Method for predicting thermal runaway by applying EMS to semi-Markov model

Abstract

The invention discloses a method for predicting thermal runaway by applying an EMS to a semi-Markov model, and the method comprises the steps: S1, monitoring the state information of a single battery in real time through the EMS, and storing the real-time and historical record data of the single battery; s2, the EMS applies real-time and historical record data to a semi-Markov model to perform thermal runaway prediction of the single battery; s3, according to a thermal runaway prediction result, determining whether the EMS takes active measures or not; s4, taking active measures; and S5, further determining whether the EMS adopts active fire-fighting triggering or not according to a thermal runaway prediction result. According to the method, the EMS of the energy storage system is used for monitoring the battery information state and reading the triggering state of the fire extinguishing system, the semi-Markov prediction model is integrated in the EMS, and the potential thermal runaway event is accurately predicted, so that measures are taken in advance, and accidents can be effectively prevented.

Description

Technical Field

[0001] The present invention relates to a thermal runaway prediction method, and in particular to a method for predicting thermal runaway by applying an EMS to a semi-Markov model. Background Art

[0002] The new energy storage industry represented by lithium-ion battery energy storage has developed rapidly in recent years. However, while the lithium-ion battery energy storage industry is developing rapidly, the safety issues caused by battery thermal runaway are becoming increasingly severe. In many major safety accidents that have occurred in lithium-ion battery energy storage power stations, battery thermal runaway is usually caused by factors such as unstable electrolytes inside the battery, damage to electrode materials, and unbalanced charging and discharging, which lead to a rapid increase in the temperature inside the battery. High temperature will cause the decomposition or volatilization of electrolytes and produce flammable gases. Once the gas meets oxygen, fire or explosion may occur. Therefore, effectively predicting the occurrence of thermal runaway of lithium-ion batteries is conducive to promoting safer and more reliable large-scale application and development of energy storage power stations.

[0003] In the prior art, the standard Markov model is used as the main means for thermal runaway prediction; the standard Markov model assumes that the transition probability between states is fixed and does not change with time or the duration of the previous state. This fixity may lead to the inability to accurately capture the actual situation and transfer rate of state changes when predicting dynamic processes such as thermal runaway; at the same time, the standard Markov model does not consider the duration distribution of the state. In the prediction of thermal runaway, the state (such as the battery being in a high temperature state) may last for a period of time, and the length of this period has an important impact on the prediction of the development of the event; due to the sporadic nature of thermal runaway events, there is usually a lack of a large amount of event data for establishing an accurate Markov model. The standard Markov model has a high data volume requirement, which limits its prediction ability and reliability in the case of data scarcity. This technology is currently mainly used in the new energy vehicle industry. If it is used in energy storage power stations, it usually faces a more complex environment and higher risks than automobiles, because the battery pack is large in scale, high in current, and long in operation. Once thermal runaway occurs, it may have a serious impact on the entire power station and the surrounding environment. Summary of the invention

[0004] In order to solve the deficiencies of the above-mentioned technologies, the present invention provides a method for predicting thermal runaway by applying EMS to a semi-Markov model.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for predicting thermal runaway using an EMS semi-Markov model,

[0006] S1, EMS monitors the status information of single cells in real time and retains the real-time and historical data of single cells;

[0007] S2, EMS applies the semi-Markov model to predict thermal runaway of single cells based on real-time and historical record data;

[0008] S3, based on the thermal runaway prediction results, decide whether the EMS should take active measures;

[0009] S4, take proactive measures;

[0010] S5, further determining whether the EMS should take active firefighting triggering according to the thermal runaway prediction result.

[0011] Furthermore, the EMS in S1 monitors the status information of the single cell in real time, including the voltage, temperature, total voltage and total current of the single cell; and retains the real-time and historical record data of the single cell, including the battery information, working status, fault records and abnormal events of the single cell.

[0012] Furthermore, the process of thermal runaway prediction of the semi-Markov model in S2 includes the following steps:

[0013] S21, define the state of the single battery;

[0014] S22, calculate the state transition probability;

[0015] S23, calculate the state duration distribution;

[0016] S24, model training;

[0017] S25, model prediction results.

[0018] Furthermore, the states of the single battery defined in S21 include: normal state, thermal runaway warning state, and thermal runaway occurrence state;

[0019] Normal status means that the voltage, temperature, total voltage and total current of the battery cells are within the normal range and there is no sign of thermal runaway;

[0020] The thermal runaway warning state indicates that the voltage, temperature, total voltage and total current of the battery cell are slightly abnormal, indicating that thermal runaway is about to occur;

[0021] The thermal runaway state indicates that the battery has entered a thermal runaway state and requires immediate attention to prevent further damage.

[0022] Furthermore, the state transition probability represents the probability of transitioning from one state to another state, and the matrix for calculating the state transition probability is:

[0023]

[0024] in, The transition rate matrix represents the state transition probability of the semi-Markov process, πaa (h) represents the transition rate of the normal state, π ab (h) represents the transition rate from the normal state to the thermal runaway warning state, π bb (h) represents the transfer rate of thermal runaway warning state, π bc (h) represents the transition rate from the thermal runaway warning state to the thermal runaway occurrence state, π cc (h) represents the transition rate of the thermal runaway state.

[0025] Furthermore, the state duration distribution represents the probability density function of the duration in a certain state. The process of calculating the state duration is to assume that the state duration follows the Weibull distribution. The Weibull distribution function is:

[0026]

[0027] Where χ is a random variable, λ>0 is a scale parameter, and k>0 is a shape parameter;

[0028] Assuming that the dwell time of each mode (s, l, k) follows the Weibull distribution, the probability density function of the dwell time h in mode s is:

[0029]

[0030] According to the battery status, the normal state is a, the thermal runaway warning state is b, and the thermal runaway state is c;

[0031] When s = a, l = b, k = c,

[0032] Furthermore, model training constructs a model through data on battery thermal runaway characteristics and normal battery behavior to evaluate and verify the model.

[0033] Furthermore, in S3, based on the thermal runaway prediction (warning) results, it is decided whether the EMS takes active measures. Specifically, if the state of the single cell battery changes from the thermal runaway warning state to the thermal runaway occurrence state, the EMS records the model prediction results in real time and takes active measures; if the state of the single cell battery does not change, continue with step S2 to perform thermal runaway prediction.

[0034] Furthermore, the proactive measures taken by S4 were to shut down the EMS and notify the operation and maintenance personnel to cut off the power supply.

[0035] Furthermore, in S5, it is further determined whether the EMS takes active firefighting triggering according to the thermal runaway prediction result: if thermal runaway occurs and the EMS is not powered off, the firefighting is not triggered, the EMS takes the initiative to trigger the firefighting, and the EMS outputs the firefighting start request instruction of the current packet sequence number to the firefighting to extinguish the fire, triggering the single-packet firefighting while the EMS records the model prediction result and the firefighting triggering message in real time; if the EMS is powered off, wait for the firefighting system to take the initiative to extinguish the firefighting.

[0036] The present invention discloses a method for applying EMS to a semi-Markov model to predict thermal runaway. The energy management system (EMS) of an energy storage system is used to monitor the battery information status and read the trigger status of a fire protection system. The semi-Markov prediction model is integrated in the EMS, and potential thermal runaway events are accurately predicted, so that measures can be taken in advance to effectively prevent accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Flowchart for applying semi-Markov model to predict thermal runaway for EMS.

[0038] Figure 2 A schematic diagram of the EMS host architecture.

[0039] Figure 3 Flowchart for thermal runaway prediction using a semi-Markov model. DETAILED DESCRIPTION

[0040] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0041] A method for predicting thermal runaway using an EMS semi-Markov model, wherein, in an energy storage system, such as Figure 2 As shown, the main control equipment of the EMS host includes: BMS, PCS, liquid cooling unit, dehumidifier, electric meter and other external equipment, fire protection, serial port screen, cooling fan control, system indicator light, emergency stop, surge, etc. Figure 1 As shown, through the coordinated work of these components, intelligent monitoring and control of the energy storage warehouse can be achieved to pursue the efficient goals of energy consumption monitoring, energy consumption management and energy consumption optimization, thereby realizing efficient management and operation of energy.

[0042] like Figure 1 The EMS application of semi-Markov model to predict thermal runaway is shown in the flowchart. The steps are as follows:

[0043] S1, EMS monitors the status information of single cells in real time and retains the real-time and historical record data of single cells; EMS monitors the status information of single cells in real time, including the single cell voltage, single cell temperature, total voltage (cluster total voltage) and total current of the single cell; retains the real-time and historical record data of single cells, including the battery information, working status, fault records and abnormal events of the single cell.

[0044] S2, EMS applies the semi-Markov model to predict the thermal runaway of the single battery based on the real-time and historical record data; the process of thermal runaway prediction of the semi-Markov model includes the following steps, such as Figure 3 As shown:

[0045] S21, defining the state of the single battery, including: normal state, thermal runaway warning state, and thermal runaway occurrence state;

[0046] Normal status means that the voltage, temperature, total voltage and total current of the battery cells are within the normal range and there is no sign of thermal runaway;

[0047] The thermal runaway warning state indicates that the voltage, temperature, total voltage and total current of the battery cell are slightly abnormal, indicating that thermal runaway is about to occur;

[0048] The thermal runaway state indicates that the battery has entered a thermal runaway state and requires immediate attention to prevent further damage.

[0049] S22, calculate the state transition probability;

[0050] definition is the transition rate matrix of the semi-Markov process, indicating that state s jumps to state l, and the duration between the two jumps is h. When πsl(h)=πsl, it means that the semi-Markov process degenerates into a standard Markov process.

[0051] The state transition probability represents the probability of transitioning from one state to another. The matrix for calculating the state transition probability is:

[0052]

[0053] in, The transition rate matrix represents the state transition probability of the semi-Markov process, π aa (h) represents the transition rate of the normal state, π ab (h) represents the transition rate from the normal state to the thermal runaway warning state, π bb (h) represents the transfer rate of thermal runaway warning state, π bc (h) represents the transition rate from the thermal runaway warning state to the thermal runaway occurrence state, π cc (h) represents the transition rate of the thermal runaway state.

[0054] S23, calculate the state duration distribution;

[0055] The state duration distribution represents the probability density function of the duration in a certain state. The process of calculating the state duration is to assume that the state duration follows the Weibull distribution. The Weibull distribution function is:

[0056]

[0057] Where χ is a random variable, λ>0 is a scale parameter, and k>0 is a shape parameter;

[0058] Assuming that the dwell time of each mode (s, l, k) follows the Weibull distribution, the probability density function of the dwell time h in mode s is:

[0059]

[0060] According to the battery status, the normal state is a, the thermal runaway warning state is b, and the thermal runaway state is c;

[0061] When s = a, l = b, k = c,

[0062] S24, model training: Since thermal runaway is an occasional event, there is no large amount of background data for training in the early stage. An effective model can be built through the data of battery thermal runaway characteristics and normal battery behavior, and the model can be evaluated and verified to ensure that it can accurately predict changes in battery status. The data should be continuously updated to optimize the model and improve accuracy.

[0063] S25, model prediction results. According to the prediction algorithm of the semi-Markov model, the possible state transitions and transition times in the future are estimated.

[0064] S3, based on the thermal runaway prediction results, decide whether the EMS takes active measures; specifically, based on the thermal runaway prediction results, decide whether the EMS takes active measures: if the single cell battery state changes from the thermal runaway warning state to the thermal runaway occurrence state, the EMS records the single cell battery state prediction results in real time and immediately takes active measures in step S4; if the single cell battery state does not change, continue with step S2 to perform thermal runaway prediction.

[0065] S4, take proactive measures; the proactive measures taken are to shut down the EMS and notify the operation and maintenance personnel to cut off the power supply, stop the energy conversion and chemical reaction inside the battery, and avoid further temperature increase leading to thermal runaway accidents.

[0066] S5, further based on the thermal runaway prediction results, decide whether the EMS should take the initiative to trigger the fire. Further based on the thermal runaway prediction results, decide whether the EMS should take the initiative to trigger the fire: if the EMS is not powered off and the fire is not triggered, the EMS actively triggers the fire, and the EMS outputs the fire start request instruction of the current packet sequence number to the fire to extinguish the fire, triggering a single packet fire. At the same time, the EMS records the model prediction results and the fire trigger message in real time; if the EMS is powered off, wait for the fire system to actively extinguish the fire.

[0067] Therefore, the present invention innovatively realizes the prediction of battery thermal runaway state through the semi-Markov model, and further realizes EMS active measures when thermal runaway warning or occurrence by integrating the semi-Markov model in the EMS, thereby realizing EMS active shutdown, active warning and rapid triggering of fire fighting.

[0068] First, the semi-Markov model is able to take into account the duration of the state, so it can more accurately predict the timing and duration of thermal runaway events. This will improve the EMS's ability to prevent thermal runaway and reduce false positives and negatives.

[0069] Secondly, the semi-Markov model allows the state transition probability to change over time, adapting to the impact of different operating conditions and environmental factors on the battery state. This enables the EMS to better cope with abnormal situations and unpredictable changes.

[0070] Finally, the semi-Markov model takes into account the transition rate of events and can adjust the prediction based on the actual observed state duration, thereby more accurately reflecting the actual probability of battery state changes. With more accurate thermal runaway predictions, EMS can quickly take early warning and control measures. By monitoring and predicting battery status in real time, the occurrence of excessive battery degradation and thermal runaway events can be reduced, thereby reducing battery maintenance and replacement costs.

[0071] The above implementation modes are not limitations of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the scope of the technical solution of the present invention also belong to the protection scope of the present invention.

Claims

1. A method for predicting thermal runaway using an EMS semi-Markov model, characterized in that: S1, EMS monitors the status information of single cells in real time and retains the real-time and historical data of single cells; S2, EMS applies the semi-Markov model to predict thermal runaway of single cells based on real-time and historical record data; S3, based on the thermal runaway prediction results, decide whether the EMS should take active measures; S4, take proactive measures; S5, further determining whether the EMS should take active firefighting triggering according to the thermal runaway prediction result.

2. The method for predicting thermal runaway using an EMS using a semi-Markov model according to claim 1, characterized in that: The EMS in S1 monitors the status information of the single cells in real time, including the voltage, temperature, total voltage and total current of the single cells; the real-time and historical record data of the single cells retained include the battery information, working status, fault records and abnormal events of the single cells.

3. The method for predicting thermal runaway using an EMS using a semi-Markov model according to claim 1, characterized in that: The process of thermal runaway prediction of the semi-Markov model in S2 includes the following steps: S21, defining the state of the single cell; S22, calculate the state transition probability; S23, calculate the state duration distribution; S24, model training; S25, model prediction results.

4. The method for predicting thermal runaway using an EMS using a semi-Markov model according to claim 3, characterized in that: The states of the single battery defined in S21 include: normal state, thermal runaway warning state, and thermal runaway occurrence state; The normal state means that the voltage, temperature, total voltage and total current of the battery cell are within the normal range and there is no sign of thermal runaway; The thermal runaway warning state indicates that the voltage, temperature, total voltage and total current of the battery cell are slightly abnormal, indicating that thermal runaway is about to occur; The thermal runaway occurrence state indicates that the battery has entered a thermal runaway state and requires immediate processing to prevent further damage.

5. The method for predicting thermal runaway using an EMS using a semi-Markov model according to claim 4, characterized in that: The state transition probability represents the probability of transitioning from one state to another state. The matrix for calculating the state transition probability is: in, The transition rate matrix represents the state transition probability of the semi-Markov process, π aa (h) represents the transition rate of the normal state, π ab (h) represents the transition rate from the normal state to the thermal runaway warning state, π bb (h) represents the transfer rate of thermal runaway warning state, π bc (h) represents the transition rate from the thermal runaway warning state to the thermal runaway occurrence state, π cc (h) represents the transition rate of the thermal runaway state.

6. The method for predicting thermal runaway using an EMS using a semi-Markov model according to claim 5, characterized in that: The state duration distribution represents the probability density function of the duration in a certain state. The process of calculating the state duration is to assume that the state duration obeys the Weibull distribution. The Weibull distribution function is: Where χ is a random variable, λ>0 is a scale parameter, and k>0 is a shape parameter; Assuming that the dwell time of each mode (s, l, k) follows the Weibull distribution, the probability density function of the dwell time h in mode s is: According to the battery status, the normal state is a, the thermal runaway warning state is b, and the thermal runaway state is c; When s = a, l = b, k = c, 7. The method for predicting thermal runaway using an EMS using a semi-Markov model according to claim 3, characterized in that: The model training constructs a model through data of battery thermal runaway characteristics and normal battery behavior, and evaluates and verifies the model.

8. The method for predicting thermal runaway using an EMS using a semi-Markov model according to claim 1, characterized in that: In S3, the EMS is determined to take active measures according to the thermal runaway prediction result. Specifically, if the state of the single battery changes from the thermal runaway warning state to the thermal runaway occurrence state, the EMS records the single battery state prediction result in real time and immediately takes active measures in step S4; if the single battery state does not change, the EMS continues to step S2 to perform thermal runaway prediction.

9. The method for predicting thermal runaway using an EMS using a semi-Markov model according to claim 1, characterized in that: The active measure taken by S4 is that the EMS actively shuts down and notifies the operation and maintenance personnel to cut off the power supply.

10. The method for predicting thermal runaway using an EMS using a semi-Markov model according to claim 1, characterized in that: In S5, it is further determined whether the EMS takes the initiative to trigger the fire according to the thermal runaway prediction result: if the EMS is not powered off and the fire is not triggered, the EMS takes the initiative to trigger the fire, and the EMS outputs the fire start request instruction of the current packet sequence number to the fire to extinguish the fire, triggering the single packet fire while the EMS records the model prediction result and the fire trigger message in real time; if the EMS is powered off, wait for the fire system to take the initiative to extinguish the fire.

Images