Energy storage battery soc estimation method and system based on charge and discharge current

Abstract

The present application relates to the technical field of energy storage battery electrical performance test, in particular to a kind of energy storage battery SOC estimation method and system based on charge-discharge current, applied to lithium ion battery, obtain the battery state time series data of current sampling time period and previous sampling time period, then obtain the abnormal characteristics of each sampling time period, according to the abnormal characteristics of the two sampling time periods, obtain the error factor of current sampling time period, finally combine the abnormal data characteristics in the two current sampling time periods, adjust the data point weight in particle filter algorithm corresponding to current sampling time period, realize the accurate adjustment of weight, can reduce the interference of noise data, improve the accuracy of particle set obtained in particle filter algorithm, thereby improve the accuracy of energy storage battery SOC estimation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage battery performance testing, and particularly relates to an energy storage battery SOC estimation method and system based on charging and discharging current. BACKGROUND

[0002] As a common energy storage battery, the lithium ion battery is usually estimated by using a particle filtering algorithm to estimate the SOC (State of Charge) of the energy storage battery, especially the SOC of the energy storage battery with the charging and discharging current at the station level. However, in the existing SOC estimation process of the energy storage battery based on the particle filtering algorithm, the particle set is too dependent, and the new particle set is obtained by resampling the weight of the particles in the previous particle set, the weight of the particles with high weight is reserved and the weight of the particles with low weight is eliminated. Since the resampling may cause the particles with low weight to be eliminated as noise interference, the particles with high weight are finally reserved more, and since the particle set depends on the previous particle set for updating, when the data in a certain time period is abnormal or mutated, the accuracy of the new particle set obtained by the previous particle set will be reduced due to inaccurate weight confirmation, thereby affecting the accuracy of the SOC estimation of the energy storage battery. SUMMARY

[0003] Therefore, in order to solve the technical problem of inaccurate weight acquisition in the particle filtering algorithm in the SOC estimation process based on the particle filtering algorithm, the present application provides an energy storage battery SOC estimation method and system based on charging and discharging current.

[0004] The technical solutions adopted are as follows:

[0005] In a first aspect, the present application provides an energy storage battery SOC estimation method based on charging and discharging current, comprising:

[0006] obtaining battery state time series data of the current sampling time period and the previous sampling time period of the energy storage battery, wherein the battery state time series data at least includes current; the current is charging current or discharging current;

[0007] obtaining an abnormal feature sequence of each sampling time period based on the battery state time series data, wherein the abnormal feature sequence includes abnormal features of each sampling time point in the sampling time period, and the abnormal features are abnormal data features or normal data features;

[0008] According to the difference between the abnormal data features in the abnormal feature sequence of the current sampling time period and the previous sampling time period, and the correlation between the abnormal data features and the normal data features in the abnormal feature sequence of the current sampling time period, an error factor of the current sampling time period is obtained, the error factor being used to represent the degree of weight deviation of the data points in the current sampling time period;

[0009] According to the difference between the isolated states of the abnormal data features in the current sampling time period and the previous sampling time period in the corresponding abnormal feature sequence, and in combination with the error factor, the data point weight in the particle filter algorithm corresponding to the current sampling time period is adjusted.

[0010] In combination with the first aspect, in some possible implementation manners, the battery state time series data further includes a battery temperature and a battery internal resistance;

[0011] Based on the battery state time series data, an abnormal feature sequence of each sampling time period is obtained, including:

[0012] According to the current, a current abnormal state of each sampling time period is obtained;

[0013] According to the current abnormal state, and the battery temperature and the battery internal resistance, an abnormal feature sequence of each sampling time period is obtained.

[0014] In combination with the first aspect, in some possible implementation manners, the process of obtaining the current abnormal state of the sampling time period includes:

[0015] The current differences between each sampling time and its adjacent two sampling times in the sampling time period are obtained respectively, to obtain a first current difference and a second current difference;

[0016] The absolute value of the difference between the first current difference and the second current difference is obtained, to obtain a first current abnormal degree of each sampling time;

[0017] The average value of the absolute values of the first current differences in the sampling time period is obtained, to obtain a second current abnormal degree;

[0018] According to the first current abnormal degree and the second current abnormal degree, a current abnormal state of each sampling time in the sampling time period is obtained.

[0019] In combination with the first aspect, in some possible implementation manners, according to the current abnormal state, and the battery temperature and the battery internal resistance, an abnormal feature sequence of each sampling time period is obtained, including:

[0020] According to the battery temperature, a battery temperature fluctuation degree and a battery temperature range of the sampling time period are obtained;

[0021] According to the battery temperature and the battery internal resistance, a change correlation of the battery temperature and the battery internal resistance in a sampling time period is obtained;

[0022] According to the current abnormal state, the battery temperature fluctuation degree, the battery temperature range and the change correlation, an abnormal feature sequence in the sampling time period is obtained; the current abnormal state, the battery temperature fluctuation degree and the battery temperature range are positively correlated with the abnormal feature sequence, and the change correlation is negatively correlated with the abnormal feature sequence.

[0023] In combination with the first aspect, in some possible implementation manners, the energy storage battery SOC estimation method further includes:

[0024] The abnormal features in the abnormal feature sequence are clustered to obtain a plurality of class clusters;

[0025] Based on the overall numerical level of the abnormal features in each class cluster, the plurality of class clusters are divided into an abnormal class cluster and a normal class cluster; the abnormal features in the abnormal class cluster are the abnormal data features, and the abnormal features in the normal class cluster are the normal data features.

[0026] In combination with the first aspect, in some possible implementation manners, the difference between the abnormal data features in the abnormal feature sequences of the current sampling time period and the previous sampling time period is a quantity proportion difference; the quantity proportion difference is a difference between a quantity proportion of the abnormal data features in the abnormal feature sequence of the current sampling time period and a quantity proportion of the abnormal data features in the abnormal feature sequence of the previous sampling time period;

[0027] The association between the abnormal data features and the normal data features in the abnormal feature sequence of the current sampling time period is a centroid distance overall level; the centroid distance overall level is an average value of distances between centroids of the normal class clusters in the current sampling time period and a centroid of the abnormal class cluster in the current sampling time period;

[0028] The error factor of the current sampling time period is obtained according to the quantity proportion difference and the centroid distance overall level; the error factor and the quantity proportion difference are positively correlated with the centroid distance overall level.

[0029] In combination with the first aspect, in some possible implementation manners, according to the difference between the isolated states of the abnormal data features in the current sampling time period and the previous sampling time period in the corresponding abnormal feature sequences, and in combination with the error factor, a data point weight in a particle filter algorithm corresponding to the current sampling time period is adjusted, including:

[0030] obtaining an isolated state of the abnormal class cluster in the abnormal feature sequence of the current sampling time period to obtain a first isolated state, and obtaining an isolated state of the abnormal class cluster in the abnormal feature sequence of the previous sampling time period to obtain a second isolated state;

[0031] comparing the first isolated state and the second isolated state;

[0032] adjusting a data point weight corresponding to the current sampling time period in a particle filter algorithm according to a comparison result of the first isolated state and the second isolated state, and in combination with the error factor.

[0033] In combination with the first aspect, in some possible implementation manners, the first isolated state is obtained by:

[0034] obtaining a distance between a centroid of the abnormal class cluster in the current sampling time period and a center point of the abnormal feature sequence in the current sampling time period as a first abnormal distance, obtaining distances between centroids of the normal class clusters in the current sampling time period and the center point of the abnormal feature sequence in the current sampling time period as first normal distances, and calculating a sum of the first abnormal distance and all the first normal distances to obtain a first total distance;

[0035] taking a ratio of the first abnormal distance to the first total distance as the first isolated state;

[0036] the second isolated state is obtained by:

[0037] obtaining a distance between a centroid of the abnormal class cluster in the previous sampling time period and a center point of the abnormal feature sequence in the previous sampling time period as a second abnormal distance, obtaining distances between centroids of the normal class clusters in the previous sampling time period and the center point of the abnormal feature sequence in the previous sampling time period as second normal distances, and calculating a sum of the second abnormal distance and all the second normal distances to obtain a second total distance;

[0038] taking a ratio of the second abnormal distance to the second total distance as the second isolated state.

[0039] In combination with the first aspect, in some possible implementation manners, the adjusting of the data point weight corresponding to the current sampling time period in the particle filter algorithm according to the comparison result of the first isolated state and the second isolated state, and in combination with the error factor, comprises:

[0040] if the first isolated state is less than the second isolated state, an adjusted data point weight of a sampling time corresponding to an abnormal data feature in the particle filter algorithm corresponding to the current sampling time period is as follows:

[0041]

[0042] The adjusted data point weight corresponding to the sampling time corresponding to the normal data feature in the particle filter algorithm corresponding to the current sampling time period is as follows:

[0043]

[0044] If the first isolated state is greater than the second isolated state, the adjusted data point weight corresponding to the sampling time corresponding to the abnormal data feature in the particle filter algorithm corresponding to the current sampling time period is as follows:

[0045]

[0046] The adjusted data point weight corresponding to the sampling time corresponding to the normal data feature in the particle filter algorithm corresponding to the current sampling time period is as follows:

[0047]

[0048] wherein, is the initial data point weight corresponding to the rth abnormal data feature in the current sampling time period, is the adjusted data point weight corresponding to the rth abnormal data feature in the current sampling time period, is the initial data point weight corresponding to the st normal data feature in the current sampling time period, is the adjusted data point weight corresponding to the st normal data feature in the current sampling time period, and G is the normalized value of the error factor of the current sampling time period.

[0049] In a second aspect, an embodiment of the present application provides a SOC estimation system for an energy storage battery based on charging and discharging currents, comprising a memory and a processor, the memory being used to store executable program codes, and the processor being used to call and run the program codes from the memory to realize the above-mentioned SOC estimation method for an energy storage battery based on charging and discharging currents.

[0050] The present application has the technical effects including but not limited to the following: according to the battery state time series data of two adjacent sampling time periods of the current sampling time period and the previous sampling time period of the energy storage battery, the abnormal feature sequence of each sampling time period is obtained, thereby obtaining the specific classification of the abnormal feature of each sampling time in the sampling time period, whether it is an abnormal data feature or a normal data feature, thereby improving the accuracy of the analysis of abnormal data; then, according to the difference of the abnormal data features in the abnormal feature sequences of the current sampling time period and the previous sampling time period, and the correlation between the abnormal data features and the normal data features in the abnormal feature sequence of the current sampling time period, the error factor of the current sampling time period is obtained, the error factor is used to represent the weight deviation degree of the data points in the current sampling time period, the error factor is a very important parameter in the subsequent weight adjustment process, by obtaining the accurate error factor reflecting the features of the current sampling time period, the accuracy of the subsequent weight adjustment can be improved; finally, according to the difference of the isolated state of the abnormal data features in the current sampling time period and the previous sampling time period in the corresponding abnormal feature sequence, the data point weight corresponding to the particle filtering algorithm in the current sampling time period is adjusted in combination with the error factor. Therefore, the method provided by the present application can accurately adjust the data point weight corresponding to the particle filtering algorithm in the current sampling time period according to the abnormal state of the data in the current sampling time period and in combination with the correlation between the previous sampling time period, can reduce the interference of noise data, improves the accuracy of the particle set obtained in the particle filtering algorithm, and thereby improves the accuracy of the SOC estimation of the energy storage battery. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 It is a hardware composition schematic diagram of an energy storage battery SOC estimation device based on charging and discharging current;

[0052] Figure 2 It is a flowchart of an energy storage battery SOC estimation method based on charging and discharging current provided by the present application;

[0053] Figure 3 It is a flowchart of the implementation process of step S2;

[0054] Figure 4 It is a flowchart of the implementation process of step S2-1;

[0055] Figure 5 It is a flowchart of the implementation process of step S2-2;

[0056] Figure 6 It is a flowchart of the division process of abnormal data features and normal data features;

[0057] Figure 7 It is a flowchart of the implementation process of step S4;

[0058] Figure 8 is a structural schematic diagram of a SOC estimation system for an energy storage battery based on charging and discharging currents;

[0059] Figure 9 is a structural schematic diagram of a computer readable storage medium;

[0060] Among them, 1 is a battery state data acquisition module, 2 is a data processor. DETAILED DESCRIPTION

[0061] The device to which the energy storage battery SOC estimation method (hereinafter referred to as the energy storage battery SOC estimation method) provided in this embodiment is applied is an energy storage battery SOC estimation device based on charging and discharging currents. The application object of the energy storage battery SOC estimation device is an energy storage battery, specifically a lithium ion battery, which is used to estimate the SOC of the lithium ion battery.

[0062] As shown in Figure 1 , the energy storage battery SOC estimation device includes a battery state data acquisition module 1 and a data processor 2. The battery state data acquisition module 1 and the data processor 2 are signal connected. They can be connected by wired signal lines or wirelessly connected by wireless communication modules.

[0063] The battery state data acquisition module 1 is used to acquire battery state time series data of the energy storage battery. The battery state data acquisition module 1 can be used as an external detection device of the energy storage battery, or can be integrated with the energy storage battery. It should be understood that a sampling time period is set, and the battery state data acquisition module 1 is used to acquire the battery state time series data of the energy storage battery in the sampling time period. The length of the sampling time period is set according to actual needs, for example: 10 seconds. A plurality of sampling instants are set in the sampling time period, and the time interval (i.e. the sampling period) between adjacent sampling instants is set according to actual needs, for example: 0.1 seconds. The battery state data acquisition module 1 acquires the battery state data of the energy storage battery at each sampling instant. The battery state time series data includes the battery state data at each sampling instant. In the above example, the sampling time period includes 100 sampling instants of battery state data.

[0064] The data processor 2 is used to perform data processing according to the acquired battery state time series data to realize the adjustment of the data point weight. The hardware device corresponding to the data processor 2 can be a computer host or a server device with certain data processing capability.

[0065] The data processor 2 is used to execute Figure 2 the energy storage battery SOC estimation method as shown in Figure 2 . As shown in , the energy storage battery SOC estimation method includes the following steps:

[0066] Step S1: obtaining battery state time series data of a current sampling time period and a previous sampling time period of the energy storage battery, the battery state time series data at least including current; the current being charging current or discharging current.

[0067] The embodiment is to adjust data point weight of an adjacent next sampling time period in the particle filter algorithm according to data of a previous sampling time period. Therefore, the embodiment sets two sampling time periods, i.e. the current time period and the previous sampling time period adjacent to the current time period. It should be understood that the end time of the current time period is the current time.

[0068] The battery state time series data of the current sampling time period and the previous sampling time period of the energy storage battery are obtained by the battery state data acquisition module 1. Since the SOC of the energy storage battery is estimated mainly by the change of the current, the current data of the energy storage battery needs to be collected. Therefore, the battery state time series data at least includes the current. Since other state information of the energy storage battery at each sampling time is also needed for estimating the SOC of the energy storage battery by the particle filter algorithm, other state data of the energy storage battery can also be collected according to specific needs, such as temperature, internal resistance, power, voltage, etc. of the energy storage battery. The other state data and the current data are collected according to the same sampling time and sampling time period. In addition, since the SOC of the energy storage battery can be estimated during charging or discharging, the current of the energy storage battery is charging current or discharging current. The charging current is used for estimating during charging, and the discharging current is used for estimating during discharging. Correspondingly, the battery state data acquisition module 1 includes a current sensor for detecting the current data of the energy storage battery.

[0069] Step S2: obtaining an abnormal feature sequence of each sampling time period based on the battery state time series data, the abnormal feature sequence including abnormal features of each sampling time in the sampling time period, the abnormal feature being abnormal data feature or normal data feature.

[0070] Since the battery state time series data represents the battery state of the energy storage battery at each sampling time, according to the battery state time series data of each sampling time period, the abnormal feature sequence of the energy storage battery in each sampling time period can be obtained. For any one of the current sampling time period and the previous sampling time period, the abnormal feature sequence includes the abnormal features of each sampling time in the sampling time period, and the abnormal feature represents the abnormal condition of the energy storage battery at the corresponding sampling time. Since the abnormal condition includes the abnormal state and the normal state, the abnormal feature is an abnormal data feature or a normal data feature. The abnormal data feature represents that the energy storage battery is in a relatively abnormal state at the corresponding sampling time, and the normal data feature represents that the energy storage battery is in a relatively normal state at the corresponding sampling time.

[0071] Step S3: obtaining an error factor of the current sampling time period according to the difference between the abnormal data features in the abnormal feature sequences of the current sampling time period and the previous sampling time period, and the correlation between the abnormal data features and the normal data features in the abnormal feature sequence of the current sampling time period, the error factor being used to represent the degree of deviation of the data point weight in the current sampling time period.

[0072] When the energy storage battery is abnormal, there will be a certain difference between the abnormal feature sequences of the current sampling time period and the previous sampling time period, that is, the difference between the abnormal feature sequences of the current sampling time period and the previous sampling time period represents the abnormal state of the energy storage battery in the current sampling time period, and the obtained abnormal state will affect the adjustment of the data point weight corresponding to the current sampling time period. Moreover, the correlation between the abnormal data features and the normal data features in the abnormal feature sequence of the current sampling time period also represents the abnormal state of the energy storage battery in the current sampling time period in one aspect. Therefore, according to the difference between the abnormal data features in the abnormal feature sequences of the current sampling time period and the previous sampling time period, and the correlation between the abnormal data features and the normal data features in the abnormal feature sequence of the current sampling time period, an error factor of the current sampling time period can be obtained, which is used to represent the degree of deviation of the data point weight in the current sampling time period, or the possibility of the existence of the data point weight deviation in the current sampling time period, and the error factor will finally determine the degree of adjustment of the data point weight.

[0073] Step S4: adjusting the data point weight in the particle filter algorithm corresponding to the current sampling time period according to the difference between the isolated states of the abnormal data features in the corresponding abnormal feature sequences in the current sampling time period and the previous sampling time period, and combining the error factor.

[0074] The isolated state of the abnormal data feature in the corresponding abnormal feature sequence in any one of the current sampling time period and the previous sampling time period reflects the abnormal situation of the energy storage battery state in the corresponding sampling time period to a certain extent. According to the difference between the isolated states of the abnormal data features in the corresponding abnormal feature sequences in the current sampling time period and the previous sampling time period, and in combination with the error factor obtained in step S3, the data point weight in the particle filter algorithm corresponding to the current sampling time period can be adjusted.

[0075] In one specific embodiment, the battery state time series data in step S1 includes not only the current of the energy storage battery, but also the battery temperature and the battery internal resistance, that is, the battery state data at each sampling time includes not only the current, but also the battery temperature and the battery internal resistance. The battery temperature can be detected by a temperature sensor arranged on the energy storage battery, and the battery internal resistance can be detected by a detection circuit for detecting the battery internal resistance. In order to realize SOC estimation, the battery state data can also include the power information, voltage information, etc. of the energy storage battery.

[0076] The energy storage battery, especially the station-level energy storage battery, requires a faster charging and discharging speed, so it has a higher power density and energy density corresponding to the energy storage battery, that is, the station-level energy storage battery has a very high charging and discharging efficiency. Higher charging and discharging current can speed up the process, but may cause the battery internal resistance to rise and the heat to increase, which in turn affects the efficiency, and lower current may cause the charging and discharging speed to slow down, and the more stable the current is, the higher the efficiency is, and the lower the degree of data interference by noise is. In addition, the change of the battery temperature will affect the chemical reaction rate inside the battery, causing the current to change more dramatically, the current fluctuation to be more dramatic, and the accuracy of the collected current data to be worse. Therefore, according to the change of the current, in combination with the relationship between the changes of the battery temperature and the battery internal resistance, the abnormal features at each sampling time are obtained. In one specific embodiment, as shown in Figure 3 The implementation process of step S2 is as follows:

[0077] Step S2-1: According to the current, the current abnormal state of each sampling time period is obtained.

[0078] Step S2-2: According to the current abnormal state, and the battery temperature and the battery internal resistance, the abnormal feature sequence of each sampling time period is obtained.

[0079] In one specific embodiment, as shown in Figure 4 The implementation process of step S2-1 is as follows:

[0080] Step S2-1-1: The current difference between each sampling time and its adjacent two sampling times in the sampling time period is obtained respectively to obtain a first current difference and a second current difference.

[0081] For the convenience of description, the current sampling time period and any one of the previous sampling time period is referred to as the target sampling time period. Any one of the sampling time in the target sampling time period is referred to as the target sampling time.

[0082] In normal circumstances, the current data of each sampling time is similar to the change difference of the current data of the adjacent sampling time before and after it. The greater the change difference of the current data of each sampling time and the current data of the adjacent sampling time before and after it, the more abnormal the current data of this moment. The analysis here is for the change difference of the current caused by the change of the state of different charging and discharging periods. For this case, it will be considered to be caused by noise, and ultimately a lower weight will be given in the resampling process to eliminate it. However, the state change result is not caused by noise, so a larger weight should be given to retain it.

[0083] The current difference between the target sampling time in the target sampling time period and the adjacent previous sampling time is set as the first current difference, and the current difference between the target sampling time in the target sampling time period and the adjacent next sampling time is set as the second current difference. In this embodiment, the first current difference is the difference between the current of the target sampling time and the current of the adjacent previous sampling time, and the second current difference is the difference between the current of the adjacent next sampling time and the current of the target sampling time. Therefore, the current difference is the difference between the current of the next sampling time and the current of the previous sampling time in the adjacent two sampling times, so the current difference will have positive and negative. Therefore, the current difference can reflect the change trend and degree of the adjacent two sampling times. If the current difference is positive and the value is larger, it means that the current of the next sampling time is greater than the current of the previous sampling time, and the difference between them is larger. If the current difference is negative and the value is larger, it means that the current of the next sampling time is less than the current of the previous sampling time, and the difference between them is larger.

[0084] Step S2-1-2: Obtain the absolute value of the difference between the first current difference and the second current difference to obtain the first current abnormality degree of each sampling time.

[0085] The absolute value of the difference between the first current difference and the second current difference is taken as the first current abnormality degree of the target sampling time, so as to obtain the first current abnormality degree of each sampling time. The greater the first current abnormality degree, the greater the possibility of abnormality of the current of the target sampling time.

[0086] Step S2-1-3: Obtain the average value of the absolute value of the first current difference in the sampling time period to obtain the second current abnormality degree.

[0087] Since the target sampling time period includes multiple sampling time points, it also includes multiple first current differences. If the target sampling time period includes N sampling time points, it includes N-1 first current differences. The absolute values of the first current differences of the target sampling time period are obtained to remove the positive and negative values. Then, the average value of the absolute values of the first current differences of the target sampling time period is calculated to obtain the second current abnormality degree. The second current abnormality degree is derived from the overall characteristics of the target sampling time period. The second current abnormality degree is obtained because the faster the current changes in the target sampling time period, the greater the possibility of abnormality in the state. Therefore, the stability and abnormality of the current are further quantified by the change difference of the current data of all adjacent time points. The greater the second current abnormality degree, the worse the stability of the current in the target sampling time period, and the greater the possibility of abnormality; otherwise, the smaller the possibility of abnormality. The current characteristics of each sampling time point are reflected by representing the overall current characteristics of the target sampling time period.

[0088] Step S2-1-4: According to the first current abnormality degree and the second current abnormality degree, the current abnormality state of each sampling time point in the sampling time period is obtained.

[0089] The calculation formula of the current abnormality state of the target sampling time point in the target sampling time period is as follows:

[0090]

[0091] wherein, represents the current abnormality state of the i-th sampling time point in the target sampling time period, represents the first current abnormality degree of the i-th sampling time point in the target sampling time period, represents the second current abnormality degree of the target sampling time period.

[0092] By using the above process, the current abnormality state of each sampling time point in the current sampling time period and the previous sampling time period is obtained.

[0093] In one specific embodiment, as shown in Figure 5 , the implementation process of step S2-2 is as follows:

[0094] Step S2-2-1: According to the battery temperature, the battery temperature fluctuation degree and the battery temperature range of the sampling time period are obtained.

[0095] The temperature rise of the energy storage battery can reduce the battery internal resistance, and increase the current. However, if the temperature is too high and quickly reaches a certain high temperature, instability can be caused, leading to current fluctuation. Therefore, the abnormal fluctuation of the current at this time is self-fluctuation rather than caused by noise interference, and thus the abnormality of the state information at each sampling time needs to be analyzed and corrected according to the change trend of the internal resistance with temperature.

[0096] The target sampling time period includes the battery temperature at each sampling time. The battery temperature fluctuation degree of the target sampling time period is obtained. The battery temperature fluctuation degree can be represented by a temperature variance. Therefore, in this embodiment, the variance of the battery temperature in the target sampling time period is obtained, and the obtained variance is taken as the battery temperature fluctuation degree of the target sampling time period. The maximum battery temperature and the minimum battery temperature in the target sampling time period are obtained, and the difference between the maximum battery temperature and the minimum battery temperature is calculated to obtain the battery temperature range of the target sampling time period.

[0097] The greater the battery temperature fluctuation degree, the more intense the temperature fluctuation. The greater the battery temperature range of the target sampling time period, the more intense the temperature change, and the more abnormal the battery state data of the target sampling time period.

[0098] Step S2-2-2: According to the battery temperature and the battery internal resistance, the change correlation of the battery temperature and the battery internal resistance in the sampling time period is obtained.

[0099] Since the battery temperature affects the internal resistance only when it reaches a certain level, and since the change of the battery internal resistance is related not only to the temperature but also to the characteristics of the battery material, the sudden change of the internal resistance when the temperature reaches a certain high temperature can reduce the accuracy of the estimation of the energy storage battery. That is, the stronger the change correlation between the temperature and the internal resistance, the higher the accuracy of the estimation of the energy storage battery based on the state information data of the temperature, and the weaker the change correlation, the greater the error in the estimation process of the energy storage battery. Accordingly, the change correlation of the battery temperature and the battery internal resistance in the target sampling time period is obtained. As a specific implementation, the battery temperature at each time in the target sampling time period is normalized, and the battery temperature sequence is obtained according to the normalized battery temperature at each time. Similarly, the battery internal resistance at each time in the target sampling time period is normalized, and the battery internal resistance sequence is obtained according to the normalized battery internal resistance at each time. The correlation of the battery temperature sequence and the battery internal resistance sequence is obtained. As a specific implementation, the Pearson correlation coefficient of the battery temperature sequence and the battery internal resistance sequence is obtained. The smaller the Pearson correlation coefficient, the greater the error in the estimation process of the energy storage battery SOC, and the more likely it is to be abnormal. As another implementation, the correlation can also be calculated by the cosine similarity.

[0100] Step S2-2-3: obtaining an abnormal feature sequence of the sampling time period according to the current abnormal state, the battery temperature fluctuation degree, the battery temperature range and the change correlation; wherein the current abnormal state, the battery temperature fluctuation degree and the battery temperature range are positively correlated with the abnormal feature of the abnormal feature sequence, and the change correlation is negatively correlated with the abnormal feature.

[0101] According to the current abnormal state, the battery temperature fluctuation degree, the battery temperature range and the change correlation obtained by the above steps S-2-1 and S-2-2, an abnormal feature sequence of the target sampling time period is obtained. The abnormal feature sequence is composed of abnormal features of each sampling time point in the target sampling time period. Moreover, it is known from the above steps S-2-1 and S-2-2 that the current abnormal state, the battery temperature fluctuation degree and the battery temperature range are positively correlated with the abnormal feature of the abnormal feature sequence, and the change correlation is negatively correlated with the abnormal feature.

[0102] As a specific embodiment, a specific quantification method of the abnormal feature is as follows:

[0103]

[0104] Wherein, represents the abnormal feature of the i-th sampling time point in the target sampling time period, represents the result of the battery temperature fluctuation degree after normalization, represents the result of the battery temperature range after normalization, represents the change correlation, and exp represents the exponential function with the natural constant e as the base, represents the normalization function. It should be understood that the normalization method in the present embodiment can be the maximum-minimum normalization method, or the following normalization method: wherein y is the normalized value and x is the input quantity.

[0105] By using the above method, the abnormal features of each sampling time point in the current sampling time period and the previous sampling time period are obtained. The abnormal feature represents the abnormal degree of the battery state data at the corresponding sampling time point. The larger the abnormal feature, the more abnormal it is, and the smaller the abnormal feature, the more normal it is. Thus, the abnormal feature sequence of the current sampling time period and the abnormal feature sequence of the previous sampling time period are obtained.

[0106] In a specific embodiment, the energy storage battery SOC estimation method further includes a division process of abnormal data features and normal data features, as shown in Figure 6 The specific process is as follows:

[0107] Step S2-3: clustering the abnormal features in the abnormal feature sequence to obtain a plurality of clusters. Wherein, a two-dimensional coordinate system is constructed with the abnormal features of each sampling time in the target sampling time period as the vertical axis and the sampling time as the horizontal axis, and based on the two-dimensional coordinate system, an abnormal feature space of the target sampling time period can be constructed, and the abnormal features of all sampling times in the target sampling time period are mapped in the abnormal feature space. The present embodiment can use existing clustering methods, such as DBSCAN clustering method based on density or K-Means clustering algorithm. If K-Means clustering algorithm is used, the number of clusters can be manually set or obtained by elbow method. DBSCAN clustering method, K-Means clustering algorithm and elbow method are all known technologies and will not be described in detail. As another embodiment, a plurality of threshold values of different sizes can be set, and the abnormal features are divided into a plurality of intervals according to the threshold values to realize the division into a plurality of clusters. It should be understood that no matter which clustering method is used, the abnormal features in the abnormal feature sequence need to be divided into at least two clusters.

[0108] Thus, a plurality of clusters of the abnormal feature sequence of the current sampling time period and a plurality of clusters of the abnormal feature sequence of the previous sampling time period are obtained.

[0109] Step S2-4: based on the overall numerical level of the abnormal features in each cluster, the plurality of clusters are divided into an abnormal cluster and a normal cluster, the abnormal features in the abnormal cluster are the abnormal data features, and the abnormal features in the normal cluster are the normal data features.

[0110] Since there is a monotonic relationship between the abnormal degree represented by the abnormal feature and the numerical value of the abnormal feature, the larger the numerical value, the higher the abnormal degree. Therefore, the overall numerical level of the abnormal features in each cluster is obtained, wherein the overall numerical level can be represented by the average value of the abnormal features in the cluster. According to the overall numerical level of the abnormal features in each cluster, the plurality of clusters are divided into an abnormal cluster and a normal cluster. As a specific embodiment, the cluster corresponding to the largest overall numerical level is determined as the abnormal cluster, and each cluster other than the abnormal cluster is determined as the normal cluster. Then, the abnormal features in the abnormal cluster are the abnormal data features, and the abnormal features in the normal cluster are the normal data features.

[0111] In the above manner, the abnormal cluster and the normal cluster of the abnormal feature sequence of the current sampling time period and the abnormal cluster and the normal cluster of the abnormal feature sequence of the previous sampling time period are obtained.

[0112] Since the battery power data at all sampling time points in the current sampling time period is obtained by predicting and analyzing the battery power data at all sampling time points in the previous sampling time period, the predicted and observed power data at each sampling time point is obtained by the difference between the predicted power data and the actual observed power data, the weight of each particle is obtained by the difference between the actual observed power data and the predicted and observed power data of each particle, and the new particle set is obtained by resampling according to the weight of the particle. Therefore, when the difference between the abnormal feature distribution of the battery state data in the current sampling time period and the previous sampling time period is greater, the accuracy of obtaining the particle set in the current sampling time period from the particle set in the previous sampling time period is lower, which reduces the accuracy of the subsequent energy storage battery SOC estimation.

[0113] In step S3, the greater the distribution difference of the abnormal class cluster in the current sampling time period and the previous sampling time period, the more the process of changing the energy storage battery state in the two sampling time periods, and thus when resampling the battery state data in the current sampling time period from the battery state data in the previous sampling time period, the battery state data with large changes in the current sampling time period is discarded, resulting in certain errors.

[0114] As a specific embodiment, in step S3, the difference between the abnormal data features in the abnormal feature sequence in the current sampling time period and the previous sampling time period is the number proportion difference, and the number proportion difference is obtained as follows: obtaining the number proportion of the abnormal data features in the abnormal feature sequence in the current sampling time period, and obtaining the number proportion of the abnormal data features in the abnormal feature sequence in the previous sampling time period, then calculating the difference between the two number proportions to obtain the number proportion difference, and a specific calculation formula is provided as follows:

[0115]

[0116] wherein, represents the number proportion difference, represents the total number of sampling time points in the current sampling time period, which is the same as the total number of sampling time points in the previous sampling time period; represents the number of abnormal data features in the abnormal class cluster in the previous sampling time period, represents the number of abnormal data features in the abnormal class cluster in the current sampling time period; represents the number proportion difference, and the greater the number proportion difference, the greater the difference between the particle state information in the current sampling time period and the previous sampling time period, that is, there is a greater error in estimating the next particle set from the previous particle set, which indicates a greater possibility of error.

[0117] In step S3, the correlation between the abnormal data features and the normal data features in the abnormal feature sequence of the current sampling time period is specifically a centroid distance overall level. The centroid distance overall level is obtained as follows: the distances between the centroids of each normal cluster of the current sampling time period and the centroid of the abnormal cluster of the current sampling time period are obtained, and then the average of the obtained distances is calculated, which is the centroid distance overall level. Therefore, the centroid distance overall level represents the correlation between each normal cluster of the current sampling time period and the abnormal cluster of the current sampling time period. The higher the centroid distance overall level, the less correlated each normal cluster of the current sampling time period is with the abnormal cluster of the current sampling time period. The greater the value of the centroid distance overall level, the greater the possibility of error.

[0118] The error factor of the current sampling time period in step S3 is obtained from the quantity proportion difference and the centroid distance overall level obtained above, and the error factor is positively correlated with the quantity proportion difference and the centroid distance overall level. A specific calculation formula is provided as follows:

[0119]

[0120] wherein, represents the error factor of the current sampling time period, i.e., the error factor between the current sampling time period and the previous sampling time period, represents the centroid distance overall level corresponding to the current sampling time period.

[0121] In one specific embodiment, as shown in FIG. 4, the implementation process of step S4 is as follows: Figure 7

[0122] Step S4-1: obtaining the isolated state of the abnormal cluster of the current sampling time period in the abnormal feature sequence of the current sampling time period to obtain a first isolated state, and obtaining the isolated state of the abnormal cluster of the previous sampling time period in the abnormal feature sequence of the previous sampling time period to obtain a second isolated state.

[0123] wherein, the isolated state can be explained as the deviation of the position of the abnormal cluster in the corresponding abnormal feature sequence from the central position of the corresponding abnormal feature sequence. The greater the deviation, the more isolated the abnormal cluster is from the central position of the abnormal feature sequence.

[0124] ​In this embodiment, the first isolation state acquisition process includes: obtaining the distance between the center of the abnormal feature sequence in the current sampling time period and the center of the abnormal cluster in the current sampling time period as the first abnormal distance, and obtaining the distance between the center of each normal cluster in the current sampling time period and the center of the abnormal feature sequence in the current sampling time period as each first normal distance, calculating the sum of the first abnormal distance and all first normal distances to obtain the first total distance. The ratio of the first abnormal distance and the first total distance is taken as the first isolation state. It should be understood that the center of the abnormal cluster, the center of the normal cluster, the center of the abnormal feature sequence (i.e. the center of the abnormal feature space), and the distance between the center and the center are obtained by mapping to a two-dimensional coordinate system.

[0125] Similarly, the second isolation state acquisition process includes: obtaining the distance between the center of the abnormal feature sequence in the previous sampling time period and the center of the abnormal cluster in the previous sampling time period as the second abnormal distance, and obtaining the distance between the center of each normal cluster in the previous sampling time period and the center of the abnormal feature sequence in the previous sampling time period as each second normal distance, calculating the sum of the second abnormal distance and all second normal distances to obtain the second total distance. The ratio of the second abnormal distance and the second total distance is taken as the second isolation state.

[0126] Step S4-2: comparing the first isolation state and the second isolation state.

[0127] Step S4-3: adjusting the data point weight in the particle filter algorithm corresponding to the current sampling time period according to the comparison result of the first isolation state and the second isolation state, in combination with the error factor.

[0128] Specifically, if the first isolation state is less than the second isolation state, it indicates that the abnormal cluster in the current sampling time period is closer to the center of all data points of the abnormal feature sequence compared with the previous sampling time period (it is more likely that the data in the abnormal cluster in the current sampling time period is abnormal data caused by state change, at this time, the data in the abnormal cluster is not caused by noise, so it needs to be preserved), at this time, the data points in the abnormal cluster can be considered as noise and preserved through the particle filter algorithm, therefore the distance between the abnormal data points in the current sampling time period and the center should be reduced, i.e. the weight of the data points in the abnormal cluster should be increased to preserve these abnormal points; at the same time, the weight of the data points in the normal cluster should be reduced to preserve the data in these normal clusters. Then, the adjusted data point weight of the sampling time corresponding to the abnormal data feature in the particle filter algorithm corresponding to the current sampling time period is as follows:

[0129]

[0130] wherein, is the initial data point weight corresponding to the rth abnormal data feature in the current sampling time period (i.e. obtained through the particle filtering algorithm), is the adjusted data point weight corresponding to the rth abnormal data feature in the current sampling time period, and G is the normalized value of the error factor of the current sampling time period, i.e. the value calculated in the above .

[0131] The adjusted data point weight corresponding to the sampling time of the normal data feature in the particle filtering algorithm in the current sampling time period is as follows:

[0132]

[0133] wherein, is the initial data point weight corresponding to the st normal data feature in the current sampling time period, is the adjusted data point weight corresponding to the st normal data feature in the current sampling time period.

[0134] Conversely, if the first isolated state is greater than the second isolated state, it indicates that the abnormal cluster in the current sampling time period is farther away from the center point of all data points of the abnormal feature sequence compared with the previous sampling time period (indicating that the data in the abnormal cluster in the current sampling time period is less likely to be abnormal data caused by state change, and at this time, the data in the abnormal cluster is very likely to be caused by noise, and therefore does not need to be retained), the distance between the abnormal data point and the center point in the current sampling time period should be increased, i.e. the weight of the data point in the abnormal cluster should be reduced, so as to discard these abnormal points. Then, the adjusted data point weight corresponding to the sampling time of the abnormal data feature in the particle filtering algorithm in the current sampling time period is as follows:

[0135]

[0136] The adjusted data point weight corresponding to the sampling time of the normal data feature in the particle filtering algorithm in the current sampling time period is as follows:

[0137]

[0138] It should be understood that if the first isolated state is equal to the second isolated state, the weight is not adjusted.

[0139] Through the above process, the corrected weight of each sampling time, i.e. each data point, i.e. each particle in the current sampling time period is obtained.

[0140] It should be understood that two adjacent sampling time periods are not limited to the current sampling time period and the previous sampling time period in time, and the focus of the present application is to adjust the weight of the next sampling time period according to the previous sampling time period. With the change of time, the "current sampling time period" will also become the "previous sampling time period" of the next sampling time period.

[0141] Based on the corrected weight of each particle, the SOC estimation of the energy storage battery is performed by using the example filtering algorithm, which belongs to the prior art, such as the power battery SOC estimation method based on the extended Kalman particle filtering algorithm disclosed in the authorized announcement No. CN103472398B, or the battery state of charge estimation method disclosed in the publication No. CN104573401A. In the present embodiment, the corrected observation value and the predicted value of the particle are analyzed to obtain the weight, and the weight is used to complete the SOC estimation of the energy storage battery. Specifically: the corrected weight of each particle is normalized (i.e. the sum is 1); the particle set is resampled (i.e. sampling with replacement) by using the normalized weight to obtain a new particle set. The new particle set is traversed until the particle set corresponding to the current sampling time period is reached, and the weight and the battery state information of the corresponding particle set are used to complete the SOC estimation of the energy storage battery.

[0142] The present embodiment also provides an energy storage battery SOC estimation system 700 based on the charging and discharging current, as shown in Figure 8 , which comprises a memory 701 and a processor 702, the memory 701 is used to store executable program code 703, and the processor 702 is used to call and run the program code 703 from the memory 701 to realize the energy storage battery SOC estimation method based on the charging and discharging current as shown in Figure 2 .

[0143] The present embodiment also provides a computer readable storage medium 800, as shown in Figure 9 , the computer readable storage medium 800 stores computer program code 801, when the computer program code 801 runs on the computer, the computer executes the energy storage battery SOC estimation method based on the charging and discharging current as shown in Figure 2 .

[0144] The above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for estimating the state of charge (SOC) of an energy storage battery based on charge / discharge current, characterized in that, The application relates to a method for adjusting data point weight in a particle filter algorithm, comprising the following steps: obtaining battery state time series data of a current sampling time period and a previous sampling time period, wherein the battery state time series data at least comprises current, and the current is charging current or discharging current; obtaining abnormal feature sequences of each sampling time period based on the battery state time series data, wherein the abnormal feature sequences comprise abnormal features of each sampling time point in the sampling time period, and the abnormal features in the abnormal feature sequences are clustered to obtain a plurality of clusters; based on the overall numerical level of the abnormal features in each cluster, the plurality of clusters are divided into an abnormal cluster and a normal cluster, the abnormal features in the abnormal cluster are abnormal data features, and the abnormal features in the normal cluster are normal data features; obtaining an error factor of the current sampling time period according to the difference between the abnormal data features in the abnormal feature sequences of the current sampling time period and the previous sampling time period and the correlation between the abnormal data features and the normal data features in the abnormal feature sequences of the current sampling time period, wherein the error factor is used to represent the weight deviation degree of the data points in the current sampling time period; wherein the difference between the abnormal data features in the abnormal feature sequences of the current sampling time period and the previous sampling time period is specifically a quantity proportion difference; and the correlation between the abnormal data features and the normal data features in the abnormal feature sequences of the current sampling time period is specifically a centroid distance overall level; adjusting the data point weight corresponding to the particle filter algorithm in the current sampling time period according to the difference between the isolated states of the abnormal data features in the current sampling time period and the previous sampling time period in the corresponding abnormal feature sequences and in combination with the error factor, and the adjustment specifically comprises the following steps: obtaining the isolated state of the abnormal cluster in the current sampling time period in the abnormal feature sequence of the current sampling time period to obtain a first isolated state, and obtaining the isolated state of the abnormal cluster in the previous sampling time period in the abnormal feature sequence of the previous sampling time period to obtain a second isolated state; comparing the first isolated state and the second isolated state; adjusting the data point weight corresponding to the particle filter algorithm in the current sampling time period according to the comparison result of the first isolated state and the second isolated state and in combination with the error factor.

2. The method of claim 1, wherein the SOC of the energy storage battery is estimated based on the charging and discharging current. The battery state time series data further comprises battery temperature and battery internal resistance; obtaining the abnormal feature sequences of each sampling time period based on the battery state time series data, comprising the following steps: obtaining current abnormal states of each sampling time period according to the current; obtaining the abnormal feature sequences of each sampling time period according to the current abnormal states, the battery temperature and the battery internal resistance.

3. The method of claim 2, wherein the SOC of the energy storage battery is estimated based on the charging and discharging current. The process for obtaining the current abnormal states of each sampling time period comprises the following steps: obtaining the current difference between each sampling time point and its adjacent two sampling time points in the sampling time period to obtain a first current difference and a second current difference; obtaining the absolute value of the difference between the first current difference and the second current difference to obtain the first current abnormal degree of each sampling time point; obtaining the average value of the absolute value of the first current difference in the sampling time period to obtain a second current abnormal degree; obtaining the current abnormal states of each sampling time point in the sampling time period according to the first current abnormal degree and the second current abnormal degree.

4. The method of claim 2, wherein the SOC of the energy storage battery is estimated based on the charging and discharging current. According to the current abnormal state, the battery temperature, and the battery internal resistance, an abnormal feature sequence of each sampling time period is obtained, including: According to the battery temperature, a battery temperature fluctuation degree and a battery temperature range of the sampling time period are obtained; According to the battery temperature and the battery internal resistance, a change correlation of the battery temperature and the battery internal resistance of the sampling time period is obtained; According to the current abnormal state, the battery temperature fluctuation degree, the battery temperature range, and the change correlation, an abnormal feature sequence of the sampling time period is obtained; wherein the current abnormal state, the battery temperature fluctuation degree, and the battery temperature range are positively correlated with the abnormal feature sequence, and the change correlation is negatively correlated with the abnormal feature sequence.

5. The method of claim 1, wherein, The number proportion difference is a difference between a number proportion of abnormal data features in the abnormal feature sequence of the current sampling time period and a number proportion of abnormal data features in the abnormal feature sequence of the previous sampling time period; The centroid distance overall level is an average value of distances between the centroid of each normal cluster of the current sampling time period and the centroid of the abnormal cluster of the current sampling time period; The error factor of the current sampling time period is obtained according to the number proportion difference and the centroid distance overall level, wherein the error factor and the number proportion difference are positively correlated with the centroid distance overall level. 6.The method of claim 1, wherein, The first isolation state is obtained by the following process: The distance between the centroid of the abnormal cluster of the current sampling time period and the center point of the abnormal feature sequence of the current sampling time period is obtained as a first abnormal distance, and the distance between the centroid of each normal cluster of the current sampling time period and the center point of the abnormal feature sequence of the current sampling time period is obtained as each first normal distance, and the sum of the first abnormal distance and all first normal distances is calculated to obtain a first total distance; The ratio of the first abnormal distance to the first total distance is taken as the first isolation state; The second isolation state is obtained by the following process: The distance between the centroid of the abnormal cluster of the previous sampling time period and the center point of the abnormal feature sequence of the previous sampling time period is obtained as a second abnormal distance, and the distance between the centroid of each normal cluster of the previous sampling time period and the center point of the abnormal feature sequence of the previous sampling time period is obtained as each second normal distance, and the sum of the second abnormal distance and all second normal distances is calculated to obtain a second total distance; The ratio of the second abnormal distance to the second total distance is taken as the second isolation state.

7. The method according to claim 6, wherein, According to the comparison result of the first isolation state and the second isolation state, and in combination with the error factor, the data point weight of the data point corresponding to the current sampling time period in the particle filtering algorithm is adjusted, including: If the first isolation state is less than the second isolation state, the adjusted data point weight of the sampling time corresponding to the abnormal data feature corresponding to the current sampling time period in the particle filtering algorithm is as follows: The adjusted data point weight of the sampling time corresponding to the normal data feature corresponding to the current sampling time period in the particle filtering algorithm is as follows: If the first isolated state is greater than the second isolated state, then the adjusted data point weight for the current sampling time period corresponding to a sampling time instant corresponding to an abnormal data feature in the particle filter algorithm is as follows: If the first isolated state is greater than the second isolated state, then the adjusted data point weight for the current sampling time period corresponding to a sampling time instant corresponding to an abnormal data feature in the particle filter algorithm is as follows: wherein, is an initial data point weight corresponding to the rth abnormal data feature in the current sampling time period, is an adjusted data point weight corresponding to the rth abnormal data feature in the current sampling time period, is an initial data point weight corresponding to the sth normal data feature in the current sampling time period, is an adjusted data point weight corresponding to the sth normal data feature in the current sampling time period, and G is a normalized value of an error factor of the current sampling time period.

8. A charge-discharge current-based energy storage battery SOC estimation system, characterized by, The SOC estimation method based on charge-discharge current of the energy storage battery according to any one of claims 1-7 is implemented by a processor calling and running executable program code stored in a memory.

Images