Hydrogen concentration estimation system

Abstract

To provide a hydrogen concentration estimation system with which it is possible to estimate a hydrogen concentration in a mixed gas that includes hydrogen.SOLUTION: A hydrogen concentration estimation system is used to estimate a hydrogen concentration in a mixed gas that includes hydrogen and a main gas composed of one or two or more kinds of component gases whose molar mass is 10 g / mol or greater, said system comprising: a first, a second, ... and an m-th flow rate sensor (m≥2) for measuring the flow rate of the mixed gas; a storage unit in which are stored the outputs of the first, second, ... and the m-th flow rate sensors when the mixed gas is caused to flow in the first, second, ... and the m-th flow rate sensors; and an estimation unit that estimates the hydrogen concentration in the mixed gas using two or more of the outputs. The first, second, ... and the m-th flow rate sensors have mutually different flow rate-output characteristics. The estimation unit includes means that estimates the hydrogen concentration using a preliminarily acquired function in which the two or more of the outputs are adopted as explanatory variables and the hydrogen concentration is adopted as an objective variable.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present invention relates to a hydrogen concentration estimation system, and more particularly to a hydrogen concentration estimation system for estimating the hydrogen concentration in a mixed gas containing hydrogen. [Background technology]

[0002] Hydrogen is produced by various methods, such as electrolysis of water, steam reforming of hydrocarbons, and separation of hydrogen from gas discharged from factories such as steelworks. The produced hydrogen is also used for various purposes, such as fuel for fuel cells and hydrogen fuel engines, raw material for manufacturing chemicals, and reducing agent. In the process of producing or using such hydrogen, a mixed gas containing hydrogen may be generated, or the mixed gas containing hydrogen may leak to the outside due to unavoidable circumstances. Furthermore, in some types of plants (e.g., boiling water nuclear power plants), hydrogen may be generated by unintended side reactions. In such a case, it may be desirable to know the hydrogen concentration in the mixed gas from the viewpoint of improving the efficiency of the hydrogen production process or use process, ensuring safety, etc. Therefore, various proposals have been made regarding a method for estimating the hydrogen concentration in a mixed gas.

[0003] For example, Patent Document 1 states: (a) providing a fuel gas supply valve and a pressure sensor in a fuel gas passage for supplying hydrogen to the fuel cell stack, and providing a heat flow sensor and a temperature sensor in a gas-liquid separator for recovering hydrogen contained in the off-gas discharged from the fuel cell stack and returning it to the fuel gas passage; (b) measuring the fluctuation range Va of the heat flow sensor output and the fluctuation range Vb of the pressure sensor output that occur when the fuel gas supply valve is periodically opened and closed; (c) The output fluctuation ratio (= Va / Vd) is corrected by the difference ΔT between the temperature on the front side (hydrogen gas side) and the temperature on the back side of the heat flow sensor. (c) Estimate the hydrogen concentration in the off-gas using the corrected output fluctuation ratio. A hydrogen concentration detection device is disclosed. The document describes that this method makes it possible to detect the hydrogen concentration with good detection accuracy even when the load or the outside air temperature changes.

[0004] Patent Document 2 discloses a method for estimating the hydrogen concentration in an anode gas circulation flow path in a fuel cell system, using the anode gas temperature in the anode gas circulation flow path, the pressure difference between the inlet and outlet of a circulation pump provided in the anode gas circulation flow path, the gas density in the anode gas flow path, and the relative humidity in the fuel cell. The document states: (A) In estimating the hydrogen concentration in the anode gas circulation flow path, if the content of water vapor is assumed to be the saturated water vapor fraction, the estimation accuracy of the hydrogen concentration decreases; and (B) When estimating the hydrogen concentration in the anode gas circulation flow path, the hydrogen concentration can be estimated with higher accuracy by using the relative humidity in the fuel cell. is stated.

[0005] Patent Document 3 states: (a) sampling a portion of the circulating gas, oxidizing hydrogen contained in the sampled gas with an oxidation catalyst, and measuring the HO concentration contained in the gas after the oxidation treatment; (b) sampling another portion of the circulating gas and measuring the HO concentration in the sampled untreated gas; (c) Calculate the hydrogen concentration in the circulating gas from the difference between the HO concentration in the gas after oxidation treatment and the HO concentration in the untreated gas. A hydrogen concentration measuring device is disclosed. The document describes that this method makes it possible to measure the hydrogen concentration simply, with high response and high accuracy.

[0006] Patent Document 4 states: (a) Data on the gas flowing into the recombiner of the boiling water nuclear power plant (flow rate data of hydrogen, oxygen, nitrogen, and water vapor contained in the gas flowing into the recombiner, temperature data of the gas flowing into the recombiner, and flow rate data of catalyst poisoning substances flowing into the recombiner), and (b) Data on the catalyst loaded in the recombiner (data on the amount of active component contained in the catalyst, activation energy and frequency factor of the catalyst). A hydrogen concentration prediction device is disclosed that predicts the hydrogen concentration in the gas discharged from the recombiner using the above. The document describes that this method makes it possible to predict the hydrogen concentration in the gas discharged from the recombiner, thereby further improving the safety of nuclear power plants.

[0007] Furthermore, Patent Document 5 discloses a method for detecting the hydrogen concentration or hydrogen partial pressure in the atmosphere of a treatment furnace in which steel surface treatment is performed, using a solid electrolyte hydrogen sensor equipped with a solid electrolyte that exhibits proton conductivity. The same document states: (A) When the hydrogen concentration in the processing furnace is detected using a thermal conduction type hydrogen sensor, a detection delay of several tens of seconds occurs in response to changes in the hydrogen concentration; and (B) The use of a solid electrolyte hydrogen sensor allows the hydrogen concentration to be detected in a timely manner in response to changes in the hydrogen concentration in the processing furnace, enabling rapid feedback control. is stated.

[0008] Patent Document 1 describes a method of estimating hydrogen concentration by periodically opening and closing a fuel gas supply valve to periodically fluctuate the output of a heat flow sensor and the output of a pressure sensor, and using the fluctuation width Va of the heat flow sensor output and the fluctuation width Vb of the pressure sensor output. However, the method described in this document cannot be used in an environment where periodic pressure fluctuations do not occur.

[0009] The methods described in Patent Documents 2 to 4 all require special equipment or are applicable only to special purposes, and are therefore poor in versatility. In addition, it is difficult to simply estimate the hydrogen concentration in a mixed gas flow containing hydrogen using these methods. Furthermore, the method described in Patent Document 5 can be used only in a high-temperature environment of 400° C. or higher, and is therefore not very versatile. [Prior art documents] [Patent documents]

[0010] [Patent Document 1] JP 2019-164918 A [Patent Document 2] Patent No. 6620510 [Patent Document 3] Patent No. 5709408 [Patent Document 4] Patent No. 5686539 [Patent Document 5] JP 2020-084219 A Summary of the Invention [Problem to be solved by the invention]

[0011] An object of the present invention is to provide a hydrogen concentration estimation system capable of estimating the hydrogen concentration in a mixed gas containing hydrogen. Another problem to be solved by the present invention is to provide a hydrogen concentration estimation system that does not require the use of special equipment and is highly versatile. [Means for solving the problem]

[0012] In order to solve the above problems, a hydrogen concentration estimation system according to the present invention has the following configuration. (1) The hydrogen concentration estimation system is used to estimate the hydrogen concentration of a mixed gas containing hydrogen and a main gas consisting of one or more component gases having a molar mass of 10 g / mol or more. (2) The hydrogen concentration estimation system comprises: first, second, ..., mth flow rate sensors (m≧2) for measuring a flow rate of the mixed gas; a storage unit that stores outputs of the first, second, ..., mth flow rate sensors when the mixed gas is caused to flow through the first, second, ..., mth flow rate sensors; an estimation unit that estimates a hydrogen concentration of the mixed gas by using any two or more of the outputs; It is equipped with: (3) The first, second, ..., mth flow rate sensors have flow rate-output characteristics different from one another. (4) The estimation unit includes means for estimating the hydrogen concentration by using a function having any two or more of the outputs acquired in advance as explanatory variables and the hydrogen concentration as a response variable.

[0013] The hydrogen concentration estimation system includes: Further comprising a temperature sensor for measuring a temperature of the mixed gas, It is preferable that the estimation unit includes a means for estimating the hydrogen concentration using a function having two or more of the outputs and the temperature, which have been acquired in advance, as explanatory variables and the hydrogen concentration as a response variable. Effect of the Invention

[0014] When the flow rate-output characteristics of a hydrogen-free main gas are measured using multiple flow rate sensors with mutually different flow rate-output characteristics, a certain functional relationship (hereinafter also referred to as the "reference function") is established between the outputs of the multiple flow rate sensors. On the other hand, when the flow rate-output characteristics of a hydrogen-containing mixed gas are measured using such multiple flow rate sensors, the data points formed by the combination of those outputs deviate from the reference function. Moreover, the distance (d) between the reference function and the data points is correlated with the hydrogen concentration.

[0015] Therefore, if a function is obtained in advance using two or more of the outputs of multiple flow sensors when mixed gases with different flow rates and / or hydrogen concentrations are flowed as explanatory variables and the hydrogen concentration as the objective variable, the hydrogen concentration can be estimated using that function. The same is true when estimating the hydrogen concentration taking into account the temperature of the mixed gas; if a function is obtained in advance that uses two or more of the outputs of multiple flow sensors and the temperature as explanatory variables and the hydrogen concentration as the objective variable, the hydrogen concentration can be estimated using that function. [Brief description of the drawings]

[0016] [Figure 1] 4 is a diagram showing the relationship between the output value of a flow sensor 1 and the output value of a flow sensor 2. FIG. [Diagram 2] FIG. 2 is an enlarged view of FIG. [Diagram 3] FIG. 13 is a graph showing the relationship between distance d and hydrogen concentration n. [Figure 4] FIG. 4 is a graph showing the relationship between the actually measured hydrogen concentration and the predicted hydrogen concentration. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] An embodiment of the present invention will be described in detail below. [1. Hydrogen concentration estimation system] The hydrogen concentration estimation system according to the present invention has the following configuration. (1) The hydrogen concentration estimation system is used to estimate the hydrogen concentration of a mixed gas containing hydrogen and a main gas consisting of one or more component gases having a molar mass of 10 g / mol or more. (2) The hydrogen concentration estimation system comprises: first, second, ..., mth flow rate sensors (m≧2) for measuring a flow rate of the mixed gas; a storage unit that stores outputs of the first, second, ..., mth flow rate sensors when the mixed gas is caused to flow through the first, second, ..., mth flow rate sensors; an estimation unit that estimates a hydrogen concentration of the mixed gas by using any two or more of the outputs; It is equipped with: (3) The first, second, ..., mth flow rate sensors have flow rate-output characteristics different from one another. (4) The estimation unit includes means for estimating the hydrogen concentration by using a function having any two or more of the outputs acquired in advance as explanatory variables and the hydrogen concentration as a response variable.

[0018] The hydrogen concentration estimation system may further include a temperature sensor for measuring a temperature of the mixed gas. In this case, it is preferable that the estimation unit includes a means for estimating the hydrogen concentration using a function having two or more of the outputs and the temperature, which are previously acquired, as explanatory variables and the hydrogen concentration as a response variable.

[0019] [1.1. Gas mixture] The term "mixed gas" refers to a gas that is the subject of inspection by the hydrogen concentration estimation system according to the present invention and that contains hydrogen and a main gas. "Main gas" means a gas consisting of one or more component gases having a molar mass of 10 g / mol or more.

[0020] Examples of component gases with a molar mass of 10 g / mol or more include methane (CH4, 16 g / mol), propane (C3H8, 44 g / mol), ethane (C2H6, 30 g / mol), butane (C4H 10 , 58g / mol), nitrogen (N2, 28g / mol), water vapor (H2O, 18g / mol), etc. The main gas may contain only one of these component gases, or it may be a mixture of two or more component gases.

[0021] However, if the mixed gas contains a large amount of other gases with a molar mass of less than 10 g / mol (e.g., helium (He) with a molar mass of 4 g / mol), the detection accuracy of the hydrogen concentration may decrease. Therefore, the mixed gas preferably has a total content of hydrogen and main gas of 99 vol% or more.

[0022] [1.2. 1st to mth flow rate sensors] [1.2.1. Definition] "First, second, ..., mth flow sensor (m≧2)" refers to a flow sensor for measuring the flow rate of a mixed gas containing hydrogen. The first, second, ..., mth flow sensors must have different flow-output characteristics from each other. When a main gas having a hydrogen concentration n of zero is flowed through each of the ith flow sensor (1≦i≦m) and the jth flow sensor (1≦j≦m, j≠i), the output Y of the ith flow sensor is i and the output Y of the jth flow sensor j increases as the flow rate of the main gas increases. Therefore, the relationship between the two is expressed by a linear equation: Y i =b×Y j It can be approximated as +c. "Having different flow rate-output characteristics" means that the increase in output when the main gas flow rate is fixed and only the hydrogen concentration n is changed is expressed as ΔY i and ΔY j Then, ΔY i / ΔY j This means that the relationship =b does not hold.

[0023] 1.2.2. Quantity The number of flow sensors (m) may be 2 or more. However, even if the number of flow sensors is increased more than necessary, there is no difference in the effect and no practical benefit. Therefore, it is preferable to select the optimal number of flow sensors depending on the composition of the mixed gas, the purpose of the inspection, etc.

[0024] Even when a single flow sensor is used, there is a tendency for the flow rate-output characteristics to change in response to changes in hydrogen concentration, but it is difficult to separate changes in the flow rate-output characteristics caused by changes in hydrogen concentration from changes in the flow rate-output characteristics caused by other factors. However, by comparing the changes in the flow rate-output characteristics of different types of flow sensors, it is possible to clarify the changes in the flow rate-output characteristics caused by changes in hydrogen concentration. Therefore, by comparing the results of multiple flow sensors, it is possible to estimate the hydrogen concentration more accurately than with conventional methods.

[0025] 1.2.3. Type In the present invention, the type of the flow sensor is not particularly limited. Examples of flow rate sensors include: (a) Mass flow rate type flow sensors such as Coriolis flow rate sensors; (b) Thermal mass flow rate type flow sensors, such as capillary flow sensors, hot wire flow sensors, thermocouple flow sensors, and semiconductor flow sensors; etc.

[0026] Among these, at least one of the first, second, . . . mth flow rate sensors is preferably a mass flow rate type flow rate sensor. Mass flow rate type flow sensors detect the mass of a gas, and the output of the sensor is determined primarily by the molar mass of the gas. Meanwhile, the molar mass of hydrogen (H2) is 2g / mol, which is about an order of magnitude lower than the molar mass of methane and other gases. Therefore, when the hydrogen concentration in a mixed gas changes, the flow rate-output characteristics of a mass flow rate type flow sensor change relatively significantly. In addition, by comparing the changes in the flow rate-output characteristics of multiple flow sensors with different flow rate-output characteristics, it is possible to clarify the changes in the flow rate-output characteristics caused by changes in hydrogen concentration. As a result, the hydrogen concentration can be estimated more accurately than with conventional methods.

[0027] At least one of the first, second, ..., mth flow rate sensors is preferably a thermal mass flow rate type flow rate sensor. The thermal conductivity of hydrogen, which has a small molar mass, at room temperature is 0.16 W / m K, which is about an order of magnitude higher than that of gases with a molar mass of 10 g / mol or more, such as methane (thermal conductivity = 0.03 W / m K). On the other hand, thermal mass flow rate sensors detect the temperature difference caused by the heat flow of gas, and the flow of hydrogen gas, which has a thermal conductivity one order of magnitude higher, has the effect of significantly reducing the temperature difference between gases upstream and downstream of the heat source (for example, the difference in thermoelectromotive force when the temperature difference is detected as the difference in thermoelectromotive force) compared to the flow of other gases.

[0028] Therefore, when the hydrogen concentration in a mixed gas changes, the flow rate-output characteristics of a thermal mass flow rate type flow sensor change relatively more significantly. Also, by comparing the changes in the flow rate-output characteristics of multiple flow sensors with different flow rate-output characteristics, it is possible to clarify the changes in the flow rate-output characteristics caused by changes in hydrogen concentration. As a result, it is possible to estimate the hydrogen concentration more accurately than with conventional methods.

[0029] [1.3. Temperature sensor] The hydrogen concentration estimation system according to the present invention may further include a temperature sensor for measuring the temperature of the mixed gas. In the present invention, the type of the temperature sensor is not particularly limited, and an optimal temperature sensor can be used depending on the purpose.

[0030] When measuring the hydrogen concentration in a mixed gas, the temperature of the mixed gas may be considered to be constant. However, if the temperature of the mixed gas changes, the flow rate of the mixed gas also changes. In order to measure the hydrogen concentration in the mixed gas more accurately, it is preferable to measure the temperature of the mixed gas and correct the hydrogen concentration using the measured temperature.

[0031] [1.4. Storage section] The "memory unit" refers to a device for storing the outputs of the first, second, ..., mth flow sensors when a mixed gas is flowed through the first, second, ..., mth flow sensors. The structure of the storage unit is not particularly limited as long as it is capable of storing the output. The output stored in the storage unit is sent to the estimation unit and used to estimate the hydrogen concentration.

[0032] [1.5. Estimation part] The "estimation unit" refers to a device for estimating the hydrogen concentration contained in a mixed gas by using the outputs of two or more of the first, second, ..., mth flow sensors when the mixed gas is flowed through the first, second, ..., mth flow sensors. The structure of the estimation section is not particularly limited as long as it is possible to estimate the hydrogen concentration using the outputs of the first, second, . . . mth flow rate sensors.

[0033] The estimation unit may include means for estimating the hydrogen concentration by using a function having any two or more previously acquired outputs as explanatory variables and the hydrogen concentration as a response variable. The estimation unit may also include means for estimating the hydrogen concentration using a function having two or more previously acquired outputs and the temperature as explanatory variables and the hydrogen concentration as a response variable. Specific examples of means for estimating hydrogen concentration include the following:

[0034] [1.5.1. Estimation method (1): Multivariate nonlinear function] The estimation unit may include a means for calculating the hydrogen concentration using the following formula (1). n[vol%] = F(Y1, Y2, …, Y m ) …(1) however, n is the hydrogen concentration of the mixed gas, Y1, Y2, … Y m are the outputs of the first, second, ..., mth flow rate sensors when the mixed gas is flowed through the first, second, ..., mth flow rate sensors, respectively. F is the Y1, Y2, ... Y m A multivariate nonlinear function having two or more of the above as explanatory variables.

[0035] When the flow rate-output characteristics of a main gas that does not contain hydrogen are measured using multiple flow rate sensors with different flow rate-output characteristics, a certain functional relationship (reference function) is established between the outputs of the multiple flow rate sensors. On the other hand, when the flow rate-output characteristics of a mixed gas that contains hydrogen are measured using multiple such flow rate sensors, the data points formed by the combination of those outputs deviate from the reference function. Moreover, the distance (d) between the reference function and the data points is correlated with the hydrogen concentration. Therefore, a certain functional relationship is established between the hydrogen concentration and the outputs of the first, second, ..., mth flow rate sensors when the premixed gas is flowed through the first, second, ..., mth flow rate sensors.

[0036] Equation (1) represents the outputs Y1, Y2, ..., Y3 of the first, second, ..., mth flow rate sensors when mixed gases with different flow rates and / or hydrogen concentrations are flowed. m It represents a function with two or more of the above as explanatory variables and hydrogen concentration n as the objective variable. F is a function of Y1, Y2, ...Y mThe function form is not particularly limited as long as it is a multivariate nonlinear function with two or more of the above as explanatory variables. For example, F may be a multivariate polynomial, an exponential function, a logarithmic function, a trigonometric function, or a combination of two or more of these.

[0037] [1.5.2. Estimation methods (2): Machine learning models] The estimation unit may include a means for calculating the hydrogen concentration using a machine learning model that has previously trained a correspondence between the hydrogen concentration and two or more of the outputs of the first, second, ..., mth flow sensors when the mixed gas is flowing.

[0038] By passing various mixed gases with different hydrogen concentrations and / or flow rates through the first, second, ..., mth flow sensors and learning the correspondence between the outputs of the first, second, ..., mth flow sensors and the hydrogen concentrations at that time through machine learning, a machine learning model can be obtained in which the outputs of the first, second, ..., mth flow sensors are used as explanatory variables and the hydrogen concentrations are used as objective variables. If such a machine learning model is constructed in advance, the hydrogen concentrations can be estimated using the machine learning model when the outputs of the first, second, ..., mth flow sensors are obtained.

[0039] [1.5.3. Estimation method (3): Linear approximation] The estimation unit may include means for calculating the hydrogen concentration using the following equations (2) and (3). d=|bY j -Y k +c| / √(b 2 +1) …(2) n [vol%] = αd + β … (3)

[0040] however, n is the hydrogen concentration of the mixed gas, Y j (1≦j≦m) is the output of the jth flow sensor when the mixed gas is flowed through the jth flow sensor, Y k (1≦k≦m, k≠j) is the output of the kth flow sensor when the mixed gas is flowed through the kth flow sensor, b and c are the outputs Y of the j flow sensor when the main gas having different flow rates and / or compositions is flowed through the j flow sensor and the k flow sensor, respectively. j ' is the x-axis, and the output Y of the kth flow sensor is k ' is the y-axis, and the Y j ' and the Y k The correspondence between Y k '=bY j ' + c) the slope and y-intercept of the linear equation when approximated, α and β are each a constant.

[0041] When the flow rate-output characteristics of the main gas are measured using two flow sensors (the jth flow sensor and the kth flow sensor), the output Y of the jth flow sensor is j ' and the output Y of the kth flow sensor k However, if the difference between the flow rate-output characteristics of the two is small, the relationship between the two can be approximated by a linear equation (see Figure 1). On the other hand, if two such flow sensors are used to measure the flow rate-output characteristics of a mixed gas containing hydrogen, the combination of their outputs (Y j , Y k ) deviates from the linear approximation (see Figure 2).

[0042] "d" in Equation (2) and Equation (3) is the combination of the approximation line and the output when measuring the mixed gas (Y j , Y k ) represents the shortest distance between the hydrogen concentration n and the output (Y j , Y k ), n can be calculated from equations (2) and (3).

[0043] The estimation of n using the above-mentioned formula (1) or the machine learning model is based on the same principle as the estimation of n using formulas (2) and (3), but mThis method differs from the estimation of n using formula (2) and formula (3) in that the relationship between n and n is described by a nonlinear function. The estimation of n using formula (2) and formula (3) has a lower estimation accuracy than the estimation of n using formula (1) or a machine learning model, but has the advantage of being able to reduce calculation costs compared to the estimation of n using formula (1) or a machine learning model.

[0044] [1.5.4. Estimation method (4): Multivariate nonlinear function taking temperature into account] In the case where the hydrogen concentration estimation system further includes a temperature sensor for measuring the temperature of the mixed gas, the estimation unit may include a means for calculating the hydrogen concentration using the following equation (4). n[vol%] = F'(Y1, Y2, ..., Y m , T) …(4) however, n is the hydrogen concentration of the mixed gas, Y1, Y2, … Y m are the outputs of the first, second, ..., mth flow rate sensors when the mixed gas is flowed through the first, second, ..., mth flow rate sensors, respectively. T is the temperature of the mixed gas. F' is any one of Y1, Y2, ... Y m and the T as the explanatory variables.

[0045] Equation (4) is used to estimate the hydrogen concentration n by using the outputs Y1, Y2, ... Y m In addition, the temperature T of the mixed gas is taken into consideration. This is what makes it different from equation (1). Other points are the same as equation (1), so a description thereof will be omitted.

[0046] [1.5.5. Estimation method (5): Machine learning model considering temperature] If the hydrogen concentration estimation system further includes a temperature sensor for measuring the temperature of the mixed gas, the estimation unit may include a means for calculating the hydrogen concentration using a machine learning model that has previously machine-learned the correspondence between two or more of the outputs of the first, second, ..., mth flow sensors when the mixed gas is flowing, the temperature, and the hydrogen concentration.

[0047] The estimation means (5) estimates the hydrogen concentration n based on the outputs Y1, Y2, ..., Y3 of the first, second, ..., m-th flow rate sensors. m In addition, the temperature T of the mixed gas is taken into consideration. This point is different from the estimation means (2). Other points are the same as the estimation means (2), so the explanation will be omitted.

[0048] [2. Effect] There are many situations in which you may want to know the hydrogen concentration in a mixed gas. For example, (a) When you want to check whether the hydrogen ratio is as desired after setting the gas flow rate to create a mixed gas, (b) If you want to know whether the hydrogen concentration at a hydrogen station is as you want, (c) If you want to detect the accidental inclusion of hydrogen gas and notify the user of the abnormality, The following are anticipated:

[0049] Normally, flow sensors measure the flow rate of a mixed gas, and it is difficult to estimate the concentration of each gas. In addition, if you want to know the concentration of each gas, you will need appropriate equipment such as an infrared spectrometer or gas chromatography device.

[0050] In contrast, when multiple flow sensors with different flow-output characteristics are used to measure the flow-output characteristics of a main gas that does not contain hydrogen, a certain functional relationship (reference function) is established between the outputs of the multiple flow sensors. On the other hand, when multiple flow sensors are used to measure the flow-output characteristics of a mixed gas that contains hydrogen, the data points formed by the combination of those outputs deviate from the reference function. Moreover, the distance (d) between the reference function and the data points is correlated with the hydrogen concentration.

[0051] Therefore, if a function is obtained in advance using two or more of the outputs of multiple flow sensors when mixed gases with different flow rates and / or hydrogen concentrations are flowed as explanatory variables and the hydrogen concentration as the objective variable, the hydrogen concentration can be estimated using that function. The same is true when estimating the hydrogen concentration taking into account the temperature of the mixed gas; if a function is obtained in advance that uses two or more of the outputs of multiple flow sensors and the temperature as explanatory variables and the hydrogen concentration as the objective variable, the hydrogen concentration can be estimated using that function.

[0052] The hydrogen concentration estimation system according to the present invention cannot know the concentrations of all gases contained in a mixed gas, but can estimate the hydrogen concentration in the mixed gas by simply using multiple simple and inexpensive flow sensors. EXAMPLES

[0053] (Examples 1 to 4) [1. Test method] Detection gases with different compositions and / or flow rates were passed through three types of flow sensors 1 to 3, which have different flow rate-output characteristics, and the sensor outputs at that time were measured. Flow sensor 1 was a thermal type flow sensor for city gas. Flow sensor 2 was a thermal type flow sensor for LP. Flow sensor 3 was a thermal type flow sensor for air. The detection gases used were various gases containing various gases in the following ratios: methane: 66-95 vol%, propane: 0-21 vol%, ethane: 0-10 vol%, butane: 0-3 vol%, hydrogen: 0-31 vol%. Hereinafter, the detection gas that does not contain hydrogen is also referred to as the "main gas", and the detection gas that contains hydrogen and a gas other than hydrogen is also referred to as the "mixed gas".

[0054] [2. Results] [2.1. Example 1: Estimation of hydrogen concentration using formulas (2) and (3)] Figure 1 shows the relationship between the output value of flow sensor 1 and the output value of flow sensor 2. Figure 1 shows that the data points for mixed gas (with hydrogen) appear to the lower right of the data points for main gas (without hydrogen). This is because the two flow sensors have slightly different flow-output characteristics.

[0055] Figure 2 shows an enlarged view of Figure 1. As shown in Figure 2, the data points of the main gas (without hydrogen) are fitted with a straight line using the least squares method. In addition, the distance between the data points of the mixed gas (with hydrogen) and the fitted straight line is defined as d. If Y1' is the output of flow sensor 1 when the main gas (without hydrogen) is flowed, and Y1' is the output of flow sensor 2, Y2', the fitted straight line can be expressed as Y2' ≒ bY1' + c. In addition, when the outputs (Y1, Y2) of the flow sensors when the mixed gas (with hydrogen) is flowed are obtained, the distance d between the obtained data points and the straight line is d = |bY1-Y2+c| / √(b 2 +1).

[0056] Figure 3 shows the relationship between distance d and hydrogen concentration n. Figure 3 also shows the result of fitting the data with a straight line using the least squares method (n [vol%] = αd + β). Figures 1 to 3 show that if the distance d of the mixed gas for which you want to know the hydrogen concentration is found, the hydrogen concentration n can be estimated from the fitted straight line (n = αd + β).

[0057] [2.2. Examples 2 to 4: Estimation of hydrogen concentration using machine learning models] Estimation of hydrogen concentration using formulas (2) and (3) is a simple estimation method, but the estimation accuracy is not good because it uses a simple formula that calculates hydrogen concentration n from the linear relationship between hydrogen concentration n and distance d. Therefore, in order to improve the estimation accuracy of hydrogen concentration, an estimation model was constructed by machine learning (neural network with statistical analysis software R) using data from flow sensor 2 and flow sensor 1 (Example 2).

[0058] In the neural network, the hydrogen concentration estimation model is constructed as a nonlinear function n = F(Y1, Y2) with the hydrogen concentration n being the variable of the mixed gas (Y1, Y2) contained in the database. Therefore, when this method is used, n [vol%] = F(Y1, Y2, ..., Y m A hydrogen concentration estimation model can also be constructed using multiple variables, including the temperature T of the mixed gas, such as the temperature T of the mixed gas. This time, we used nnet (which uses a sigmoid function as the activation function) as the library for neural networks, and the hyperparameters are size=10, decay=0.009, maxit=100000, and reltol=1.0e. -9 We constructed an estimation model as follows. In addition, we evaluated the generalization performance by performing 10-fold cross-validation for machine learning.

[0059] Figure 4 shows the relationship between the measured hydrogen concentration and the predicted hydrogen concentration. Note that Figure 4 shows both predictions using distance d and predictions using a machine learning model. The outputs of flow sensors 1 and 2 were used to build the machine learning model. Figure 4 shows that the predictions using the machine learning model introduce nonlinear effects into the estimation, and therefore have improved estimation accuracy compared to estimations using d. The "hydrogen concentration (%)" on the vertical axis of Fig. 3 corresponds to the "measured hydrogen concentration (%)" on the horizontal axis of Fig. 4. The predicted hydrogen concentration using d shown in Fig. 4 was calculated by substituting d into the straight line in Fig. 3 (a straight line approximating the experimental results using the least squares method).

[0060] The root mean square error (RMSE) between the actual and predicted hydrogen concentration values ​​in 10-fold cross-validation for the predicted hydrogen concentration values ​​estimated using the machine learning model was CV ) and the coefficient of determination (R CV 2 The results are shown in Table 1. RMSE CV and R CV 2 Since RMSE is the average value of each of the 10 test data, it is not related to the learning data used to construct the estimation model, and indicates how well it can be estimated in a generalized manner. CV = 1.774%, which shows that a model with low estimation error for hydrogen concentration has been constructed.

[0061] Table 1 also shows the evaluation results (Example 3) when the hydrogen concentration was estimated by machine learning using flow sensor 1 and flow sensor 3, and the evaluation results (Example 4) when the hydrogen concentration was estimated by machine learning using flow sensor 2 and flow sensor 3. From Table 1, it can be seen that the RMSE CV It can be seen that an estimation model with low error could be constructed. In this way, it was found that the hydrogen concentration can be estimated easily and with high accuracy by using multiple flow sensors with different flow rate-output characteristics.

[0062] [Table 1]

[0063] Although the embodiment of the present invention has been described in detail above, the present invention is not limited to the above embodiment, and various modifications are possible without departing from the gist of the present invention. [Industrial Applicability]

[0064] The hydrogen concentration estimation system according to the present invention can be used for estimating the hydrogen concentration in a mixed gas, detecting leaks of a mixed gas containing hydrogen, and the like.

Claims

1. A hydrogen concentration estimation system comprising the following configurations (1) to (4): (1) The hydrogen concentration estimation system is used to estimate the hydrogen concentration of a mixed gas containing hydrogen and a main gas consisting of one or more component gases having a molar mass of 10 g / mol or more. (2) The hydrogen concentration estimation system includes: first, second, ..., mth flow rate sensors (m≧2) for measuring a flow rate of the mixed gas; a storage unit that stores outputs of the first, second, ..., mth flow rate sensors when the mixed gas is caused to flow through the first, second, ..., mth flow rate sensors; an estimation unit that estimates a hydrogen concentration of the mixed gas by using any two or more of the outputs; It is equipped with: (3) The first, second, . . . mth flow rate sensors have flow rate-output characteristics different from one another. Here, the term "having different flow rate-output characteristics" means that The output Y of the i-th flow sensor (1≦i≦m) and the j-th flow sensor (1≦j≦m, j≠i) when the main gas having the hydrogen concentration of zero is passed through the i-th flow sensor (1≦i≦m) and the j-th flow sensor (1≦j≦m, j≠i) at different flow rates. i and the output Y of the jth flow sensor j The relationship is expressed as a linear equation: Y i = b x Y j In the case of approximation with +c, The increase in output when the mixed gas in which only the hydrogen concentration is changed is flowed through the i flow sensor and the j flow sensor with the flow rate of the main gas fixed, is represented by ΔY i and ΔY j Then, ΔY i / ΔY j This means that the relationship =b does not hold. (4) The estimation unit includes means for estimating the hydrogen concentration by using a function having any two or more of the outputs acquired in advance as explanatory variables and the hydrogen concentration as a response variable.

2. 2. The hydrogen concentration estimation system according to claim 1, wherein the estimation unit includes a means for calculating the hydrogen concentration using the following equation (1). (Y) 1 Y 2 ...... m ) …(1) however, n is the hydrogen concentration of the mixed gas, Y 1 , Y 2 , … Y m are the outputs of the first, second, ..., mth flow rate sensors when the mixed gas is flowed through the first, second, ..., mth flow rate sensors, respectively. F is the Y 1 , Y 2 , … Y m A multivariate nonlinear function having two or more of the above as explanatory variables.

3. The hydrogen concentration estimation system of claim 1 or 2, wherein the estimation unit includes a means for calculating the hydrogen concentration using a machine learning model that has previously machine-learned the correspondence between the hydrogen concentration and two or more of the outputs of the first, second, ..., mth flow sensors when the mixed gas is flowing.

4. 4. The hydrogen concentration estimation system according to claim 1, wherein the estimation unit includes a means for calculating the hydrogen concentration using the following equations (2) and (3). d=|bY j -Y k +c| / √(b 2 +1) …(2) n[vol%]=αd+β…(3) however, n is the hydrogen concentration of the mixed gas, Y j (1≦j≦m) is the output of the jth flow sensor when the mixed gas is flowed through the jth flow sensor, Y k (1≦k≦m, k≠j) is the output of the kth flow sensor when the mixed gas is flowed through the kth flow sensor, b and c are the outputs Y of the j flow sensor when the main gases having different flow rates and / or compositions are flowed through the j flow sensor and the k flow sensor, respectively. j ' is the x-axis, and the output Y k ' is the y-axis, j ' and the above Y k The correspondence relationship with Y k '=by j ' + c) the slope and y-intercept of the linear equation when approximated, α and β are each a constant.

5. Further comprising a temperature sensor for measuring a temperature of the mixed gas, A hydrogen concentration estimation system as described in any one of claims 1 to 4, wherein the estimation unit includes a means for estimating the hydrogen concentration using a function having two or more of the outputs and the temperature acquired in advance as explanatory variables and the hydrogen concentration as a response variable.

6. 6. The hydrogen concentration estimation system according to claim 5, wherein the estimation unit includes a means for calculating the hydrogen concentration using the following equation (4). n[vol%]=F'(Y 1 Y 2 ...... m () … () however, n is the hydrogen concentration of the mixed gas, Y 1 , Y 2 , … Y m are the outputs of the first, second, ..., mth flow rate sensors when the mixed gas is flowed through the first, second, ..., mth flow rate sensors, respectively. T is the temperature of the mixed gas; F′ is the Y 1 , Y 2 , … Y m A multivariate nonlinear function having two or more of the above and the T as explanatory variables.

7. The hydrogen concentration estimation system of claim 5 or 6, wherein the estimation unit includes a means for calculating the hydrogen concentration using a machine learning model that has previously machine-learned the correspondence between two or more of the outputs of the first, second, ..., mth flow sensors when the mixed gas is flowing, the temperature, and the hydrogen concentration.

8. 8. The hydrogen concentration estimation system according to claim 1, wherein at least one of the first, second, ..., mth flow rate sensors is a mass flow rate type flow rate sensor.

9. 9. The hydrogen concentration estimation system according to claim 1, wherein at least one of the first, second, ..., mth flow rate sensors is a thermal mass flow rate type flow rate sensor.

10. 10. The hydrogen concentration estimation system according to claim 1, wherein the mixed gas has a total content of the hydrogen and the main gas of 99 vol % or more.

Images