KR102850998B1 - Smart battery management system for fire prevention - Google Patents

Abstract

A smart battery management system is provided. According to one embodiment of the present invention, a smart battery management system (S-BMS) for integrated management of one or more batteries comprises: a battery state estimation module for estimating a battery state (SOC) using a hybrid Kalman filter (HKF), and predicting R and C values of a battery, which are state variables input to the hybrid Kalman filter, using a deep learning technique; a fire prevention information collection module for collecting cell temperatures of each battery, fire safety inspection item signals, and relay contact signals; and a power control module for controlling charging/discharging of a battery by synthesizing signals of the battery state estimation module and the fire prevention information collection module.

Description

Smart battery management system for fire prevention

The present invention relates to a smart battery management system for fire prevention, and more specifically, to a smart battery management system capable of preventing fire by estimating the SOC (State of Charge) of each battery with improved accuracy and synthesizing collected fire prevention information of the battery.

In systems powered by lithium-ion secondary batteries, such as electric vehicles, electric ships, and energy storage devices, the BMS ensures the safety and reliability of the secondary batteries that supply the power required by the system. In particular, maintaining the battery's state of charge (SOC) within a constant range (0% to 100%) is crucial for extending battery life. If the SOC remains too low or too high, battery deterioration progresses more rapidly than if the SOC is maintained at an intermediate level. Overdischarging can cause component deterioration, while overcharging can lead to battery overheating. Furthermore, fires occurring in energy storage devices can escalate into large-scale fires, resulting in significant economic losses.

Therefore, accurately estimating the remaining capacity of a battery is essential for battery usage efficiency, lifespan extension, performance optimization, and even fire prevention. Since battery usage conditions change significantly from moment to moment, and battery parameters also change accordingly, accurate estimation technology is required depending on the degree of battery aging.

As such a technology, "Battery state estimation method, device performing the method, and computer program (Registration No.: 10-2016715)" discloses a method for improving the accuracy of battery state estimation by combining an ampere-hour counting method and a Kalman filter (hybrid Kalman filter method), a device performing the method, and a computer program performing the method.

As described in the aforementioned literature, a conventional method for estimating battery SOC has been using the ampere-hour counting method. The ampere-hour counting method is a method for measuring battery SOC by converting actual current values into standard current values using the Peukert equation, as shown in [Mathematical Equation 1] below, and then integrating the values.

Here, SOC ₀ is the initial state of charge of the battery, Q _n is the nominal capacity of the battery, I is the current flowing through the battery in real time according to time (t), and a is the current efficiency factor (typically set to 1) calculated and deduced by the Peukert equation.

This current integration method has the advantage of being easy to implement, but it has the problem of accumulating battery SOC estimation errors due to initial value problems and sensor noise, so regular calibration, such as resetting the battery SOC when the battery is fully charged, is required. In other words, since the battery SOC is estimated by the current measured by the current sensor installed in the battery's charge/discharge path, accurate sensing of the current sensor is very important. However, the current sensor has a problem in that a current offset occurs due to causes such as performance degradation or deterioration, where the measured values of the actual current and the measured current differ, and the error range between the actual battery SOC and the estimated battery SOC increases over time due to the accumulation of this current offset.

A conventional battery condition estimation device can be modeled as an equivalent circuit that reflects the electrochemical characteristics of the battery, as shown in Fig. 1. In Fig. 1, E(t) represents the open circuit voltage (OCV), I(t) represents the current, R ₁ represents the internal resistance of the battery, R ₂ represents the polarization resistance, and C represents the polarization capacity.

By applying Kirchhoff's law and current integration method to the equivalent circuit of the battery, the following [Mathematical Equations 2] to [Mathematical Equations 5] can be obtained.

Here, S(t) is the SOC value, Q ₀ is the nominal capacity, and F[S(t)] is a function of SOC and OCV.

In order to solve the above-described problem, a method using the Kalman filter (KF) algorithm was studied, as shown in Fig. 2. The Kalman filter (KF) includes the steps of constructing a battery equivalent circuit, calculating a system equation including the SOC of the battery and the polarization capacity and resistance of the battery equivalent circuit as state variables, calculating a measurement equation including the observed SOC of the battery as an output variable, and applying the measured current, system equation, and measurement equation for the battery to the Kalman filter (KF) algorithm to estimate the SOC of the battery at a preset cycle.

Here, the system equation and the measurement equation can be expressed as [Mathematical Equation 6] and [Mathematical Equation 7] below, respectively.

Here, S(k+1) and u _c (k+1) are the SOC and the polarization capacity of the battery at the current stage, i.e., the predicted values of the SOC and the polarization capacity of the battery, respectively, and S(k) and u _c (k) are the SOC and the polarization capacity of the battery at the previous stage, respectively. In addition, I(k) represents the previous battery current, w ₁ (k) and w ₂ (k) are the noise of the system, v(k) is the voltage measurement noise, and R ₁ , R ₂ , and C are the internal resistance, polarization resistance, and polarization capacity of the battery equivalent circuit, respectively.

Meanwhile, a method for predicting battery SOC using a hybrid Kalman filter (HKF) that combines the current integration method and the Kalman filter has been studied to improve the accuracy of battery state estimation. That is, the hybrid Kalman filter (HKF) is a method for predicting SOC using a Kalman filter (KF) that applies the current integration method by including current observation noise in the measured current during the step of calculating the measurement equation, and calculates a measurement equation as shown in [Mathematical Formula 8] below.

Here, z _k is the measurement equation, S(t ₀ ) is the initial SOC value, Q _n is the nominal battery capacity, and H(k) is the current measurement noise.

In order to estimate accurate SOC using this hybrid Kalman filter (HKF), it is important to reduce the error by substituting accurate values for R ₂ and C in [Mathematical Formula 3].

In the past, in order to estimate the values of R ₂ and C, as shown in Fig. 3, there was a method of measuring V ₁ and V ₂ based on the starting point of the curve by giving a pulse input during the discharge of the battery, and calculating the values of R ₂ and C input in [Mathematical Formula 3] using this.

However, according to the above-described method, there was a problem that the results differed greatly and the error range increased depending on which point was set as the starting point of the curve, as shown in Fig. 4(a). In addition, although the R and C values change depending on environmental change factors such as aging and performance degradation due to accumulated charging and discharging of the battery, these environmental change factors are not taken into account, and the SOC is continuously estimated using the initially calculated R ₂ and C values. Therefore, there is a limitation that the reliability of the SOC prediction may decrease as the number of battery charging/discharging cycles increases. Therefore, there is an urgent need for the development of new technologies for this.

KR 10-2016715 B1

Accordingly, the present invention is a technology for improving the problems of the prior art, and aims to provide a new type of smart battery management system for fire prevention that improves the accuracy of battery SOC prediction by accurately estimating the R and C values of the battery, which are state variables input to the Hybrid Kalman Filter (HKF) algorithm, and prevents battery fire by synthesizing the collected fire prevention information of the battery.

In order to achieve the above object, a smart battery management system according to an embodiment of the present invention is a smart battery management system (S-BMS) that comprehensively manages one or more batteries, comprising: a battery state estimation module that estimates a state of charge (SOC) of a battery using a hybrid Kalman filter (HKF), and predicts R and C values of a battery, which are state variables input to the hybrid Kalman filter, using a deep learning technique; a fire prevention information collection module that collects cell temperatures of each battery, fire safety inspection item signals, and relay contact signals; and a power control module that controls charging/discharging of a battery by synthesizing signals of the battery state estimation module and the fire prevention information collection module; wherein the battery state estimation module comprises: a 1-1 step of measuring R and C values of n sampling batteries sampled from among all batteries; a 2-1 step of calculating SOC _n of each sampling battery through a hybrid Kalman filter algorithm that uses the measured R and C values as input state variables; The method is characterized in that the algorithm of the step 3-1 of finding the optimal approximate SOC by comparing the SOC _n of the sampling battery with the ideal SOC and the step 4-1 of determining the R, C values of the optimal approximate SOC as the R, C values of the entire battery is performed to accumulate the R, C measured values of the sampling battery and the calculated optimal R, C values to obtain big data, and the deep learning algorithm is performed using the big data, and the SOC of each battery is predicted by performing the hybrid Kalman filter algorithm using the optimal R, C values and the input state variables through the deep learning algorithm. (Ideal SOC: the optimal value calculated by the battery design value or the optimal value estimated by the experiment)

delete

The battery state estimation module of the smart battery management system is characterized in that it performs the following algorithms: 1-2 steps of measuring R, C values of one sampling battery sampled from among all batteries; 2-2 steps of calculating SOC m of the sampling battery through a hybrid Kalman filter algorithm that uses the measured R, C values and R _m , C _m , which are obtained by adding or subtracting micro-incremental values ΔR, ΔC _m to the measured R, C values, as input state variables; 3-2 steps of comparing the SOC _m of the sampling battery with an ideal SOC to find an optimal approximate SOC; and 4-2 steps of determining the R, C values of the optimal approximate SOC as the R, C values of the entire battery; thereby accumulating optimal R, C values for the sampling batteries to obtain big data, and using the big data to perform a deep learning algorithm, thereby predicting the SOC of each battery that has performed the hybrid Kalman filter algorithm that uses the optimal R, C values and the hybrid Kalman filter algorithm that uses the optimal R, C values as input state variables through the deep learning algorithm. (Ideal SOC: Optimal value calculated by battery design value or optimal value estimated by experiment)

The battery status estimation module of this smart battery management system is characterized in that the battery sampling time is the initial stage of battery installation.

The battery state estimation module of the smart battery management system is characterized in that battery sampling is performed periodically and repeatedly after battery installation to obtain optimal R, C values according to battery usage time as aging big data, and a deep learning algorithm is performed using the aging big data, and the SOC of each battery is predicted by performing a hybrid Kalman filter algorithm using the optimal R, C values that change according to battery usage through the deep learning algorithm and the hybrid Kalman filter algorithm that uses these as input state variables.

The fire safety inspection item signal of the smart battery management system is characterized in that it includes at least one of a power signal, a ground fault detection and tripping (GPT) signal, a power conversion device (PCS) signal, a surge protection device (SPD) signal, an insulation monitoring device (IMD) signal, and an overcurrent tripping (fuse) signal.

The power control module of this smart battery management system is characterized in that it blocks the connection to each battery when it detects a fire risk in the battery.

According to the smart battery management system for fire prevention according to the present invention, the R and C values of the battery, which are state variables input when applying the Kalman filter algorithm, are predicted through deep learning, so that the accuracy of estimating the R and C values of the battery, which are state variables, can be improved, and accordingly, the error range of the SOC prediction can be reduced, thereby improving the accuracy of the SOC prediction.

Therefore, it is possible to prevent and block problems in advance, such as continuous overcharging or overdischarging of the battery due to SOC prediction errors, and overheating due to accumulated stress in the battery leading to fire.

It has the effect of collecting various signals that may cause a fire in the battery and synthesizing them with battery SOC prediction information to prevent battery fire.

Additionally, the smart battery management system according to the present invention has the advantage of being able to monitor the status of battery current, voltage, temperature, etc. in real time to predict the remaining capacity and replacement time, and utilize this for system control, and secure the battery life through battery performance equalization control, and reduce maintenance costs by securing system reliability and safety through battery protection and self-diagnosis.

Figure 1 is an equivalent circuit diagram reflecting the electrochemical characteristics of a battery.
Figure 2 is a flowchart of a conventional Kalman filter (KF) algorithm.
Figure 3 is a time-voltage graph when a pulse input is given during discharge of a battery to estimate the values of R ₂ and C.
Figure 4 is a battery SOC prediction graph using a conventional hybrid Kalman filter (HKF) and a deep learning technique according to an embodiment of the present invention.
Figure 5 is a block diagram of a smart battery management system according to one embodiment of the present invention.
Figure 6 is a block diagram of a battery SOC prediction algorithm according to one embodiment of the present invention.
Figure 7 is a block diagram of a battery SOC prediction algorithm according to another embodiment of the present invention.
Figure 8 is an optimal configuration diagram for collecting fire safety inspection item data according to one embodiment of the present invention.
Figure 9 is a configuration diagram of a smart battery management system according to one embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings, FIGS. 1 to 9. Meanwhile, in the drawings and detailed descriptions, the illustrations and references to the configurations and operations that can be easily understood by those skilled in the art from general BMS, SOC, Kalman filters, etc., are briefly or omitted. In particular, in the illustrations and detailed descriptions of the drawings, detailed descriptions and illustrations of specific technical configurations and operations of elements not directly related to the technical features of the present invention are omitted, and only the technical configurations related to the present invention are briefly illustrated or described.

A smart battery management system for fire prevention according to an embodiment of the present invention is configured to include a battery status estimation module (100), a fire prevention information collection module (200), and a power control module (300), as shown in FIG. 5.

The battery state estimation module (100) predicts the battery SOC.

According to the above-described "Battery State Estimation Method, Device and Computer Program Performing the Method (Registration Number: 10-2016715)", as shown in Fig. 4(a), there was a problem in that the results differed significantly and the error range increased depending on which point was selected as the starting point of the curve, and environmental change factors such as aging and performance degradation due to accumulated charging and discharging of the battery must be taken into consideration.

Therefore, in the embodiment of the present invention, by using a deep learning technique that takes into account the number of charge and discharge cycles and the environmental changes of the battery accordingly, V ₁ and V ₂ are measured at various points in FIG. 3, and the values of R ₂ and C are calculated accordingly to select the most appropriate values, thereby reducing the error range of SOC prediction, as shown in FIG. 4(b).

That is, the battery SOC prediction algorithm according to the embodiment of the present invention can be represented as shown in Fig. 6. The R and C values of n sampling batteries sampled from among all batteries are measured, and the SOC _n of each sampling battery is calculated through a hybrid Kalman filter algorithm that uses the measured R and C values as input state variables. The SOC _n thus calculated is compared with the ideal SOC selected as the optimal value calculated by the battery design value or the optimal value estimated by experiment, and the optimal approximate SOC is found, and the R and C values of the optimal approximate SOC are set as the R and C values of the entire battery.

At this time, big data can be acquired by accumulating the R, C measured values of the sampled batteries and the calculated optimal R, C values. Using the acquired big data, a deep learning algorithm is performed, and by performing the hybrid Kalman filter algorithm with the optimal R, C values and these as input state variables, it becomes possible to predict each SOC for the entire battery to be predicted.

By utilizing this battery SOC prediction technique, more precise predictions are possible, and the optimal values of the input state variables are automatically extracted, thereby improving user convenience.

A battery SOC prediction algorithm according to another embodiment of the present invention can be represented as shown in Fig. 7. First, a hybrid Kalman filter algorithm is performed using the measured initial R, C values of one battery sampled from among all batteries as input state variables to calculate SOC ₁ .

Next, the values obtained by adding and subtracting the micro-incremental values ΔR and ΔC m times from the measured R ₁ and C ₁ are set as the predicted R _m and C _m values, and a hybrid Kalman filter algorithm that uses numerous predicted R _m and C _m values as input state variables is performed to calculate the SOC _m for each predicted R _m and C _m value, and the calculated SOC _m is compared with the ideal SOC to find the optimal approximate SOC, and the R and C values of the optimal approximate SOC are set as the R and C values of the entire battery.

At this time, big data can be acquired by accumulating optimal R and C values for sampled batteries. Using the acquired big data, a deep learning algorithm is performed, and by performing a hybrid Kalman filter algorithm with the optimal R and C values and these as input state variables, each SOC can be predicted for the entire battery to be predicted.

In particular, by generating a large amount of data using a single sampled actual data and converting it into big data, SOC can be predicted more simply than in the case of the embodiment as shown in FIG. 6 described above.

The battery condition estimation module (100) initially predicts the SOC of the battery at the initial stage of battery installation using the method described above, but allows battery sampling to be performed repeatedly and periodically. This is because battery performance may deteriorate or environmental changes may occur over time as the battery is used, and the R and C values may change accordingly, thereby causing the optimal SOC value to also change.

Therefore, by periodically repeating battery sampling according to the aging of the battery, the optimal R and C values according to the battery usage time can be obtained as aging big data.

When predicting each SOC of the entire battery by the method illustrated in FIG. 6 or FIG. 7, by performing a deep learning algorithm using the acquired aging big data, it is possible to calculate optimal R and C values that take into account the battery usage environment conditions that change according to battery usage, and by performing a hybrid Kalman filter algorithm that uses this as an input state variable, it is possible to predict the SOC that takes into account the aging of each battery.

A smart battery management system according to one embodiment of the present invention predicts the SOC of a battery using the above-described method through a battery state estimation module (100), and detects risk factors that may cause fire in advance through a fire prevention information collection module (200).

The fire prevention information collection module (200) is intended to intensively inspect factors with a high risk of fire in a power system, and collects the temperature of each battery cell, fire safety inspection item signals, and relay contact signals.

Among them, the fire safety inspection item signals include at least one of the following: a power signal, a ground fault detection and tripping (GPT) signal, a power conversion device (PCS) signal, a surge protection device (SPD) signal, an insulation monitoring device (IMD) signal, and an overcurrent tripping (fuse) signal, and these inspection items can be configured as shown in Fig. 7 so that signals can be collected.

As described above, it has the effect of collecting various signals that may cause a fire in a battery and synthesizing them with battery SOC prediction information to prevent a fire in the battery.

A smart battery management system according to one embodiment of the present invention is configured to predict the SOC of a battery in a battery state estimation module (100), detect risk factors that may cause fire in advance in a fire prevention information collection module (200), and further include a power control module (300) so that fire can be prevented from spreading in advance when it occurs.

The power control module (300) integrates signals from the battery status estimation module (100) and the fire prevention information collection module (200) to control the charging/discharging of each battery. In particular, when a fire risk is detected in a battery, the power control module (300) blocks the connection to the battery, thereby preventing the spread of fire.

In addition, if a monitoring system capable of monitoring the status of battery current, voltage, temperature, etc. in real time is further provided through the smart battery management system according to the present invention, the status of the battery can be monitored in real time by the manager, and the remaining capacity and replacement time of the battery can be easily predicted, and this can be utilized for power system control, or the reliability and safety of the battery management system can be secured and maintenance costs can be reduced by securing the battery life through performance balance control of the battery.

In addition, a cloud storage device or a black box may be additionally installed to analyze the cause after a fire has occurred and prevent recurrence. The cloud storage device or black box stores battery monitoring data received from the battery status estimation module (100), data on each safety inspection item received from the fire prevention information collection module (200), and battery charge/discharge and operation data received from the power control module (300) in a database, and enables post-confirmation and analysis, thereby analyzing the cause when a fire occurs and preventing recurrence of the fire.

Finally, the smart battery management system according to the present invention, which includes a battery status estimation module (100), a fire prevention information collection module (200), a power control module (300), a monitoring system, a cloud storage device, or a black box, can be configured as shown in FIG. 8.

Therefore, the smart battery management system for fire prevention of the present invention can prevent the accumulation of stress due to continuous overcharging or overdischarging of the battery and block the occurrence of battery overheating by precisely predicting the SOC by considering the aging changes of the battery through the battery condition estimation module. In addition, it has the effect of collecting various signals that may cause fire in the battery and synthesizing them with battery SOC prediction information to prevent the risk of battery fire in advance.

Although the smart battery management system for fire prevention according to the embodiment of the present invention as described above has been illustrated in accordance with the above description and drawings, this is merely an example, and those skilled in the art will readily understand that various changes and modifications are possible within the scope that does not depart from the technical spirit of the present invention.

100: Battery status estimation module
200: Fire Prevention Information Collection Module
300: Power control module

Claims (7)

In a smart battery management system (S-BMS) that manages one or more batteries,
A battery state estimation module that estimates the battery state (SOC) using a hybrid Kalman filter (HKF), and predicts the R and C values of the battery, which are state variables input to the hybrid Kalman filter, using a deep learning technique;
A fire prevention information collection module that collects the temperature of each battery cell, fire safety inspection item signal, and relay contact signal; and
It is configured to include a power control module that controls the charging/discharging of the battery by synthesizing the signals of the battery status estimation module and the fire prevention information collection module;
The above battery status estimation module,
Step 1-1 of measuring the R and C values of n sampled batteries from among all batteries;
Step 2-1 of calculating the SOC _n of each sampled battery through a hybrid Kalman filter algorithm using the measured R and C values as input state variables;
Step 3-1: Finding the optimal approximate SOC by comparing the SOC _n of the sampled battery with the ideal SOC; and
Step 4-1 of determining the R, C values of the above optimal approximate SOC as the R, C values of the entire battery is performed by performing the algorithm.
Big data is obtained by accumulating the R, C measured values of the above sampling battery and the calculated optimal R, C values,
By using the above big data, a deep learning algorithm is performed,
A smart battery management system for fire prevention characterized by predicting the SOC of each battery by performing a hybrid Kalman filter algorithm using optimal R and C values and these as input state variables through a deep learning algorithm.
(Ideal SOC: Optimal value calculated by battery design value or optimal value estimated by experiment) delete In the first paragraph,
The above battery status estimation module,
Step 1-2: measuring the R and C values of one sampled battery among all batteries;
Step 2-2 of calculating the SOC m of the sampling battery through a hybrid Kalman filter algorithm that uses the measured R, C values and R _m and C _m obtained by adding or subtracting the micro-incremental values ΔR, ΔC to the measured R, C values _m times as input state variables;
Step 3-2: Finding the optimal approximate SOC by comparing the SOC _m of the sampled battery with the ideal SOC; and
The algorithm of step 4-2 is performed to determine the R, C values of the above optimal approximate SOC as the R, C values of the entire battery.
Obtain big data by accumulating optimal R and C values for the above sampling batteries,
A smart battery management system for fire prevention, characterized in that it performs a deep learning algorithm using the above big data, and predicts the SOC of each battery by performing a hybrid Kalman filter algorithm using the optimal R and C values and these as input state variables through the deep learning algorithm.
(Ideal SOC: Optimal value calculated by battery design value or optimal value estimated by experiment) In the third paragraph,
The above battery status estimation module,
A smart battery management system for fire prevention characterized by a battery sampling time that is early in the battery installation. In paragraph 4,
The above battery status estimation module,
Battery sampling is performed periodically after battery installation.
Obtain optimal R, C values according to battery usage time using big data on aging.
A smart battery management system for fire prevention characterized by performing a deep learning algorithm using the above-mentioned aging big data, and predicting the SOC of each battery by performing a hybrid Kalman filter algorithm using the optimal R, C values that change according to battery use through the deep learning algorithm and the input state variables. In the first paragraph,
The above fire safety inspection item signal is,
A smart battery management system for fire prevention, characterized in that it includes at least one of a power signal, a ground fault detection and tripping (GPT) signal, a power conversion device (PCS) signal, a surge protection device (SPD) signal, an insulation monitoring device (IMD) signal, and an overcurrent tripping (fuse) signal. In the first paragraph,
The above power control module,
A smart battery management system for fire prevention, characterized in that when a fire risk is detected in each of the above batteries, the connection to the corresponding battery is blocked.