMA model Combining AR(p) and MA(q) yields the autoregressive moving average model (p, q): ; A random time series can be represented by an autoregressive moving average model, that is, the series can be explained by its own process or lagged values ​​and random disturbances. If the series is stationary, that is, its behavior does not change over time, then we can use the past behavior of the series to predict the future.

[0043] (4) ARIMA model AR and MA in ARIMA are AR model and MA model respectively, and I is the difference method. The difference calculation ensures the stability of the data. The autoregressive model (AR), moving average model (MA) and difference method (I) are combined to form the difference autoregressive moving average model ARIMA (p, d, q).

[0044] In this embodiment, in step S5, the ARIMA model is trained by setting a training data set; the performance of the trained ARIMA model is evaluated by setting a validation data set to generate an evaluation index; the training is completed when the evaluation index reaches the set standard, and the trained ARIMA model is obtained; Specifically, the data of the day before the date to be predicted is obtained as the model training data set, the real value of the predicted date is obtained as the verification data set, and the sliding window method is used to input the pressure difference data to train the ARIMA model for prediction. Then, the model is trained multiple times with the training set, and the model test is checked to minimize the prediction error. Finally, the model performance is evaluated using the verification data set. When the evaluation index reaches the set standard, the training is completed and the trained ARIMA model is obtained.

[0045] In this embodiment, the metrics used to evaluate the predicted data are the root mean square error (RMSE) and the mean absolute error (MAE). RMSE stands for Root Mean Square Error, which represents the sample standard deviation of the difference between the predicted value and the observed value. The root mean square error is used to illustrate the degree of sample dispersion. When performing nonlinear fitting, the smaller the RMSE, the better. Its mathematical expression is: ; In the formula, is the root mean square error; is the sample size; is the predicted value of the current sample; is the actual value of the current sample; MAE stands for Mean Absolute Error, which represents the average of the absolute errors between the predicted value and the observed value. MAE is a linear score, and all individual differences have equal weight on the average value. Therefore, RMSE has a higher difference penalty than MAE in residual calculation. The data expression of MAE is: ; In the formula, is the mean absolute error.

[0046] Both prediction data evaluation metrics can be evaluated. The RMSE evaluation result is more credible than the MAE. However, in order to ensure the accuracy of the evaluation, two different metrics are introduced for evaluation.

[0047] In this embodiment, in step S6, the prediction data is input into the trained ARIMA model for prediction processing to obtain pressure difference prediction data.

[0048] Specifically, Figure 2 As shown, the number of data obtained each time is set, and then each time it slides back, most of the last data is retained, and a small amount of new data is accessed. For example, 60 prediction data are used as a window, and 5 data points are predicted each time, then the next window will have 5 new data updates. Use x1, x2, ...., xi, ...., xn to represent the input set of each prediction data. First, the ARIMA model is established. In order to perform batch data prediction, the values ​​of p and q need to be automatically determined. This study uses the Bayesian Information Criterion (BIC) to determine the p and q values ​​of the ARIMA model. Similar to AIC, the penalty term for BIC used for model selection is larger than that of AIC, and the number of samples is taken into account. When the number of samples is too large, it can effectively prevent the model complexity from being too high due to excessive model accuracy. The formula for BIC is expressed as: ; when hour, Therefore, BIC penalizes model parameters more than AIC in large samples, which makes BIC more inclined to choose simple models with fewer parameters. The data set x1,x2,....,xi,....,xn is used by the ARIMA model to generate predicted data y1,y2,....,yi,....,yn. Finally, the predicted pressure difference data of each time is written into the file in sequence to generate complete pressure difference prediction data.

[0049] In summary, the present invention acquires voltage data by collecting the highest and lowest voltage data of the single battery of the energy storage device; encapsulates the voltage data to acquire time series data; cleans the time series data by setting a processing strategy to acquire cleaned time series data; performs ADF stationarity verification on the cleaned time series data, and if the cleaned time series data is stable, performs the next step of processing; if the cleaned time series data is unstable, performs differential calculation, and re-performs ADF stationarity verification until the cleaned time series data becomes stable; performs Ljung-Box white noise test on the cleaned time series data, and if the cleaned time series data is unstable, performs differential calculation, and re-performs ADF stationarity verification until the cleaned time series data becomes stable; performs Ljung-Box white noise test on the cleaned time series data, and if the cleaned time series data is unstable, performs ADF stationarity verification again .... If the data is a white noise sequence, the cleaned time series data is used as the prediction data; if the cleaned time series data is not a white noise sequence, the cleaned time series data is used as unpredictable data, and the voltage data is recollected; an ARIMA model is constructed, and the parameters of the ARIMA model are determined by setting a confirmation strategy; the ARIMA model is trained by setting a training data set; the performance of the trained ARIMA model is evaluated by setting a verification data set, and an evaluation index is generated; when the evaluation index reaches the set standard, the training is completed and a trained ARIMA model is obtained; the prediction data is input into the trained ARIMA model for prediction processing, and the pressure difference prediction data is obtained. The model provided by the present invention is simple and easy to understand, easy to understand and apply, and the ARIMA model prediction belongs to the time series prediction of a single variable, and only the lag and difference of the data are considered, which can well capture the trend and seasonal changes of the data, and data prediction can be performed without huge data, which greatly reduces the resource consumption of the platform. According to the characteristics of ARIMA model prediction, a single prediction of a small amount of data is more accurate, and the amount of data required for prediction is small. The sliding window method is used to realize the training of long-distance prediction of data sets, and a longer and more accurate prediction data set can be generated. The ARIMA model predicts pressure difference data through the sliding window method to solve the problem that the model cannot establish long-term dependencies between time series, so as to provide more accurate pressure difference prediction data. The present invention has strong portability. After training according to the actual situation of the location where the model is deployed, it supports migrating the pressure difference prediction model to different environments and platforms, and can flexibly adapt to different field environments. The model provided by the present invention requires a small amount of data for prediction. Only pressure difference time series data is required, which is a univariate prediction model only for time series data, so the performance requirements of the machine are very small, which greatly reduces resource consumption. The model provided by the present invention is simple and easy to understand, easy to explain and apply. The model only needs to adjust 3 parameters, among which the d difference is determined in the ADF verification, and the determination of p and q can also be directly determined using BIC, which greatly reduces the process of manual parameter adjustment. The ARIMA model can handle nonlinear relationships and unconventional data.Taking into account the lag and difference of data, the present invention can effectively capture the trend and seasonal changes of data.

[0050] It should be noted that the method of the embodiment of the present disclosure can be performed by a single device, such as a computer or a server. The method of the present embodiment can also be applied in a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present disclosure, and the multiple devices will interact with each other to complete the described method.

[0051] It should be noted that the above describes some embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0052] Example 2

[0053] See also Figure 3 Embodiment 2 of the present invention further provides a pressure difference time series data prediction device based on an improved ARIMA model, comprising: The data acquisition and cleaning processing module 001 is used to collect the highest and lowest voltage data of the single battery of the energy storage device to obtain voltage data; encapsulate the voltage data to obtain time series data; clean the time series data by setting a processing strategy to obtain cleaned time series data; ADF stationarity verification module 002 is used to perform ADF stationarity verification on the cleaned time series data. If the cleaned time series data is stable, the next step of processing is performed; if the cleaned time series data is unstable, differential calculation is performed and ADF stationarity verification is performed again until the cleaned time series data becomes stable; Ljung-Box white noise test module 003, used to perform Ljung-Box white noise test on the cleaned time series data, if the cleaned time series data is a white noise sequence, the cleaned time series data is used as prediction data; if the cleaned time series data is not a white noise sequence, the cleaned time series data is used as unpredictable data, and the voltage data is re-collected; ARIMA model building module 004, used to build an ARIMA model and determine the parameters of the ARIMA model by setting a confirmation strategy; ARIMA model training module 005 is used to train the ARIMA model by setting a training data set; to perform performance evaluation on the trained ARIMA model by setting a validation data set to generate an evaluation index; and to complete the training when the evaluation index reaches a set standard to obtain a trained ARIMA model; The pressure difference prediction data acquisition module 006 is used to input the prediction data into the trained ARIMA model for prediction processing to obtain the pressure difference prediction data.

[0054] In this embodiment, in the data acquisition and cleaning processing module 001, in the process of cleaning the time series data through the set processing strategy, the missing voltage data is supplemented by taking the nearest voltage value through the upward filling strategy; the data with missing timestamps is supplemented with timestamps; the abnormal values ​​are identified through the data screening threshold strategy, and the abnormal values ​​are processed by adopting the strategy of deleting and then filling upwards.

[0055] In this embodiment, in the ADF stationarity verification module 002, when the cleaned time series data is unstable, when the difference calculation is performed, the difference calculation formula is: ; ; In the formula, is the observed value; is the initial observation value of the time series; is the counting sequence number; is the time interval; It is the forward difference calculation formula, and the differential data is obtained by subtracting the function value at the previous moment from the function value at the next moment.

[0056] In this embodiment, in the ARIMA model construction module 004, in the process of determining the parameters of the ARIMA model through the setting confirmation strategy, the expression of the ARIMA model is: ARIMA (p, d, q); parameter p is the number of autoregressive terms; parameter q is the number of sliding average terms; parameter d is the difference order; parameter d is determined through the ADF test process; by drawing autocorrelation and partial autocorrelation graphs, and determining parameters p and q according to the truncation and tailing forms of the autocorrelation and partial autocorrelation graphs.

[0057] In this embodiment, in the ARIMA model training module 005, in the process of performing performance evaluation on the trained ARIMA model by setting the verification data set and generating the evaluation index, the evaluation index is the root mean square error and the mean absolute error; the mathematical expression of the root mean square error is: ; In the formula, is the root mean square error; is the sample size; is the predicted value of the current sample; is the actual value of the current sample; The mathematical expression of the mean absolute error is: ; In the formula, is the mean absolute error.

[0058] It should be noted that the information interaction, execution process and other contents between the modules of the above-mentioned system are based on the same concept as the method embodiment in Example 1 of the present application, and the technical effects they bring are the same as those of the method embodiment of the present application. For specific contents, please refer to the description in the method embodiment shown above in the present application, and will not be repeated here.

[0059] Example 3

[0060] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, in which a program code for a pressure difference time series data prediction method based on an improved ARIMA model is stored. The program code includes instructions for executing embodiment 1 or any possible implementation thereof, a pressure difference time series data prediction method based on an improved ARIMA model.

[0061] Computer-readable storage media can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0062] Example 4

[0063] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor; The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute a pressure difference time series data prediction method based on an improved ARIMA model in Example 1 or any possible implementation thereof.

[0064] Specifically, the processor can be implemented by hardware or by software. When implemented by hardware, the processor can be a logic circuit, an integrated circuit, etc.; when implemented by software, the processor can be a general-purpose processor implemented by reading software codes stored in a memory. The memory can be integrated in the processor or can be located outside the processor and exist independently.

[0065] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable systems. The computer instructions can be stored in a computer-readable storage medium, or transmitted from a computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center.

[0066] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing system, they can be concentrated on a single computing system, or distributed on a network composed of multiple computing systems, and optionally, they can be implemented by a program code executable by a computing system, so that they can be stored in a storage system and executed by the computing system, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0067] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A method for predicting pressure difference time series data based on an improved ARIMA model, characterized in that: include: Collect the highest and lowest voltage data of the single battery of the energy storage device to obtain voltage data; Packaging the voltage data to obtain time series data; Cleaning the time series data by setting a processing strategy to obtain cleaned time series data; Perform ADF stationarity verification on the cleaned time series data. If the cleaned time series data is stable, proceed to the next step. If the cleaned time series data is unstable, perform differential calculation and re-perform ADF stationarity verification until the cleaned time series data becomes stable. Performing a Ljung-Box white noise test on the cleaned time series data, if the cleaned time series data is a white noise sequence, taking the cleaned time series data as prediction data; if the cleaned time series data is not a white noise sequence, taking the cleaned time series data as unpredictable data, and re-collecting the voltage data; Constructing an ARIMA model and determining the parameters of the ARIMA model by setting a confirmation strategy; The ARIMA model is trained by setting a training data set; the performance of the trained ARIMA model is evaluated by setting a validation data set to generate an evaluation index; the training is completed when the evaluation index reaches the set standard to obtain a trained ARIMA model; The predicted data is input into the trained ARIMA model for prediction processing to obtain pressure difference prediction data.

2. The method for predicting pressure difference time series data based on the improved ARIMA model according to claim 1, characterized in that: In the process of cleaning the time series data by using the set processing strategy, the missing voltage data is supplemented by taking the most recent voltage value by using the upward filling strategy; Supplement timestamp processing for data with missing timestamps; Outliers are identified by filtering data threshold strategies, and are processed by removing and then filling up strategies.

3. The method for predicting pressure difference time series data based on the improved ARIMA model according to claim 2, characterized in that: When the cleaned time series data is unstable, the differential calculation is performed, and the differential calculation formula is: ； ； In the formula, is the observed value; is the initial observation value of the time series; is the counting sequence number; is the time interval; It is the forward difference calculation formula, and the differential data is obtained by subtracting the function value at the previous moment from the function value at the next moment.

4. The method for predicting pressure difference time series data based on the improved ARIMA model according to claim 3, characterized in that: In the process of determining the parameters of the ARIMA model through the setting confirmation strategy, the expression of the ARIMA model is: ARIMA (p, d, q); parameter p is the number of autoregressive terms; parameter q is the number of sliding average terms; parameter d is the difference order; parameter d is determined through the ADF test process; parameters p and q are determined by drawing autocorrelation and partial autocorrelation graphs and based on the truncation and tailing forms of the autocorrelation and partial autocorrelation graphs.

5. The method for predicting pressure difference time series data based on the improved ARIMA model according to claim 4, characterized in that: In the process of performing performance evaluation on the trained ARIMA model by setting the validation data set and generating the evaluation index, the evaluation index is the root mean square error and the mean absolute error; the mathematical expression of the root mean square error is: ； In the formula, is the root mean square error; is the sample size; is the predicted value of the current sample; is the actual value of the current sample; The mathematical expression of the mean absolute error is: ； In the formula, is the mean absolute error.

6. A pressure difference time series data prediction device based on an improved ARIMA model, adopting a pressure difference time series data prediction method based on an improved ARIMA model as described in any one of claims 1 to 5, characterized in that: include: The data collection and cleaning processing module is used to collect the highest and lowest voltage data of the single battery of the energy storage device to obtain voltage data; Encapsulating the voltage data to obtain time series data; cleaning the time series data by setting a processing strategy to obtain cleaned time series data; An ADF stationarity verification module is used to perform ADF stationarity verification on the cleaned time series data. If the cleaned time series data is stable, the next step of processing is performed; if the cleaned time series data is unstable, differential calculation is performed and ADF stationarity verification is performed again until the cleaned time series data becomes stable; A Ljung-Box white noise test module is used to perform a Ljung-Box white noise test on the cleaned time series data. If the cleaned time series data is a white noise sequence, the cleaned time series data is used as prediction data; if the cleaned time series data is not a white noise sequence, the cleaned time series data is used as unpredictable data and the voltage data is re-collected; An ARIMA model building module is used to build an ARIMA model and determine the parameters of the ARIMA model by setting a confirmation strategy; An ARIMA model training module is used to train the ARIMA model by setting a training data set; to perform performance evaluation on the trained ARIMA model by setting a validation data set to generate an evaluation index; and to complete the training when the evaluation index reaches a set standard to obtain a trained ARIMA model; The pressure difference prediction data acquisition module is used to input the prediction data into the trained ARIMA model for prediction processing to obtain the pressure difference prediction data.

7. The pressure difference time series data prediction device based on the improved ARIMA model according to claim 6, characterized in that: In the data acquisition and cleaning processing module, in the process of cleaning the time series data by using the set processing strategy, the missing voltage data is supplemented by taking the nearest voltage value by using the upward filling strategy; Supplement timestamp processing for data with missing timestamps; Outliers are identified by filtering data threshold strategies, and are processed by removing and then filling up strategies.

8. The pressure difference time series data prediction device based on the improved ARIMA model according to claim 7, characterized in that: In the ADF stationarity verification module, when the cleaned time series data is unstable, the differential calculation is performed, and the differential calculation formula is: ； ； In the formula, is the observed value; is the initial observation value of the time series; is the counting sequence number; is the time interval; It is the forward difference calculation formula, and the differential data is obtained by subtracting the function value at the previous moment from the function value at the next moment.

9. The pressure difference time series data prediction device based on the improved ARIMA model according to claim 8, characterized in that: In the ARIMA model construction module, in the process of determining the parameters of the ARIMA model through the setting confirmation strategy, the expression of the ARIMA model is: ARIMA (p, d, q); parameter p is the number of autoregressive terms; parameter q is the number of sliding average terms; parameter d is the difference order; parameter d is determined through the ADF test process; and parameters p and q are determined by drawing autocorrelation and partial autocorrelation graphs and according to the truncation and tailing forms of the autocorrelation and partial autocorrelation graphs.

10. A pressure difference time series data prediction device based on an improved ARIMA model according to claim 9, characterized in that: In the ARIMA model training module, in the process of performing performance evaluation on the trained ARIMA model by setting the validation data set and generating the evaluation index, the evaluation index is the root mean square error and the mean absolute error; the mathematical expression of the root mean square error is: ； In the formula, is the root mean square error; is the sample size; is the predicted value of the current sample; is the actual value of the current sample; The mathematical expression of the mean absolute error is: ； In the formula, is the mean absolute error.

Images