A method for estimating methane leakage

Abstract

The invention discloses a methane leakage estimation method, which belongs to the technical field of methane leakage estimation. The invention divides the leakage area into grids and sets data collection points in each grid to realize multi-point collection of the leakage area, and uses the collected methane concentration data to predict the methane distribution range, and fills the uncollected area, thereby overcoming the data shortage problem caused by limited sampling points in the traditional method, thereby obtaining the methane concentration in the complete methane distribution range, and estimating the future methane leakage based on the methane concentration and methane pressure in the methane distribution range based on an estimation model, and realizing the prediction of the future methane leakage based on the relationship between the methane concentration and the methane pressure, so as to give an early warning before the methane leakage exceeds the amount.

Description

Technical Field

[0001] The present invention relates to the technical field of methane leakage estimation, and in particular to a methane leakage estimation method. Background Art

[0002] Methane leakage has become an increasingly serious environmental and safety issue during oil and gas extraction. As a potent greenhouse gas, methane has a global warming potential 28-36 times that of carbon dioxide on a 100-year scale, which has a significant impact on climate change. In extraction scenarios, methane leakage not only causes the loss of valuable energy resources, but can also cause fires and explosions, threatening the lives of workers.

[0003] In the prior art, fixed data collection points are set to collect methane concentration, thereby estimating the current methane concentration leakage situation, but it is impossible to estimate the future methane leakage amount. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, a methane leakage estimation method provided by the present invention solves the problem that the prior art cannot estimate the future methane leakage.

[0005] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a methane leakage estimation method, comprising the following steps:

[0006] S1. Divide the spatial area of ​​methane leakage into multiple grid areas;

[0007] S2. Set a data collection point in each grid area to collect methane concentration;

[0008] S3. Predict the methane distribution range based on the methane concentration at each collection point;

[0009] S4. Fill the methane concentration in the blank area where no data collection point is set according to the methane distribution range;

[0010] S5. Estimate the future methane leakage based on the estimation model according to the methane concentration and methane pressure within the methane distribution range.

[0011] Furthermore, S3 includes the following sub-steps:

[0012] S31, finding the location of the maximum methane concentration in each row of the grid area as the row marker grid;

[0013] S32, obtaining the methane distribution length on the left side of the row according to the methane concentration of the grid area on the left side of the row mark grid;

[0014] S33, according to the methane concentration of the grid area on the right side of the row mark grid, obtain the methane distribution length on the right side of the row;

[0015] S34, finding the location of the maximum methane concentration in each column grid area as the column marker grid;

[0016] S35, obtaining the methane distribution length on the upper side of the row according to the methane concentration of the grid area on the upper side of the column mark grid;

[0017] S36, obtaining the methane distribution length at the lower side of the row according to the methane concentration of the grid area at the lower side of the column mark grid;

[0018] S37. Obtain the methane distribution range according to the methane distribution lengths of each row and column.

[0019] Furthermore, S32, S33, S35, and S36 all include the following sub-steps:

[0020] A1. According to the distance between other grid areas and the marked grid, the distance is used as the independent variable and the methane concentration in the grid area is used as the dependent variable to fit the methane concentration change function: y = ax + b, where y is the methane concentration, x is the distance, a is the slope, and b is the maximum methane concentration;

[0021] A2. In the methane concentration change function, y is 0, and the methane distribution length is calculated.

[0022] Furthermore, the S4 comprises the following sub-steps:

[0023] S41, filling the methane concentration of the blank area on the left where no data collection points are set according to the methane concentration change function on the left side of each row of the grid area;

[0024] S42, filling the methane concentration of the blank area on the right side where no data collection point is set according to the methane concentration change function on the right side of each row of the grid area;

[0025] S43, filling the methane concentration of the blank area on the upper side where no data collection points are set according to the methane concentration change function on the upper side of each column of the grid area;

[0026] S44, filling the methane concentration of the blank area on the lower side where no data collection points are set according to the methane concentration change function on the upper and lower sides of each column of the grid area.

[0027] Furthermore, the S5 comprises the following sub-steps:

[0028] S51, adding up all methane concentrations in the methane distribution range to obtain the methane amount;

[0029] S52. Construct a methane quantity sequence based on the methane quantity at each moment: X={x t-i}, X is the methane amount sequence, i is an integer, i is an integer from 0 to (N-1), N is the length of the sequence, x t-i is the amount of methane at the ti-th moment, and t is the current moment;

[0030] S53, according to the methane amount sequence: X={x t-i}, construct the methane change sequence: R={r t-i}, r t-i = x t-i -x t-i-1 , R is the methane variation sequence, r t-i is the methane change at the ti moment, x t-i-1 is the amount of methane at the moment ti-1;

[0031] S54. According to the methane pressure at each moment, construct a pressure sequence: P={p t-i}, where P is the pressure sequence, p t-i is the methane pressure at the moment ti;

[0032] S55, according to the pressure sequence: P = {p t-i}, construct the pressure variation sequence: S={c t-i},c t-i = p t-i -p t-i-1 , S is the pressure variation sequence, c t-i is the pressure change at the ti moment, p t-i-1 is the methane pressure at the moment ti-1;

[0033] S56, according to the methane amount sequence: X={x t-i}、Methane change sequence: R={r t-i}、Pressure sequence: P={p t-i} and pressure variation sequence: S={c t-i}, based on the estimation model, the future methane leakage is estimated.

[0034] Furthermore, the estimation model includes: a first LSTM layer, a second LSTM layer, a third LSTM layer, a fourth LSTM layer, a first fusion layer, a second fusion layer, a Concat layer, a Tanh layer, a Sigmoid layer and an output layer;

[0035] The input end of the first LSTM layer is used to input a methane amount sequence, the input end of the second LSTM layer is used to input a methane change sequence, the input end of the third LSTM layer is used to input a pressure sequence, and the input end of the fourth LSTM layer is used to input a pressure change sequence;

[0036] The first input end of the first fusion layer is connected to the output end of the first LSTM layer, and the second input end of the first fusion layer is connected to the output end of the second LSTM layer; the first input end of the second fusion layer is connected to the output end of the third LSTM layer, and the second input end of the second fusion layer is connected to the output end of the fourth LSTM layer; the input end of the Concat layer is respectively connected to the output end of the first fusion layer and the output end of the second fusion layer, and the output end of the Concat layer is respectively connected to the input end of the Tanh layer and the input end of the Sigmoid layer; the input end of the output layer is connected to the output end of the Tanh layer and the output end of the Sigmoid layer, and the output end of the output layer serves as the output end of the estimation model.

[0037] Furthermore, the input of each Cell unit in the first LSTM layer is an element x in the methane amount sequence t-i , the input of each Cell in the second LSTM layer is an element r in the methane change sequence t-i , the input of each Cell unit in the third LSTM layer is an element p in the pressure sequence t-i , the input of each Cell unit in the fourth LSTM layer is an element c in the pressure change sequence t-i .

[0038] Furthermore, the expression of the first fusion layer or the second fusion layer is: , where Y is the sequence output by the fusion layer, H1 is the sequence input by the first input terminal of the fusion layer, H2 is the sequence input by the second input terminal of the fusion layer, W1 is the weight sequence of H1, B1 is the bias sequence of H1, W2 is the weight sequence of H2, and B2 is the bias sequence of H2. is the element-wise multiplication of two sequences, Add two sequences element-wise.

[0039] Furthermore, the expression of the Tanh layer is: , where z1 is the first estimator, Tanh is the hyperbolic tangent activation function, and y j is the jth element output by the Concat layer, ω 1,j is the y in the Tanh layer j The weight of b 1,j is the y in the Tanh layer j , M is the number of elements output by the Concat layer, and j is a positive integer;

[0040] The expression of the Sigmoid layer is: , where z2 is the second estimator, Sigmoid is the S-type activation function, ω 2,j is y in the Sigmoid layer jThe weight of b 2,j is y in the Sigmoid layer j The bias.

[0041] Furthermore, the expression of the output layer is: , where h is the future methane leakage, ω z1 is the weight of z1, ω z2 is the weight of z2.

[0042] The beneficial effects of the present invention are as follows: the present invention divides the leakage area into grids and sets data collection points in each grid to realize multi-point collection of the leakage area, uses the collected methane concentration data to predict the methane distribution range, and fills the uncollected area, thereby overcoming the problem of insufficient data caused by limited sampling points in traditional methods, thereby obtaining the methane concentration over the complete methane distribution range, and based on the methane concentration and methane pressure of the methane distribution range and based on the estimation model, the future methane leakage amount is estimated, and based on the relationship between the methane concentration and the methane pressure, the future methane leakage amount is predicted, thereby giving an early warning before the methane leakage exceeds the amount. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flow chart of a methane leakage estimation method;

[0044] Figure 2 It is a schematic diagram of the division of step S2;

[0045] Figure 3 Schematic diagram of the estimation model. DETAILED DESCRIPTION

[0046] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0047] like Figure 1 As shown, a methane leakage estimation method comprises the following steps:

[0048] S1. Divide the spatial area of ​​methane leakage into multiple grid areas;

[0049] S2. Set a data collection point in each grid area to collect methane concentration;

[0050] S3. Predict the methane distribution range based on the methane concentration at each collection point;

[0051] S4. Fill the methane concentration in the blank area where no data collection point is set according to the methane distribution range;

[0052] S5. Estimate the future methane leakage based on the estimation model according to the methane concentration and methane pressure within the methane distribution range.

[0053] In this embodiment, if Figure 2 As shown, the spatial region with the highest methane concentration is selected and the spatial region is divided into multiple grid regions.

[0054] In this embodiment, S3 includes the following sub-steps:

[0055] S31, finding the location of the maximum methane concentration in each row of the grid area as the row marker grid;

[0056] S32, obtaining the methane distribution length on the left side of the row according to the methane concentration of the grid area on the left side of the row mark grid;

[0057] S33, according to the methane concentration of the grid area on the right side of the row mark grid, obtain the methane distribution length on the right side of the row;

[0058] S34, finding the location of the maximum methane concentration in each column grid area as the column marker grid;

[0059] S35, obtaining the methane distribution length on the upper side of the row according to the methane concentration of the grid area on the upper side of the column mark grid;

[0060] S36, obtaining the methane distribution length at the lower side of the row according to the methane concentration of the grid area at the lower side of the column mark grid;

[0061] S37. Obtain the methane distribution range according to the methane distribution lengths of each row and column.

[0062] like Figure 2 As shown, taking point B in the figure as an example, point B is used as the row mark grid and the column mark grid, and there are data in four directions that need to be filled.

[0063] The present invention identifies the maximum methane concentration position in each row and column as a landmark grid and calculates the methane distribution length in four directions (left, right, top, and bottom) respectively. This method can more accurately capture the spatial characteristics of methane diffusion and improve the accuracy of methane distribution range estimation. By considering the horizontal and vertical distributions separately, this method can better adapt to leakage scenarios of different shapes and complexities, including irregular or multi-point leakage.

[0064] In this embodiment, the area where multiple grid areas are located is the area with the most serious methane leakage. The methane leakage is in a diffuse state. Therefore, the methane leakage in the middle area is large. However, due to the limited number of detection sensors, the methane concentration at the edge needs to be inferred through the methane concentration change function.

[0065] In this embodiment, S32, S33, S35, and S36 all include the following sub-steps:

[0066] A1. According to the distance between other grid areas and the marked grid, the distance is used as the independent variable and the methane concentration in the grid area is used as the dependent variable to fit the methane concentration change function: y = ax + b, where y is the methane concentration, x is the distance, a is the slope, and b is the maximum methane concentration;

[0067] In step S32, A1 is to fit the methane concentration change function according to the distance between the grid area on the left side of the row mark grid and the row mark grid, with the distance as the independent variable and the methane concentration in the grid area on the left side of the row mark grid as the dependent variable; A1 is to fit the methane concentration change function according to the distance between the grid area on the right side of the row mark grid and the mark grid, with the distance as the independent variable and the methane concentration in the grid area on the right side of the row mark grid as the dependent variable; in S35, A1 is to fit the methane concentration change function according to the distance between the grid area on the upper side of the column mark grid and the column mark grid, with the distance as the independent variable and the methane concentration in the grid area on the upper side of the column mark grid as the dependent variable; in S36, A1 is to fit the methane concentration change function according to the distance between the grid area on the lower side of the column mark grid and the column mark grid, with the distance as the independent variable and the methane concentration in the grid area on the lower side of the column mark grid as the dependent variable;

[0068] A2. In the methane concentration change function, y is 0, and the methane distribution length is calculated.

[0069] Methane distribution length: can be obtained by calculating the x value in the function when y = 0. This length represents the distance from the maximum concentration to the point where the methane concentration drops to 0.

[0070] The present invention achieves accurate quantification of the methane diffusion pattern by expressing the relationship between methane concentration and distance as a linear function y=ax+b. By calculating the distance when the methane concentration is zero, the boundary of the methane distribution can be accurately defined.

[0071] At an initial concentration (b) of 100 ppm and a slope (a) of -2 ppm / dm, the concentration decreases rapidly with distance, indicating a stronger but smaller leak source. The methane concentration function is y = -2x + 100.

[0072] When the initial concentration (b) is 80 ppm and the slope (a) is -1 ppm / dm, the concentration decrease rate is moderate, representing a medium-intensity leakage source. The methane concentration change function is y = -x + 80.

[0073] When the initial concentration (b) is 60 ppm and the slope (a) is -0.5 ppm / dm, the concentration decreases slowly, which may indicate a weak but continuous leakage source, or a situation where it diffuses over a larger area. The methane concentration change function is: y = -0.5x + 60.

[0074] In this embodiment, S4 includes the following sub-steps:

[0075] S41, filling the methane concentration of the blank area on the left where no data collection points are set according to the methane concentration change function on the left side of each row of the grid area;

[0076] S42, filling the methane concentration of the blank area on the right side where no data collection point is set according to the methane concentration change function on the right side of each row of the grid area;

[0077] S43, filling the methane concentration of the blank area on the upper side where no data collection points are set according to the methane concentration change function on the upper side of each column of the grid area;

[0078] S44, filling the methane concentration of the blank area on the lower side where no data collection points are set according to the methane concentration change function on the upper and lower sides of each column of the grid area.

[0079] Taking step S41 as an example, the filling method is specifically described: in this embodiment, the blank area is also filled in a way of one grid area length, which is consistent with the division method. For the methane concentration change function on the left side of the row: y = ax + b, the farthest distance between the grid area on the left and the row mark grid is d, and the grid areas on the left are set to be 5 in total. Therefore, the first filling position of the blank area of ​​the row is d + d / 5, and the nth filling position is d + n (d / 5), until the boundary is reached, n is the number of the filling position, so that the methane concentration after filling is evenly distributed in the methane distribution range, such as Figure 2 As shown, the distances from B1 to B, from B2 to B1, from B3 to B2, and from B4 to B3 are all the same, and the points of the filling position are B3 and B4.

[0080] In this embodiment, the filling method may also be other methods, not limited to the filling method proposed in this embodiment, the purpose is to make the methane concentration after filling and the collected methane concentration reflect the methane concentration of each local area within the methane distribution range.

[0081] The present invention starts from each marking grid and fills the blank areas in the grid area where no data collection points are set from four directions (left, right, top and bottom), thereby ensuring the comprehensive coverage and integrity of the methane concentration distribution data, and also improving the accuracy of the data by utilizing the concentration change function in a specific direction.

[0082] In the present invention, according to the methane concentration change function, the blank area on one side of a row or a column can be filled until the boundary of the methane distribution range. Taking the methane concentration change function: y = -2x + 100 as an example, when y=0, x=50, then in the row or column, the boundary on one side is 50dm. In the blank area, the distance of the blank area is brought into the methane concentration change function to obtain the methane concentration of the filling position point.

[0083] In this embodiment, S5 includes the following sub-steps:

[0084] S51, adding up all methane concentrations in the methane distribution range to obtain the methane amount;

[0085] In this embodiment, the present invention fills S4, which is equivalent to obtaining the methane concentration at each point on a surface, and the surface is the methane distribution range;

[0086] In this embodiment, step S51 may also take the average of all methane concentrations as the methane amount;

[0087] S52. Construct a methane quantity sequence based on the methane quantity at each moment: X={x t-i}, X is the methane amount sequence, i is an integer, i is an integer from 0 to (N-1), N is the length of the sequence, x t-i is the amount of methane at the ti-th moment, and t is the current moment;

[0088] S53, according to the methane amount sequence: X={x t-i}, construct the methane change sequence: R={r t-i}, r t-i = x t-i -x t-i-1 , R is the methane variation sequence, r t-i is the methane change at the ti moment, x t-i-1 is the amount of methane at the moment ti-1;

[0089] S54. According to the methane pressure at each moment, construct a pressure sequence: P={p t-i}, where P is the pressure sequence, p t-i is the methane pressure at the moment ti;

[0090] S55, according to the pressure sequence: P = {p t-i}, construct the pressure variation sequence: S={c t-i},c t-i = p t-i -p t-i-1 , S is the pressure variation sequence, c t-i is the pressure change at the ti moment, p t-i-1 is the methane pressure at the moment ti-1;

[0091] S56, according to the methane amount sequence: X={x t-i}、Methane change sequence: R={r t-i}、Pressure sequence: P={p t-i} and pressure variation sequence: S={c t-i}, based on the estimation model, the future methane leakage is estimated.

[0092] In this embodiment, the methane pressure is the leak point pressure, which refers to the pressure at the actual leak point of methane. Due to the pressure loss caused by the leak, this pressure is usually lower than the leak source pressure. It can be measured by installing a pressure sensor near the leak point.

[0093] When methane leaks, the change in the pressure at the leak point often directly reflects the dynamic conditions at the leak point, and can reflect the change in the leakage intensity of methane. Therefore, based on the pressure sequence, a pressure change sequence is constructed. The pressure change sequence reflects the methane leakage at the leak point. Based on the methane amount sequence, a methane change sequence is constructed. The methane amount sequence reflects the methane leakage at various historical moments. The methane change sequence reflects the change in methane leakage at various historical moments. Therefore, an estimation model is used to process the methane amount sequence, methane change sequence, pressure sequence and pressure change sequence to improve the estimation accuracy of future methane leakage. This accurate prediction enables managers to take targeted measures in a timely manner, such as strengthening inspections in specific areas or implementing emergency repairs, thereby effectively preventing large-scale leakage accidents and significantly reducing environmental risks and economic losses.

[0094] In this embodiment, if Figure 3 As shown, the estimation model includes: a first LSTM layer, a second LSTM layer, a third LSTM layer, a fourth LSTM layer, a first fusion layer, a second fusion layer, a Concat layer, a Tanh layer, a Sigmoid layer and an output layer;

[0095] The input end of the first LSTM layer is used to input a methane amount sequence, the input end of the second LSTM layer is used to input a methane change sequence, the input end of the third LSTM layer is used to input a pressure sequence, and the input end of the fourth LSTM layer is used to input a pressure change sequence;

[0096] The first input end of the first fusion layer is connected to the output end of the first LSTM layer, and the second input end of the first fusion layer is connected to the output end of the second LSTM layer; the first input end of the second fusion layer is connected to the output end of the third LSTM layer, and the second input end of the second fusion layer is connected to the output end of the fourth LSTM layer; the input end of the Concat layer is respectively connected to the output end of the first fusion layer and the output end of the second fusion layer, and the output end of the Concat layer is respectively connected to the input end of the Tanh layer and the input end of the Sigmoid layer; the input end of the output layer is connected to the output end of the Tanh layer and the output end of the Sigmoid layer, and the output end of the output layer serves as the output end of the estimation model.

[0097] The present invention uses multiple LSTM layers to process the methane amount, methane change, pressure and pressure change sequence respectively, and the model can effectively capture the time dependency and long-term pattern of each sequence. The introduction of the fusion layer allows the model to learn the relationship between the methane amount and its change, and the relationship between the pressure and its change, to achieve the fusion of the characteristic value of the methane amount and the characteristic value of its change, and to achieve the fusion of the characteristic value of the pressure and the characteristic value of its change, and then use the Concat layer to combine the output of the first fusion layer and the output of the second fusion layer, use two activation functions, and use two paths for prediction, and then use the output layer to combine the two prediction results to improve the prediction accuracy.

[0098] The input of each Cell unit in the first LSTM layer is an element x in the methane amount sequence t-i , the input of each Cell in the second LSTM layer is an element r in the methane change sequence t-i , the input of each Cell unit in the third LSTM layer is an element p in the pressure sequence t-i , the input of each Cell unit in the fourth LSTM layer is an element c in the pressure change sequence t-i .

[0099] In the present invention, one LSTM layer processes one sequence, and by separating different types of sequences, the LSTM layer can independently extract features from each sequence, avoiding direct interference between different types of data. This ensures that the features of each sequence can be fully learned and represented.

[0100] The expression of the first fusion layer or the second fusion layer is: , where Y is the sequence output by the fusion layer, H1 is the sequence input by the first input terminal of the fusion layer, H2 is the sequence input by the second input terminal of the fusion layer, W1 is the weight sequence of H1, B1 is the bias sequence of H1, W2 is the weight sequence of H2, and B2 is the bias sequence of H2. is the element-wise multiplication of two sequences, Add two sequences element-wise.

[0101] The present invention can effectively integrate information from different sources and capture the relationship between them by weighted fusion of two input sequences (H1 and H2).

[0102] In this embodiment, the expression of the Tanh layer is: , where z1 is the first estimator, Tanh is the hyperbolic tangent activation function, and y j is the jth element output by the Concat layer, ω 1,j is the y in the Tanh layer j The weight of b 1,j is the y in the Tanh layer j , M is the number of elements output by the Concat layer, and j is a positive integer;

[0103] In this embodiment, the expression of the Sigmoid layer is: , where z2 is the second estimator, Sigmoid is the S-type activation function, ω 2,j is y in the Sigmoid layer j The weight of b 2,j is y in the Sigmoid layer j The bias.

[0104] In this embodiment, the expression of the output layer is: , where h is the future methane leakage, ω z1 is the weight of z1, ω z2 is the weight of z2.

[0105] In this embodiment, the estimation model can be trained by a gradient descent method to obtain weights and biases in the estimation model.

[0106] The present invention divides the leakage area into grids and sets data collection points in each grid to achieve multi-point collection of the leakage area, uses the collected methane concentration data to predict the methane distribution range, and fills the uncollected area, thereby overcoming the problem of insufficient data caused by limited sampling points in traditional methods, thereby obtaining the methane concentration over the complete methane distribution range, and estimating the future methane leakage based on the methane concentration and methane pressure in the methane distribution range based on an estimation model, and based on the relationship between methane concentration and methane pressure, the future methane leakage is predicted, thereby giving an early warning before the methane leakage exceeds the limit.

[0107] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A methane leakage estimation method, characterized in that: The following steps are involved: S1. Divide the spatial area of ​​methane leakage into multiple grid areas; S2. Set a data collection point in each grid area to collect methane concentration; S3. Predict the methane distribution range based on the methane concentration at each collection point; S4. Fill the methane concentration in the blank area where no data collection point is set according to the methane distribution range; S5. Estimate the future methane leakage based on the estimation model according to the methane concentration and methane pressure within the methane distribution range; The S3 comprises the following sub-steps: S31, finding the location of the maximum methane concentration in each row of the grid area as the row marker grid; S32, obtaining the methane distribution length on the left side of the row according to the methane concentration of the grid area on the left side of the row mark grid; S33, according to the methane concentration of the grid area on the right side of the row mark grid, obtain the methane distribution length on the right side of the row; S34, finding the location of the maximum methane concentration in each column grid area as the column marker grid; S35, obtaining the methane distribution length on the upper side of the row according to the methane concentration of the grid area on the upper side of the column mark grid; S36, obtaining the methane distribution length at the lower side of the row according to the methane concentration of the grid area at the lower side of the column mark grid; S37, obtaining a methane distribution range according to the methane distribution lengths of each row and column; S32, S33, S35, and S36 all include the following sub-steps: A1. According to the distance between other grid areas and the marked grid, the distance is used as the independent variable and the methane concentration in the grid area is used as the dependent variable to fit the methane concentration change function: y = ax + b, where y is the methane concentration, x is the distance, a is the slope, and b is the maximum methane concentration; A2. In the methane concentration change function, y is 0, and the methane distribution length is calculated; The S4 comprises the following sub-steps: S41, filling the methane concentration of the blank area on the left where no data collection points are set according to the methane concentration change function on the left side of each row of the grid area; S42, filling the methane concentration of the blank area on the right side where no data collection point is set according to the methane concentration change function on the right side of each row of the grid area; S43, filling the methane concentration of the blank area on the upper side where no data collection points are set according to the methane concentration change function on the upper side of each column of the grid area; S44, filling the methane concentration of the blank area on the lower side where no data collection point is set according to the methane concentration change function on the upper and lower sides of each column of the grid area; The S5 comprises the following sub-steps: S51, adding up all methane concentrations in the methane distribution range to obtain the methane amount; S52. Construct a methane quantity sequence based on the methane quantity at each moment: X={x t-i }, X is the methane amount sequence, i is an integer, i is an integer from 0 to (N-1), N is the length of the sequence, x t-i is the amount of methane at the ti-th moment, and t is the current moment; S53, according to the methane amount sequence: X={x t-i }, construct the methane change sequence: R={r t-i }, r t-i = x t-i -x t-i-1 , R is the methane variation sequence, r t-i is the methane change at the ti moment, x t-i-1 is the amount of methane at the moment ti-1; S54. According to the methane pressure at each moment, construct a pressure sequence: P={p t-i }, where P is the pressure sequence, p t-i is the methane pressure at the moment ti; S55, according to the pressure sequence: P = {p t-i }, construct the pressure variation sequence: S={c t-i },c t-i = p t-i -p t-i-1 , S is the pressure variation sequence, c t-i is the pressure change at the ti moment, p t-i-1 is the methane pressure at the moment ti-1; S56, according to the methane amount sequence: X={x t-i }、Methane change sequence: R={r t-i }、Pressure sequence: P={p t-i } and pressure variation sequence: S={c t-i }, based on the estimation model, estimate the future methane leakage; The estimation model includes: a first LSTM layer, a second LSTM layer, a third LSTM layer, a fourth LSTM layer, a first fusion layer, a second fusion layer, a Concat layer, a Tanh layer, a Sigmoid layer and an output layer; The input end of the first LSTM layer is used to input a methane amount sequence, the input end of the second LSTM layer is used to input a methane change sequence, the input end of the third LSTM layer is used to input a pressure sequence, and the input end of the fourth LSTM layer is used to input a pressure change sequence; The first input end of the first fusion layer is connected to the output end of the first LSTM layer, and the second input end thereof is connected to the output end of the second LSTM layer; the first input end of the second fusion layer is connected to the output end of the third LSTM layer, and the second input end thereof is connected to the output end of the fourth LSTM layer; the input end of the Concat layer is respectively connected to the output end of the first fusion layer and the output end of the second fusion layer, and the output end thereof is respectively connected to the input end of the Tanh layer and the input end of the Sigmoid layer; the input end of the output layer is connected to the output end of the Tanh layer and the output end of the Sigmoid layer, and the output end thereof serves as the output end of the estimation model; The input of each Cell unit in the first LSTM layer is an element x in the methane amount sequence t-i , the input of each Cell in the second LSTM layer is an element r in the methane change sequence t-i , the input of each Cell unit in the third LSTM layer is an element p in the pressure sequence t-i , the input of each Cell unit in the fourth LSTM layer is an element c in the pressure change sequence t-i ; The expression of the first fusion layer or the second fusion layer is: , where Y is the sequence output by the fusion layer, H1 is the sequence input by the first input terminal of the fusion layer, H2 is the sequence input by the second input terminal of the fusion layer, W1 is the weight sequence of H1, B1 is the bias sequence of H1, W2 is the weight sequence of H2, and B2 is the bias sequence of H2. is the element-wise multiplication of two sequences, Add two sequences element-wise; The expression of the Tanh layer is: , where z1 is the first estimator, Tanh is the hyperbolic tangent activation function, and y j is the jth element output by the Concat layer, ω 1,j is the y in the Tanh layer j The weight of b 1,j is the y in the Tanh layer j , M is the number of elements output by the Concat layer, and j is a positive integer; The expression of the Sigmoid layer is: , where z2 is the second estimator, Sigmoid is the S-type activation function, ω 2,j is y in the Sigmoid layer j The weight of b 2,j is y in the Sigmoid layer j Bias of The expression of the output layer is: , where h is the future methane leakage, ω z1 is the weight of z1, ω z2 is the weight of z2.

Images