Energy storage system gas sensor arrangement optimization method based on gas detection

Abstract

An energy storage system gas sensor arrangement optimization method based on gas detection comprises the following steps: selecting a fault point in an energy storage system, arranging a gas sensor at a monitoring point, overcharging a battery at the fault point to trigger thermal runaway to release hydrogen, and obtaining hydrogen concentration data of the monitoring point; the method comprises the following steps: carrying out node division on a monitoring area in an energy storage system, placing a gas sensor on each node, obtaining hydrogen concentration data of each node so as to obtain a measurement matrix, and carrying out time division on the hydrogen concentration data of each node so as to obtain a loss coefficient and a corresponding loss coefficient matrix; and based on the obtained loss coefficient matrix, performing expectation optimization on the monitoring area by using a minimum expectation optimization algorithm to obtain the optimal installation position and average detection time when different numbers of gas sensors are installed on the energy storage system. The method is suitable for different energy storage systems, the gas detection efficiency is remarkably improved, high-sensitivity thermal runaway early warning can be achieved, and the safety and stability of the energy storage systems can be greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of early warning of thermal runaway of batteries in energy storage systems, and in particular to a method for optimizing the arrangement of gas sensors in energy storage systems based on gas detection. Background Art

[0002] As the main force in promoting my country's dual carbon goals, the new energy industry is in urgent need of moving towards low-carbonization in terms of both energy security and energy environmental protection. As a new type of power equipment, the energy storage system is a key member of this energy green transformation action. Lithium-ion batteries have been widely used in the field of large-scale energy storage due to their advantages such as high energy density, high charging efficiency, light weight, and long cycle life, such as electric vehicles and energy storage cabins that are being rapidly promoted around the world.

[0003] However, due to the flammable and explosive active chemical properties of lithium-ion battery electrolytes, lithium batteries pose considerable safety risks during storage and use, especially under typical fault conditions such as overcharging and low temperature. In addition, the large number of batteries in electric vehicle battery packs and energy storage compartments makes it very easy for large-scale fires caused by the spread of thermal runaway of single cells to occur. In recent years, reports of electric vehicle fire accidents have emerged one after another, and according to incomplete statistics from CNESA, there have been nearly 60 grid-level energy storage fire accidents in the past five years; therefore, the optimization method of early warning technology for thermal runaway of energy storage system batteries is a key issue that needs to be urgently resolved in the current energy storage safety field.

[0004] In recent years, gas detection technology has attracted widespread attention and has been widely developed in the field of lithium battery thermal runaway warning technology. Gas detection has the advantages of high accuracy, fast response speed, and good economy in the application of lithium battery thermal runaway warning.

[0005] Thermal runaway of lithium batteries usually produces a large amount of side reaction gases, most of which are extremely low in air; the gas diffusion is generally wide and fast. So far, the effectiveness of the warning of characteristic gases in the energy storage compartment is still unclear. Lithium-ion batteries are usually assembled in battery clusters after packing. In addition, the internal space of the energy storage compartment is huge. The impact of these factors on the diffusion of characteristic gases is still unclear. In addition, there is no theoretical guidance on the installation strategy of gas sensors in the energy storage compartment, and there is no report on the optimization method of the number and installation location of gas sensors in the energy storage system. Summary of the invention

[0006] The purpose of the present invention is to address the problems existing in the above-mentioned prior art and to propose a method for optimizing the arrangement of gas sensors in an energy storage system based on gas detection.

[0007] The technical concept of the present invention is as follows: After the lithium battery is overcharged, due to the saturation of lithium intercalation in the negative electrode, excess lithium ions begin to precipitate on the surface of the negative electrode, forming lithium dendrites; moreover, through research, it is found that at room temperature, lithium dendrites will react with the PVDF binder to release hydrogen; when the side reaction gases inside the battery accumulate to a certain extent, the safety valve will be broken through and characteristic gases will be released; these gases are transmitted to the position of the gas sensor by diffusion. Regarding the gas diffusion mechanism, Fick's law can describe the gas diffusion behavior and can explain that the gas diffusion rate is related to the concentration. The greater the diffusion coefficient and concentration gradient, the faster the gas diffuses; in addition, the pressure gradient will also affect gas diffusion, which can also be explained by Fick's law.

[0008] The method of the present invention includes optimizing the installation quantity of gas sensors and optimizing the installation positions of gas sensors. The main gas to be detected is hydrogen, and by detecting the hydrogen concentration in the target area, it is judged whether the target battery has a thermal runaway failure.

[0009] The technical solution of the present invention includes the following steps:

[0010] S1. Select two single cells at different positions in the energy storage system as fault points, and select several monitoring points to arrange gas sensors. Overcharge the battery at the fault point to trigger thermal runaway to release hydrogen, and obtain the hydrogen concentration data of the monitoring points.

[0011] S2. Establish an energy storage system model, use the ANSYS Fluent fluid numerical analysis method to carry out gas diffusion simulation of the energy storage system, obtain the hydrogen concentration data of the monitoring points selected in S1, and compare and fit with the hydrogen concentration data obtained from the experiment in S1 to prove the reliability of the simulation.

[0012] S3. Divide the monitoring area in the energy storage system into nodes, place gas sensors on each node, obtain the hydrogen concentration data of each node, thereby obtaining a measurement matrix, and divide the hydrogen concentration data of each node by time to obtain the loss coefficient and its corresponding loss coefficient matrix.

[0013] S4. Based on the loss coefficient matrix obtained in S2, use the minimum expectation optimization algorithm to optimize the expectation of the monitoring area, and obtain the optimal installation positions and average detection times when different numbers of gas sensors are installed in the energy storage system.

[0014] S5. Compare the hydrogen concentration data obtained in S1 and S2 with the data in S3. S3 is the simulation of the hydrogen diffusion behavior of the battery pack. Only when the simulation data in S3 fits the data in S1 and S2 can the reliability of the simulation be proved.

[0015] The energy storage system described in step S1 of the present invention includes: an electric vehicle battery pack and energy storage compartments of different shapes.

[0016] The fault point described in step S1 of the present invention can be randomly selected, and the monitoring points are randomly selected in the monitoring area. Among them, the monitoring area of the electric vehicle battery pack is the planar area 0.3 m above the bottom of the battery pack, and the monitoring area of the energy storage cabin is the symmetric plane area in the vertical direction of the energy storage cabin.

[0017] The gas diffusion simulation based on the ANSYS Fluent fluid numerical analysis method described in step S2 of the present invention introduces the standard k-ε equation model to simulate the turbulent process during gas movement. Among them, k represents the turbulent kinetic energy, and ε represents the turbulent dissipation rate; the velocity inlet boundary condition is used to simulate the escape of hydrogen, and a small hole is used to simulate the leakage port; the formula

[0018]

[0019] simulates the hydrogen escape velocity V, is the adiabatic index; is the gas constant; is the gas temperature, which is approximated as a linear function of time here and approximated as a first-order equation of ; is the ambient pressure; is the pressure before gas leakage. According to the ideal gas state equation , assuming that the gas is produced uniformly inside the battery, then is a quadratic equation about .

[0020] The energy storage system model described in step S2 of the present invention is established one-to-one according to the electric vehicle battery pack and the energy storage cabin. The fault points and monitoring points are set the same as in S1, and the obtained hydrogen concentration data are compared to prove that the numerical analysis method adopted in step S2 is feasible and can reliably simulate the gas diffusion behavior of the energy storage system.

[0021] In step S3 of the present invention, a monitoring network containing several candidate nodes is constructed in the monitoring area, the detection data of the sensor is substituted as the observed value to form a measurement matrix, and finally the objective function is solved to complete the expected optimization of the monitoring points.

[0022] For the gas monitoring optimization simulation of the electric vehicle battery pack in step S3 of the present invention, the mirror symmetry method is adopted. Half of the monitoring area is evenly divided into square grids with a side length of 10 cm. There are a total of 35 vertices of all grids, which are used as 35 candidate monitoring points to form a measurement matrix containing 35 nodes. A gas sensor is placed at each node, and the gas concentration data of the measured nodes are filled into the measurement matrix.

[0023] After optimizing the detection positions in half of the area, the optimized simulation model of the monitoring points is symmetrically processed by the mirror symmetry method to obtain the globally optimal expected optimization.

[0024] The optimization of the gas sensor layout in the present invention is mainly to detect the characteristic gas more quickly, that is, it is necessary to minimize the detection time (the time from the release of the gas to the gas sensor first detecting the gas). Using this simulation method, the time for the gas to diffuse to each candidate monitoring point (mainly discrete) can be obtained, that is, the detection time for the gas sensor at this position to detect the characteristic gas.

[0025] In the present invention, the first time the gas concentration at the candidate monitoring point exceeds 30 ppm is recorded as the detection time, that is, the alarm threshold of the gas sensor is 30 ppm.

[0026] The minimum expectation algorithm described in step S4 of the present invention processes the monitoring data of the gas sensor. This algorithm is based on the solution method of the integer programming problem, and the established objective function is as follows:

[0027]

[0028] s.t.

[0029]

[0030]

[0031]

[0032]

[0033]

[0034] Among them, A is the set of all leakage scenarios; represents the probability of leakage scenario a occurring; it is assumed that each single cell has the same TR probability; corresponds to the candidate monitoring position under leakage scenario a, is the loss coefficient when the leakage scenario a is first detected at position i; here, represents the time to detect scenario a at position i;

[0035] The loss coefficient matrix described in step S3 is obtained here; is a decision variable. If the scenario a is detected at position i, then, is 1; otherwise, is 0; is a decision variable. When the gas sensor is installed at position it is 1; otherwise, is 0. p is the gas sensor number;

[0036] Equation 1 is the objective function, representing the expected value of the required performance index (average detection time);

[0037] Equation 3 ensures that only when the gas sensor is installed at position i can it detect the leakage scenario a first; Equation 4 means that only one position among all candidate positions can detect the leakage scenario a first.

[0038] In the present invention, the loss coefficient of each node and its corresponding coefficient matrix are obtained according to the simulation data, substituted into the objective function, multiplied by the probability matrix to obtain the expected observation value, and the minimum expectation algorithm is used to optimize the monitoring expectation to obtain the minimum expected value.

[0039] The minimum expected value obtained by the present invention indicates that in all possible diffusion scenarios, the average detection time of the gas sensor is minimized (assuming that the occurrence probabilities of different gas leakage diffusion scenarios are the same, and the average detection time is the average of the sum of the detection times of these scenarios).

[0040] In the gas monitoring optimization simulation for the energy storage cabin in step S2 of the present invention, there are 150 battery modules on each side of the energy storage cabin. 75 battery modules on one side are selected for gas diffusion simulation, and the monitoring area is divided into 76 candidate monitoring points.

[0041] The method and steps for optimizing the expectation of the monitoring points of the energy storage cabin in the present invention are the same as those of the electric vehicle battery pack.

[0042] The target area refers to any space inside the battery pack and the energy storage cabin where hydrogen can diffuse; the target battery refers to a single battery that has experienced overcharge thermal runaway failure.

[0043] The device for hydrogen detection relies on a gas sensor. The gas sensor completes the collection and identification of hydrogen, collects hydrogen concentration data, and forms a real-time change curve of hydrogen concentration and an expected curve of detection time.

[0044] The present invention combines experiments with simulations. The simulations complete the optimization work of sensor layout that cannot be continued by experiments through extended experiments, and the experiments prove the effectiveness of the simulations.

[0045] On the premise of proving the effectiveness of the hydrogen warning battery thermal runaway, the present invention studies the diffusion behavior characteristics of gases; the simulation can further study the gas diffusion characteristics and complete the optimization of monitoring points, that is, the optimization of the gas sensor installation strategy. The optimization of the monitoring points uses the minimum expectation algorithm to analyze and process the monitoring data of the sensors, and the minimum expectation algorithm is based on the solution method of integer programming problems.

[0046] The beneficial effects of the present invention are as follows: while realizing the warning behavior, it can further accelerate the gas detection speed, thereby shortening the warning delay of battery thermal runaway. Compared with the existing sensor layout strategies, the present invention has obvious advantages, is applicable to different energy storage systems, significantly improves the gas detection efficiency, and can achieve highly sensitive thermal runaway warnings. Therefore, the present invention will lead to the development of a gas sensor layout mode for energy storage systems with wide application scenarios and good commercial prospects, and can greatly improve the safety and stability of energy storage systems. Brief Description of the Drawings

[0047] Figure 1 It is a schematic diagram of an electric vehicle battery pack model;

[0048] Figure 2 It is a schematic diagram of the fault point of an electric vehicle battery pack model;

[0049] Figure 3 It is a schematic diagram of the candidate monitoring points of an electric vehicle battery pack;

[0050] Figure 4 It is the optimal layout diagram when 1 gas sensor is installed in the electric vehicle battery pack;

[0051] Figure 5 It is the optimal layout diagram when 2 gas sensors are installed in the electric vehicle battery pack;

[0052] Figure 6 It is the optimal layout diagram when 3 gas sensors are installed in the electric vehicle battery pack;

[0053] Figure 7 It is a schematic diagram of the relationship between the number of sensors in the battery pack and the expected detection time;

[0054] Figure 8 It is a 3D model diagram of the energy storage cabin and a schematic diagram of the candidate detection points;

[0055] Figure 9 It is a schematic diagram of the candidate detection points of the energy storage cabin;

[0056] Figure 10 It is a schematic diagram of the average detection time and the maximum detection time results of the energy storage cabin;

[0057] Figure 11It is the first optimized solution for the energy storage module layout;

[0058] Figure 12 It is the second optimized solution for the energy storage module layout. Specific implementation manners

[0059] Next, the technical solutions in the embodiments of the present invention will be described clearly and completely. Obviously, the described embodiments should not be construed as limiting the present invention.

[0060] Embodiment

[0061] For the energy storage system of an electric vehicle battery pack, an experimental platform for battery thermal runaway and gas diffusion is built, gas diffusion and thermal runaway warning experiments are designed to verify the effectiveness of early warning of the characteristic gas hydrogen in a small, complex and enclosed energy storage system such as a battery pack. It is found that before the safety valve opens, a small amount of hydrogen leaks, but its diffusion rate is very slow. After the safety valve opens, a large amount of hydrogen leaks, the gas concentration increases significantly, resulting in an increase in the diffusion rate, and the hydrogen concentration at the monitoring point increases significantly. It is worth mentioning that there are significant differences in the hydrogen concentrations detected by gas sensors installed at different positions. It is also found in the experiment that the jet direction of the safety valve is random; further, gas diffusion simulations at different fault point positions are carried out, and it is found that when the position of the fault point is changed, the time taken for hydrogen to diffuse throughout the cabin varies greatly, which is related to the randomness of the safety valve jet direction mentioned above; further, in order to more comprehensively study the gas diffusion characteristics in a small enclosed space, the FLUENT software is used to conduct a strict simulation analysis of the diffusion behavior of hydrogen. The simulation results are basically consistent with the above experimental results, verifying that the simulation can effectively simulate the gas diffusion in the battery pack, so subsequent optimization of the monitoring points can be carried out using the simulation; further, in order to find the optimal detection position of the gas sensor, the mixed integer linear programming formula is used to optimize the detection point position.

[0062] For the large energy storage system of the energy storage module, an experimental platform for battery thermal runaway and gas diffusion is built, gas diffusion and thermal runaway warning experiments are designed, and the faulty single battery is set at different positions to verify the effectiveness of the distributed gas sensor for thermal runaway warning in the energy storage module environment. It is found that there are significant differences in the alarm time delays of gas sensors installed at different positions, and the latest alarm time is close to the battery thermal runaway time, indicating the necessity of optimizing the monitoring points; due to the installation method of the distributed gas sensors used in the experiment, it cannot fully reflect the gas concentration and diffusion conditions at other positions in the cabin. Further, an energy storage module gas diffusion simulation model is established to comprehensively explore the hydrogen diffusion in the entire energy storage module space, and an optimization simulation of the designed monitoring points is carried out; further, the optimization simulation results of the monitoring points are analyzed, the gas sensor installation strategy is studied, and an optimization method for gas sensor installation (including installation position and quantity) is proposed.

[0063] This embodiment conducts gas diffusion tests and simulations for two scenarios of an electric vehicle battery pack and an energy storage cabin. The steps are as follows:

[0064] S1. Select a single battery in a typical electric vehicle battery pack as the fault point, and select three monitoring points. Overcharge the battery at the fault point to trigger thermal runaway and release hydrogen, and obtain the hydrogen concentration data at the monitoring points.

[0065] S2. Change the position of the fault point, release hydrogen signals at different positions, and obtain the hydrogen concentration data at the monitoring points.

[0066] S3. Establish a model of a typical electric vehicle battery pack, use the ANSYS Fluent fluid numerical analysis method to conduct optimization simulations of the monitoring points of the electric vehicle battery pack, and use the minimum expectation optimization algorithm to optimize the expectation of the monitoring area, that is, the comprehensive optimization of the number and installation position of gas sensors, to obtain the optimal layout plan of the gas sensors in the electric vehicle battery pack.

[0067] S4. Randomly select two single batteries at different positions in a typical electric vehicle battery pack as the fault points, and select four monitoring points. Overcharge the batteries at two different positions successively to trigger thermal runaway, and obtain the hydrogen concentration data at the monitoring points.

[0068] S5. Establish a model of a typical electric vehicle battery pack, use the ANSYS Fluent fluid numerical analysis method to conduct optimization simulations of the monitoring points of the electric vehicle battery pack, and use the minimum expectation optimization algorithm to optimize the expectation of the monitoring area, that is, the comprehensive optimization of the number and position of gas sensors, to obtain the optimal layout plan of the gas sensors in the electric vehicle battery pack.

[0069] In step S1, the battery packs are centrally placed in three regions (Region I, II, III), as Figure 1 shown. Among them, the batteries are arranged in two layers, upper and lower. The position of the faulty battery is as Figure 2 shown, the fault point is located at position 1, and it is overcharged to trigger thermal runaway. The coordinates of the three monitoring points are (59.6, 30, -18), (0, 30, -0.4), and (-44.4, 30, 0) respectively, and the hydrogen concentration in the target area is monitored in real time.

[0070] In step S2, change the fault point to Figure 2 position 2 in, and repeat the overcharge and monitoring operations described in step S1. The overcharge and monitoring operations are carried out three times, the hydrogen concentration is obtained and the results are analyzed. It can be concluded that the diffusion speed of the gas is related to its concentration, and the greater the concentration, the faster the diffusion; moreover, the cracking direction of the safety valve is random, which leads to the random direction of the gas ejection, so there is a large optimization space for the installation position of the sensor.

[0071] In step S3, the battery pack model is established in proportion to the actual battery pack; the simulation uses FLUENT numerical analysis software, and the standard k-ε equation model is introduced to simulate the turbulent process during gas movement; the simulation uses the velocity inlet boundary condition to simulate the hydrogen escape situation, and uses small holes to simulate the leakage ports; the formula is used to simulate the hydrogen escape velocity, where is the adiabatic index; is the gas constant; is the gas temperature, which is fitted with a linear function here and approximated as a first-order equation of time . is the ambient pressure; is the pressure before gas leakage. According to the ideal gas state equation , assuming that the gas in the battery is produced at a constant speed, then is a quadratic equation about . Based on the above settings, the simulation is run to obtain the hydrogen concentration data at the monitoring points. According to the simulation results, it is found that the hydrogen concentration curve obtained by the simulation is basically consistent with the hydrogen monitoring curve of the actual battery pack, indicating that this method can effectively simulate the hydrogen diffusion behavior in the battery pack. In the optimization of the simulation at the monitoring points, the method of mirror symmetry is adopted. The monitoring area is the plane area 0.3 m above the battery pack. Half of the monitoring area is evenly divided into square grids with a side length of 10 cm. There are a total of 35 vertices of all grids, which are selected as 35 candidate monitoring points, forming a measurement matrix containing 35 nodes. A sensor is placed at each node to record the monitoring data in real time and fill it into the measurement matrix, as shown in Figure 3 . After optimizing the detection positions in half of the area, the method of mirror symmetry is adopted to symmetrically process the simulation model of the monitoring points to obtain the globally optimal expected optimization.

[0072] The minimum expectation algorithm processes the monitoring data of the sensors. This algorithm is based on the solution method of integer programming problems, and the established objective function is as follows:

[0073]

[0074] s.t.

[0075]

[0076]

[0077]

[0078]

[0079]

[0080] Among them, A is the set of all leakage scenarios; represents the probability of the occurrence of leakage scenario a. Here, we assume that each single battery has the same TR probability; corresponds to the candidate monitoring location under leakage scenario a, is the loss coefficient when the leakage scenario a is first detected at position i. Here, represents the time when scenario a is detected at position i. is a decision variable. If scenario a is detected at position i, then, is 1; otherwise, is 0. is a decision variable. When the sensor is installed at position then, is 1; otherwise, is 0. p is the sensor number. The objective function represents the expected value (average detection time) of the required performance index. Equation 3 ensures that only when the sensor is installed at position i can it first detect leakage scenario a. Equation 4 means that only one position among all candidate positions can first detect leakage scenario a.

[0081] According to the simulation data, the loss coefficient of each node and its corresponding coefficient matrix are obtained, substituted into the objective function, multiplied by the probability matrix to obtain the expected observation value, and the minimum expectation algorithm is used to optimize the monitoring expectation to obtain the minimum expected value. In addition, if the error of the result is large, the threshold can be appropriately increased.

[0082] After optimizing the detection positions in the left half area, the mirror symmetry method is adopted to symmetrically process the monitoring point optimization simulation model to obtain the global optimal expectation optimization.

[0083] The key point of the monitoring point optimization simulation is to explore the hydrogen diffusion behavior released by the underlying 62 batteries and optimize the monitoring point positions to ensure that the early warning time delay can be minimized to the greatest extent.

[0084] As Figure 4 shown, for the optimal gas sensor layout scheme in the battery pack, when setting one sensor, the optimal installation position is in the middle of the battery pack, with the coordinates (0, 30, 10), and the coordinate unit is cm, and the expected time to capture hydrogen is 409.8 s; as Figure 5 shown, when setting two sensors, the optimal installation positions are above Region I and III respectively, with the coordinates (40, 30, 0) and (-40, 30, 0), and the expected time to capture hydrogen is 285.8 s;

[0085] As Figure 6 shown, when three sensors are set, the optimal installation positions are respectively above the I, II, and III regions, with coordinates (40, 30, 0), (-40, 30, 0), and (0, 30, -10), and the expected time to capture hydrogen is 264.5 s.

[0086] As Figure 7 shown, the relationship between the number of sensors and the expected value of the detection time; when there are 2 sensors, the decreasing speed of the expected time to detect hydrogen changes from fast to slow; selecting 2 gas sensors can achieve a good early warning effect while taking cost into account, reducing the number of sensor installations while ensuring reliability. To compare the differences before and after the optimization of the gas sensor positions, two gas sensors are randomly set, and the obtained expected time is 365.8 s. It can be seen that after the optimization of the sensor positions, an average of 80 s of early warning for battery failure can be achieved.

[0087] In step S4, a total of four gas sensors are set in the energy storage cabin, one near the target module, and the other three on the central axis of the top of the energy storage cabin; the target battery is a 52 Ah square lithium iron phosphate battery, and overcharging is carried out to trigger thermal runaway, and the hydrogen concentration data at the monitoring points are obtained; the results show that the times for the sensors at 4 different positions to detect hydrogen are different because the distances of the sensors from the thermally runaway battery are different, and the gas is continuously diluted during diffusion, indicating that the installation positions of the sensors are very important for thermal runaway early warning.

[0088] In step S5, the simulation sets two different gas generation positions, compares the diffusion characteristics under different gas release positions, and conducts experimental verification; the means adopted in the simulation is basically the same as that in S3, and the curve trend of its hydrogen concentration detection is basically the same as that of the experiment, indicating that the simulation can correctly simulate the gas diffusion behavior in the energy storage cabin. In addition, the simulation also shows that hydrogen diffuses very slowly in the long side direction of the energy storage cabin and relatively fast in the short side direction of the energy storage cabin, indicating that the gas sensors should be arranged along the long side direction of the energy storage cabin.

[0089] In the optimization simulation of the monitoring points in the energy storage cabin, as Figure 8 shown, the simulation constructs a one-to-one simulation model for a certain energy storage cabin, with 150 battery modules on each side. To reduce the calculation time, 75 different battery modules are randomly selected for gas diffusion simulation, and 76 candidate monitoring points are selected to form a measurement matrix. The early warning threshold is set to 30 ppm, and the time when the hydrogen concentration at different monitoring points first exceeds 30 ppm is recorded as the detection time, and all data are recorded.

[0090] Analysis of the data reveals that the simulation consistency of different modules under the same battery cluster is relatively high. This is mainly because the cluster is relatively enclosed, restricting the diffusion of leaked gas in the long side direction of the energy storage cabin. Moreover, the opening of the battery cluster is small. No matter which position it is, after hydrogen is released, the internal space of the battery cluster will be filled first, and then it will slowly diffuse into the cabin.

[0091] For the monitoring points on the surface of each battery cluster, using the point plotting method, plot the gas concentration detection curve and find the minimum value. The minimum value of the gas concentration detection curve means the minimum expectation. An optimal monitoring point is found in each battery cluster, and a total of 5 optimal monitoring points are obtained.

[0092] As Figure 10 shown, when different numbers of detection points are selected, the average detection time and maximum detection time curves. The curve is based on the minimum expectation algorithm, whose principle and function are consistent with the above description. Using mixed-integer linear programming to calculate the position of the detection point and the average detection time, this method can more accurately study the selection strategy of the position and number of detection points.

[0093] When optimizing the monitoring points, a total of two groups of candidate monitoring points are selected. The first group is as Figure 9 shown, and the second group is the uppermost monitoring point of Figure 9 . Using the minimum expectation algorithm, substitute the data of all candidate monitoring points into the objective function, multiply the probability matrix by the loss coefficient matrix to obtain the expected observation value. And name the average detection time and maximum detection time after optimizing the first group as and . The average detection time reflects the average ability of the sensor to give early warnings to each battery in the cabin; the maximum detection time represents the early warning ability of the sensor to the battery at the farthest distance. Name the corresponding detection times of the second group as and .

[0094] As can be seen from the curve, selecting 3 - 5 sensors can detect hydrogen before thermal runaway (within 200 s); is a little smaller than , indicating that there are more reasonable detection positions in the first group of candidate monitoring points. When selecting less than 4 sensors, is larger than , which is mainly caused by the diffusion behavior of hydrogen. After hydrogen is released, the concentration gradient will cause hydrogen to diffuse in all directions. The property that the density of hydrogen is small will also make it more inclined to diffuse towards the top of the cabin and then diffuse in all directions at the top.

[0095] Figure 11 Figure 12 ​​As shown, the optimization results of the two groups of solutions. The optimization results of the first group of candidate monitoring points are mainly distributed at the waist of the energy storage cabin; when the number of sensors is small, the faulty battery is far from any sensor, and since hydrogen mainly diffuses at the top of the energy storage cabin, the detection time of the waist sensors is relatively late; while when the number of sensors is large, the faulty battery is close to the sensors, and hydrogen quickly diffuses to the nearest sensor under the influence of the concentration gradient. This shows that when the number of sensors is small, the optimization results of the second group of solutions should be selected.

[0096] Based on this, using distributed gas sensors can significantly reduce the detection time of hydrogen. It is more appropriate to install 3 - 5 sensors in the energy storage cabin; when the number of sensors is large, they should be installed at the waist of the energy storage cabin; when the number of sensors is small, they should be installed at the top of the energy storage cabin.

[0097] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution of the present invention and its inventive concept, makes equivalent substitutions or changes, and all should be covered by the protection scope of the present invention.

Claims

1. A method for optimizing the layout of gas sensors in an energy storage system based on gas detection, characterized in that The steps include: S1. Select two single cells at different locations in the energy storage system as fault points, and select several monitoring points to arrange gas sensors. Overcharge the battery at the fault point to trigger thermal runaway and release hydrogen, and obtain hydrogen concentration data at the monitoring point. S2. Establish an energy storage system model, use ANSYS Fluent fluid numerical analysis method to carry out gas diffusion simulation of the energy storage system, obtain the hydrogen concentration data of the monitoring points selected in S1, compare and fit with the hydrogen concentration data obtained in the S1 experiment, and prove the reliability of the simulation; S3, dividing the monitoring area in the energy storage system into nodes, placing a gas sensor on each node, obtaining the hydrogen concentration data of each node, thereby obtaining a measurement matrix, and dividing the hydrogen concentration data of each node by time to obtain a loss coefficient and its corresponding loss coefficient matrix; S4. Based on the loss coefficient matrix obtained in S3, the minimum expectation optimization algorithm is used to perform expectation optimization on the monitoring area to obtain the optimal installation position and average detection time when different numbers of gas sensors are installed in the energy storage system; S5. Compare the hydrogen concentration data obtained by S1 and S2 with the data in S3. S3 is a simulation of the hydrogen diffusion behavior of the energy storage system. The simulation data in S3 fits the data of S1 and S2 to prove the reliability of the simulation.

2. The method for optimizing the arrangement of gas sensors in an energy storage system based on gas detection according to claim 1, characterized in that: The energy storage system described in step S1 includes an electric vehicle battery pack and an energy storage compartment.

3. The method for optimizing the arrangement of gas sensors in an energy storage system based on gas detection according to claims 1 and 2, characterized in that: The fault point described in step S1 can be randomly selected, and the monitoring point is randomly selected in the monitoring area, wherein the monitoring area of ​​the electric vehicle battery pack is the plane area 0.3m above the bottom of the battery pack, and the monitoring area of ​​the energy storage cabin is the symmetrical plane area in the vertical direction of the energy storage cabin.

4. A method for optimizing the arrangement of gas sensors in an energy storage system based on gas detection as claimed in claim 1, characterized in that: The gas diffusion simulation based on the ANSYS Fluent fluid numerical analysis method described in step S2 introduces a standard k-ε equation model to simulate the turbulent process of gas movement, wherein k represents the turbulent kinetic energy and ε represents the turbulent dissipation rate; a velocity inlet boundary condition is used to simulate the escape of hydrogen, and a small hole is used to simulate the leakage port; formula Simulate the hydrogen escape velocity V, k is the adiabatic index; R is the gas constant; T is the gas temperature, which is approximated as a linear equation of time t by a linear function fit; P0 is the ambient pressure; P1 is the pressure before gas leakage, according to the ideal gas state equation PV=nRT, assuming that the gas inside the battery is produced at a uniform rate, then P1 is a quadratic equation about t.

5. A method for optimizing the arrangement of gas sensors in an energy storage system based on gas detection as described in claims 1 and 2, characterized in that: The energy storage system model described in step S2 is established based on a one-to-one ratio between the electric vehicle battery pack and the energy storage cabin, and the fault point and monitoring point are set consistent with S1. The obtained hydrogen concentration data are compared to prove that the numerical analysis method used in step S2 is feasible and can reliably simulate the gas diffusion behavior of the energy storage system.

6. A method for optimizing the arrangement of gas sensors in an energy storage system based on gas detection as claimed in claim 1, characterized in that: In step S3, for the gas monitoring optimization simulation of the electric vehicle battery pack, the mirror symmetry method is adopted to evenly divide half of the monitoring area into square grids with a side length of 10 cm. There are 35 vertices in all grids, which serve as 35 candidate monitoring points to form a measurement matrix containing 35 nodes. A gas sensor is placed at each node, and the gas concentration data of the measured node is filled into the measurement matrix.

7. A method for optimizing the arrangement of gas sensors in an energy storage system based on gas detection as claimed in claim 6, characterized in that: After optimizing the detection positions of half of the area, the monitoring point optimization simulation model is symmetrically processed by using a mirror symmetry method to obtain the global optimal expected optimization.

8. A method for optimizing the arrangement of gas sensors in an energy storage system based on gas detection as claimed in claim 1, characterized in that: The first time that the gas concentration at the candidate monitoring point exceeds 30 ppm is recorded as the detection time, that is, the alarm threshold of the gas sensor is 30 ppm.

9. A method for optimizing the arrangement of gas sensors in an energy storage system based on gas detection as claimed in claim 1, characterized in that: The minimum expectation optimization algorithm described in step S4 processes the monitoring data of the gas sensor. The algorithm is based on the solution method of integer programming problem, and the objective function is set as follows: Among them, A is the set of all leakage scenarios; α a represents the probability of leakage scenario a occurring; assuming that each single cell has the same TR probability; Corresponding to the candidate monitoring locations under leakage scenario a, d a,i is the loss coefficient when leakage scenario a is first detected at location i; here, d a,i represents the time when scene a is detected at position i; The loss coefficient matrix described in step S3 is obtained here; x a,i is a decision variable. If scene a is detected at position i, then x a,i is 1; otherwise, x a,i is 0;s l is the decision variable. When the gas sensor is installed at position l, s l is 1; otherwise, s l is 0; p is the gas sensor number; Formula 1 is the objective function, which represents the expected value of the required performance indicator, that is, the average detection time; Formula 3 shows that leakage scenario a can be detected first only when the gas sensor is installed at position i; Formula 4 shows that leakage scenario a is detected first at only one position among all candidate positions.

10. A method for optimizing the arrangement of gas sensors in an energy storage system based on gas detection as claimed in claim 9, characterized in that: According to the simulation data, the loss coefficient of each node and its corresponding coefficient matrix are obtained, which are substituted into the objective function and multiplied by the probability matrix to obtain the expected observation value. The minimum expectation algorithm is used to optimize the monitoring expectation and obtain the minimum expected value.

Images