Oil shale pyrolysis control system and method

Abstract

The invention relates to an oil shale pyrolysis control system and method. The system comprises a pyrolysis furnace, a controller, an electronic nose, a methane gas storage tank, a gas separator, a gas collection tank, a washing tower, a gas valve, a methane gas path and a mixed gas path, wherein the pyrolyzing furnace comprises a feeding hole, a smoke outlet, a waste outlet, a preheating chamber, a dry distillation chamber, a combustion chamber, a cooling chamber, an annular pipeline, an air inlet and a temperature sensor; according to the method, gas collected by an electronic nose is subjected to component recognition through an RNN-SVM model, the complex nonlinear relation between gas data and the temperature is learned, more accurate and rapid oil shale pyrolysis process control is achieved, and therefore more efficient energy utilization is achieved.

Description

Technical Field

[0001] The invention relates to the technical field of oil shale pyrolysis, and in particular to an oil shale pyrolysis control system and method. Background Art

[0002] Oil shale pyrolysis is a complex chemical reaction process, which usually includes multiple reactions such as cracking, gasification and recombination, and the pyrolysis temperature is in the range of 350℃–550℃. This is an important unconventional oil and gas extraction method that can effectively utilize unconventional oil and gas resources such as shale.

[0003] Temperature control of oil shale pyrolysis is a key technical link, and the effect of temperature control directly affects the yield and quality of the reaction. In the traditional oil shale pyrolysis process, temperature control is usually achieved through a preset fixed temperature or a rough temperature curve, which has problems such as low temperature control accuracy, low pyrolysis efficiency, high energy consumption, and unstable product quality. In order to solve these problems, in recent years, people have begun to explore the use of advanced sensor technology to monitor the pyrolysis state of oil shale in real time and accurately control the pyrolysis temperature to improve pyrolysis efficiency and product quality.

[0004] The present invention proposes an oil shale pyrolysis control system and method, which uses an electronic nose to monitor and record the gas composition data generated during the pyrolysis process of oil shale in real time, designs a RNN-SVM data analysis model combining a recurrent neural network (RNN) and a support vector machine (SVM), judges the pyrolysis maturity of oil shale, and realizes accurate and rapid control of the pyrolysis process of oil shale. Summary of the invention

[0005] The present invention proposes an oil shale pyrolysis control system and method, which uses an RNN-SVM model to identify the components of the gas collected by an electronic nose, learns the complex nonlinear relationship between gas data and temperature, and realizes accurate and rapid oil shale pyrolysis process control, thereby improving pyrolysis efficiency and product quality.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An oil shale pyrolysis control system, the system comprising: a pyrolysis furnace, a controller, an electronic nose, a methane gas storage tank, a gas separator, a gas collecting tank, a washing tower, a gas valve, a methane gas path and a mixed gas path;

[0008] The pyrolysis furnace includes: a feed inlet, a smoke outlet, a waste outlet, a preheating chamber, a dry distillation chamber, a combustion chamber, a cooling chamber, a ring duct, an air inlet and a temperature sensor;

[0009] The shale particles are heated and dried by heat conduction between the cooling chamber and the preheating chamber. The dried shale particles and the shale ash from the combustion chamber are mixed and heat-transferred in the retorting chamber to reach the optimum temperature of 505°C required for shale retorting. The shale semi-coke produced after retorting enters the combustion chamber through an annular pipe (the annular structure can effectively utilize the space and provide a larger heat exchange area), and is burned in the combustion chamber with the air entering through the air inlet to make the temperature of the shale ash reach 740°C. The smoke generated by the combustion is discharged from the smoke outlet outside the furnace, and part of the shale ash enters the cooling chamber for cooling, and at the same time, part of the heat is transferred to the preheating chamber through heat conduction; the gas collecting tank and the retorting chamber are connected to the retorting chamber. The chamber is connected, the hydrocarbon vapor and other gases produced by dry distillation are discharged into the gas collecting tank, and a part of them are introduced into the electronic nose. The electronic nose detects gas characteristics through its sensor array, identifies gas components and sends data to the controller. The washing tower is connected to the gas collecting tank and is used to wash away impurities in the remaining gas in the gas collecting tank. The washing tower is connected to the gas separator through a mixed gas path. The gas separator is used to separate the combustible gas methane. The gas separator is connected to the methane gas storage tank, and the methane gas storage tank provides auxiliary fuel methane to the combustion chamber through a methane gas path and a gas valve; the controller determines the opening and closing of the gas valve according to the signal fed back by the temperature sensor.

[0010] As a further technical solution of the present invention, the pyrolysis furnace adopts a horizontal rotary kiln structure, the furnace body is slightly inclined, and contains two concentric cylinders, which are divided into four chambers: a preheating chamber, a distillation chamber, a combustion chamber, and a cooling chamber. The chambers are isolated by sealed channels, and the sealed channels can control the gas flow between the chambers.

[0011] Another object of the present invention is to provide an oil shale pyrolysis control method, which is implemented based on the above-mentioned oil shale pyrolysis control system and includes the following steps:

[0012] Step 1: Place oil shale particles (<10 mm) into a pyrolysis furnace through a feed port;

[0013] Step 2: heating the pyrolysis furnace according to a preset program to achieve pyrolysis of the oil shale;

[0014] Step 3: During the pyrolysis process, volatile organic compounds are volatilized from the oil shale, and the volatile organic compounds are collected in a gas collecting tank;

[0015] Step 4: The gas in the gas collecting tank is introduced into the electronic nose through the gas pipeline. The electronic nose detects gas characteristics through its sensor array, identifies gas components and sends data to the controller;

[0016] Step 5: The controller identifies the volatile components in the oil shale by analyzing the output data of the electronic nose, identifies the gas composition ratio by building an RNN model, and determines the pyrolysis maturity of the oil shale;

[0017] Step 6: Use the SVM model to learn the complex nonlinear relationship between gas data and temperature to control the pyrolysis temperature, thereby accurately controlling the pyrolysis process of oil shale and improving production efficiency.

[0018] As a further technical solution of the present invention, in the step 2, the initial temperature of the pyrolysis furnace is set according to the oil shale of different composition and origin; the oil shale and hot rock ash are mixed and heated to 505°C to produce shale oil, gas and shale semi-coke, the oil and gas are extracted and condensed to generate shale oil and high calorific value dry distillation gas, the shale semi-coke enters the combustion chamber through an annular pipe, burns in the air to become shale ash, part of the hot shale ash returns to the dry distillation chamber through a circulation system as a solid heat carrier to heat the oil shale, forming a cycle, and the other part of the hot shale ash is discharged from the cooling chamber, and the preheating chamber is heated during the discharge process.

[0019] As a further technical solution of the present invention, in step three, the volatile organic compounds generated by pyrolysis of oil shale include carbon monoxide, carbon dioxide, methane, and hydrocarbon gases.

[0020] As a further technical solution of the present invention, in step 4, the electronic nose is composed of 16 gas sensors, and the sensors are in contact with the gas products to generate voltage signals, which are received and converted into digital signals by the acquisition card, and the data are sent to the controller, including the following specific steps:

[0021] S1. Collect gas composition data and build a structured dataset;

[0022] S2. Calibrate the collected data values;

[0023] S3. Normalize the data, the formula is as follows:

[0024]

[0025] Where x' is the standardized value, μ is the mean value of the data, and σ is the variance of the data.

[0026] As a further technical solution of the present invention, the step five is to construct an RNN model to process the gas data collected by the electronic nose, so as to identify the gas composition ratio and judge the pyrolysis maturity of the oil shale, including the following specific steps:

[0027] S1. Construct an RNN model, use gas composition data for training, and save the identification parameters of the oil shale pyrolysis stage after training until the results converge;

[0028] S2. Use the saved RNN model to predict the stage of oil shale pyrolysis and obtain the accurate value of the composition ratio of pyrolysis gas products;

[0029] S3. The RNN model includes parameters related to the number of hidden layers, the number of hidden neural nodes, the activation function, the optimizer, the learning rate, the random inactivation layer and the number of iterations and the loss function training model;

[0030] The expression of the RNN model is as follows:

[0031] (1) Forward propagation:

[0032] a (t) =b+Wh (t-1) +Ux (t) ;

[0033] h (t) =tanh(a (t) );

[0034] o (t) =c+Vh (t) ;

[0035] y^ (t) =softmax(o (t) );

[0036] a (t) represents the hidden state at the tth time step, x (t) represents the input of the tth time step, y^ (t) represents the output of the t-th time step, b and c represent the bias vector and weight matrix respectively, U, V and W correspond to the connections from input to hidden, hidden to output and hidden to hidden respectively, and tanh is the hidden layer activation function;

[0037] (2) Loss function:

[0038] E t (y t ,y^ t )=-y t logy^ t ;

[0039] E(y,y^)=∑ t E t (y t ,y^ t )-∑ t y t logy^ t ;

[0040] The loss function of an element: called the cross entropy loss function (Cross Entropy Loss);

[0041] (3) Reverse calculation:

[0042] Parameter optimization method: Gradient descent, same as traditional neural networks;

[0043] Calculate the derivative of the loss function with respect to the parameters;

[0044] Each output has an effect on the parameters;

[0045] The derivatives of the parameters are the outputs The sum of derivatives with respect to the parameters;

[0046] Solutions to gradient disappearance and gradient explosion:

[0047] 1) Use Leaky ReLU activation function:

[0048] f(x) = max(αx, x);

[0049] α is a small constant;

[0050] 2) Gradient clipping and regularization: This is mainly proposed to address gradient explosion. The idea is to set a gradient clipping threshold, and then when updating the gradient, if the gradient exceeds this threshold, it will be forced to be limited within this range.

[0051] As a further technical solution of the present invention, the step of controlling the pyrolysis temperature by learning the complex nonlinear relationship between gas data and temperature through the SVM model includes:

[0052] Construct an SVM model and use known gas data and temperature data to train the SVM model. The SVM model learns the complex nonlinear relationship between gas data and temperature.

[0053] The SVM model is as follows:

[0054] Given a training set D containing N samples, D = {(x1, y1), (x2, y2), ..., (x N ,y N ),}, where x i represents the feature vector of the i-th sample, y i Represents the category label of the i-th sample; feature vector x i Has n dimensions, namely x i =(x 1i ,x 2i ,…,x ni );

[0055] The hyperplane is expressed as follows:

[0056] w·x+b=0;

[0057] w represents the normal vector of the hyperplane, which determines the direction of the hyperplane, b represents the intercept of the hyperplane, and w·x represents the inner product of w and x;

[0058] The goal is to find w and b so that all sample points satisfy:

[0059] y i (w·x i +b)≥1;

[0060] At the same time, the boundary should be maximized, so the optimization problem can be expressed as:

[0061] min1 / 2‖w‖ 2 ;

[0062] In order to solve the outliers, the slack variable (x i ,ε i ), so the optimization problem becomes:

[0063] min1 / 2‖w‖ 2 +C∑ε i ;

[0064] The constraints are:

[0065] y i (w·x i +b)≥1-ε i ,i=1,2,…,N;

[0066] ε i ≥0,i=1,2,…,N;

[0067] ‖w‖ represents the norm of vector w, y i Represents the category label of the i-th sample point, and C is a regularization parameter;

[0068] For linearly inseparable data, the kernel function is introduced to map the sample points from the original feature space to a high-dimensional feature space. The kernel function is expressed as:

[0069]

[0070] x i , x j is a sample point in the input space, It is a function that maps sample points to feature space.

[0071] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention designs an oil shale pyrolysis control system and method, which uses the RNN-SVM model to identify the components of the gas collected by the electronic nose, learns the complex nonlinear relationship between gas data and temperature, and realizes more accurate and rapid oil shale pyrolysis process control, thereby achieving more efficient energy utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1The figure is a schematic diagram of the structure of an oil shale pyrolysis control system.

[0073] Figure 2 A flow chart for implementing an oil shale pyrolysis control method.

[0074] Notes on figure numbers: 1-feed inlet, 2-smoke exhaust outlet, 3-waste outlet, 4-preheating chamber, 5-retorting chamber, 6-combustion chamber, 7-cooling chamber, 8-annular duct, 9-air inlet, 10-temperature sensor, 11-controller, 12-electronic nose, 13-methane gas storage tank, 14-gas separator, 15-gas collecting tank, 16-washing tower, 17-gas valve, 18-methane gas path, 19-mixing gas path. DETAILED DESCRIPTION

[0075] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0076] See also Figure 1 , an embodiment of the present invention provides an oil shale pyrolysis control system, the system comprising: a pyrolysis furnace, a controller 11, an electronic nose 12, a methane gas storage tank 13, a gas separator 14, a gas collecting tank 15, a washing tower 16, a gas valve 17, a methane gas path 18 and a mixed gas path 19;

[0077] The pyrolysis furnace includes: a feed port 1, a smoke outlet 2, a waste outlet 3, a preheating chamber 4, a dry distillation chamber 5, a combustion chamber 6, a cooling chamber 7, a ring duct 8, an air inlet 9 and a temperature sensor 10;

[0078] The shale particles are heated and dried by heat conduction between the cooling chamber 7 and the preheating chamber 4. The dried shale particles and the shale ash from the combustion chamber 6 are mixed and heat-transferred in the distillation chamber 5 to reach the optimal temperature of 505°C required for shale distillation. The shale semi-coke produced after distillation enters the combustion chamber 6 through the annular pipe 8 (the annular structure can effectively utilize the space and provide a larger heat exchange area), and is burned in the combustion chamber 6 with the air entering through the air inlet 9 to make the temperature of the shale ash reach 740°C. The smoke generated by the combustion is discharged out of the furnace from the smoke exhaust port 2, and part of the shale ash enters the cooling chamber 7 for cooling, while part of the heat is transferred to the preheating chamber 4 through heat conduction; the gas collecting tank 15 is connected to the distillation chamber 5, and the hydrocarbon vapor generated by the distillation and other gases are discharged into the gas collecting tank 15, and a part of them is introduced into the electronic nose 12. The electronic nose 12 detects gas characteristics through its sensor array, identifies gas components and sends data to the controller 11. The washing tower 16 is connected to the gas collecting tank 15, and is used to wash away impurities in the remaining gas in the gas collecting tank 15. The washing tower 16 is connected to the gas separator 14 through the mixed gas path 19. The gas separator 14 is used to separate the combustible gas methane. The gas separator 14 is connected to the methane gas storage tank 13. The methane gas storage tank 13 provides auxiliary fuel methane to the combustion chamber 6 through the methane gas path 18 and the gas valve 17; the controller 11 determines the opening and closing of the gas valve 17 according to the signal fed back by the temperature sensor 10.

[0079] In this embodiment, the pyrolysis furnace adopts a horizontal rotary kiln structure with a slightly inclined furnace body. It contains two concentric cylinders and is divided into four chambers: a preheating chamber 4, a distillation chamber 5, a combustion chamber 6, and a cooling chamber 7. Each chamber is isolated by a sealed channel, which can control the gas flow between the chambers.

[0080] See also Figure 2 Another object of the embodiment of the present invention is to provide an oil shale pyrolysis control method, the control method is implemented based on the above-mentioned oil shale pyrolysis control system, and the control method includes the following steps:

[0081] Step 1: Place oil shale particles (<10 mm) into the pyrolysis furnace through the feed port 1. Under the action of the rotating motion of the cylinder, the oil shale particles roll forward along the cylinder wall, so that the pyrolysis can be fully carried out;

[0082] Step 2: heating the pyrolysis furnace according to a preset program to achieve pyrolysis of the oil shale;

[0083] In the step 2, the initial temperature of the pyrolysis furnace is set according to the oil shale of different composition and origin; the oil shale and hot rock ash are mixed and heated to 505° C. to produce shale oil, gas and shale semi-coke, the oil and gas are led out, condensed to generate shale oil and high calorific value dry distillation gas, the shale semi-coke enters the combustion chamber 6 through the annular pipe 8, burns in the air to become shale ash, part of the hot shale ash returns to the dry distillation chamber 5 through the circulation system as a solid heat carrier to heat the oil shale, forming a cycle, and the other part of the hot shale ash is discharged from the cooling chamber 7, and the preheating chamber 4 is heated during the discharge process;

[0084] Step 3: During the pyrolysis process, volatile organic compounds are volatilized from the oil shale and collected in a gas collecting tank 15; the volatile organic compounds include carbon monoxide, carbon dioxide, methane, and hydrocarbon gases.

[0085] Step 4: The gas in the gas collecting tank 15 is introduced into the electronic nose 12 through the gas pipeline. The electronic nose 12 detects gas characteristics through its sensor array, identifies gas components and sends data to the controller 11;

[0086] In step 4, the electronic nose 12 is composed of 16 gas sensors, the specific models are MP135, MP905, MP-7, MP-3B, MP-2, MP503, MP402, TGS2600, TGS2602, TGS2603, TGS2610, TGS2611, TGS2612, TGS2620, WSP1110, WSP2110, the sensor contacts the gas product to generate a voltage signal, which is received and converted into a digital signal by the acquisition card, and the data is sent to the controller 11, including the following specific steps:

[0087] S1. Collect gas composition data and build a structured dataset;

[0088] S2. Calibrate the collected data values;

[0089] S3. Normalize the data, the formula is as follows:

[0090]

[0091] Where x' is the standardized value, μ is the mean value of the data, and σ is the variance of the data.

[0092] Step 5: The controller 11 identifies the volatile components in the oil shale by analyzing the output data of the electronic nose 12, identifies the gas composition ratio by building an RNN model, and determines the pyrolysis maturity of the oil shale;

[0093] The step five is to construct an RNN model to process the gas data collected by the electronic nose 12, so as to identify the gas composition ratio and determine the pyrolysis maturity of the oil shale, including the following specific steps:

[0094] S1. Construct an RNN model, use gas composition data for training, and save the identification parameters of the oil shale pyrolysis stage after training until the results converge;

[0095] S2. Use the saved RNN model to predict the stage of oil shale pyrolysis and obtain the accurate value of the composition ratio of pyrolysis gas products;

[0096] S3. The RNN model includes parameters related to the number of hidden layers, the number of hidden neural nodes, the activation function, the optimizer, the learning rate, the random inactivation layer and the number of iterations and the loss function training model;

[0097] The expression of the RNN model is as follows:

[0098] (1) Forward propagation:

[0099] a (t) =b+Wh (t-1) +Ux (t) ;

[0100] h (t) =tanh(a (t) );

[0101] o (t) =c+Vh (t) ;

[0102] y^ (t) =softmax(o (t) );

[0103] a (t) represents the hidden state at the tth time step, x (t) represents the input of the tth time step, y^ (t) represents the output of the t-th time step, b and c represent the bias vector and weight matrix respectively, U, V and W correspond to the connections from input to hidden, hidden to output and hidden to hidden respectively, and tanh is the hidden layer activation function;

[0104] (2) Loss function:

[0105] E t (y t ,y^ t )=-y t logy^ t ;

[0106] E(y,y^)=∑ t E t(y t ,y^ t )-∑ t y t logy^ t ;

[0107] The loss function of an element: called the cross entropy loss function (Cross Entropy Loss);

[0108] (3) Reverse calculation:

[0109] Parameter optimization method: Gradient descent, same as traditional neural networks;

[0110] Calculate the derivative of the loss function with respect to the parameters;

[0111] Each output has an effect on the parameters;

[0112] The derivatives of the parameters are the outputs The sum of derivatives with respect to the parameters;

[0113] Solutions to gradient disappearance and gradient explosion:

[0114] 1) Use Leaky ReLU activation function:

[0115] f(x) = max(αx, x);

[0116] α is a small constant, usually taken as 0.01;

[0117] 2) Gradient clipping and regularization: This is mainly proposed to address gradient explosion. The idea is to set a gradient clipping threshold, and then when updating the gradient, if the gradient exceeds this threshold, it will be forced to be limited within this range.

[0118] Step 6: Use the SVM model to learn the complex nonlinear relationship between gas data and temperature to control the pyrolysis temperature, thereby accurately controlling the pyrolysis process of oil shale and improving production efficiency.

[0119] Construct an SVM model and use known gas data and temperature data to train the SVM model. The SVM model learns the complex nonlinear relationship between gas data and temperature.

[0120] The SVM model is as follows:

[0121] Given a training set D containing N samples, D = {(x1, y1), (x2, y2), ..., (x N ,y N ),}, where x i represents the feature vector of the i-th sample, y i Represents the category label of the i-th sample; feature vector xi Has n dimensions, namely x i =(x 1i ,x 2i ,…,x ni );

[0122] The hyperplane is expressed as follows:

[0123] w·x+b=0;

[0124] w represents the normal vector of the hyperplane, which determines the direction of the hyperplane, b represents the intercept of the hyperplane, and w·x represents the inner product of w and x;

[0125] The goal is to find w and b so that all sample points satisfy:

[0126] y i (w·x i +b)≥1;

[0127] At the same time, the boundary should be maximized, so the optimization problem can be expressed as:

[0128] min1 / 2‖w‖ 2 ;

[0129] In order to solve the outliers, the slack variable (x i ,ε i ), so the optimization problem becomes:

[0130] min1 / 2‖w‖ 2 +C∑ε i ;

[0131] The constraints are:

[0132] y i (w·x i +b)≥1-ε i ,i=1,2,…,N;

[0133] ε i ≥0,i=1,2,…,N;

[0134] ‖w‖ represents the norm of vector w, y i Represents the category label of the i-th sample point, and C is a regularization parameter;

[0135] For the linearly inseparable data in this design, the kernel function is introduced to map the sample points from the original feature space to a high-dimensional feature space. The kernel function is expressed as:

[0136]

[0137] x i , x jis a sample point in the input space, It is a function that maps sample points to feature space. The key property of the kernel function is that it can directly calculate the inner product of two samples in the feature space without knowing the specific mapping function. What is it.

[0138] In summary, the present invention designs an oil shale pyrolysis control system and method, uses an electronic nose to record the gas composition data generated during the oil shale pyrolysis process, constructs an RNN-SVM model to identify the composition of the gas collected by the electronic nose, learns the complex nonlinear relationship between gas data and temperature, and realizes the control of the oil shale pyrolysis process. Not only is the model simple and low-cost, but it can also achieve higher energy utilization.

[0139] It should be noted that, in this article, the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of more restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0140] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An oil shale pyrolysis control system, characterized in that: The system comprises: a pyrolysis furnace, a controller, an electronic nose, a methane gas storage tank, a gas separator, a gas collecting tank, a scrubber, a gas valve, a methane gas path and a mixed gas path; The pyrolysis furnace includes: a feed inlet, a smoke outlet, a waste outlet, a preheating chamber, a retorting chamber, a combustion chamber, a cooling chamber, a ring duct, an air inlet and a temperature sensor; The shale particles are heated and dried by heat conduction between the cooling chamber and the preheating chamber. The dried shale particles and the shale ash from the combustion chamber are mixed and heat-transferred in the distillation chamber to reach the optimal temperature of 505°C required for shale distillation. The shale semi-coke produced after distillation enters the combustion chamber through an annular pipe and is burned in the combustion chamber with the air entering through the air inlet to make the temperature of the shale ash reach 740°C. The smoke generated by the combustion is discharged out of the furnace from the smoke exhaust port. Part of the shale ash enters the cooling chamber for cooling, and at the same time, part of the heat is transferred to the preheating chamber through heat conduction. The gas collecting tank is connected to the distillation chamber, and the hydrocarbon vapor and other Other gases are discharged into a gas collecting tank, and a part of them are introduced into an electronic nose. The electronic nose detects gas characteristics through its sensor array, identifies gas components and sends data to a controller. The washing tower is connected to the gas collecting tank and is used to wash away impurities in the remaining gas in the gas collecting tank. The washing tower is connected to a gas separator through a mixed gas path. The gas separator is used to separate the combustible gas methane. The gas separator is connected to the methane gas storage tank. The methane gas storage tank provides auxiliary fuel methane to the combustion chamber through a methane gas path and a gas valve. The controller determines the opening and closing of the gas valve according to the signal fed back by the temperature sensor.

2. The oil shale pyrolysis control system according to claim 1, characterized in that: The pyrolysis furnace adopts a horizontal rotary kiln structure with a slightly inclined furnace body. It contains two concentric cylinders and is divided into four chambers: a preheating chamber, a dry distillation chamber, a combustion chamber, and a cooling chamber. Each chamber is isolated by a sealed channel, which controls the gas flow between the chambers.

3. A method for controlling pyrolysis of oil shale, characterized in that: The control method is implemented based on the oil shale pyrolysis control system according to claim 1, and the control method comprises the following steps: Step 1: placing oil shale particles into a pyrolysis furnace through a feed port; Step 2: heating the pyrolysis furnace according to a preset program to achieve pyrolysis of the oil shale; Step 3: During the pyrolysis process, volatile organic compounds are volatilized from the oil shale, and the volatile organic compounds are collected in a gas collecting tank; Step 4: The gas in the gas collecting tank is introduced into the electronic nose through the gas pipeline. The electronic nose detects the gas characteristics through its sensor array, identifies the gas composition and sends the data to the controller; Step 5: The controller identifies the volatile components in the oil shale by analyzing the output data of the electronic nose, identifies the gas composition ratio by building an RNN model, and determines the pyrolysis maturity of the oil shale; Step 6: Use the SVM model to learn the complex nonlinear relationship between gas data and temperature to control the pyrolysis temperature, thereby accurately controlling the pyrolysis process of oil shale and improving production efficiency.

4. The oil shale pyrolysis control method according to claim 3, characterized in that: In the step 2, the initial temperature of the pyrolysis furnace is set according to the oil shale of different composition and origin; the oil shale and hot rock ash are mixed and heated to 505° C. to produce shale oil, gas and shale semi-coke, the oil and gas are extracted and condensed to generate shale oil and high calorific value dry distillation gas, the shale semi-coke enters the combustion chamber through an annular pipeline, burns in the air to become shale ash, part of the hot shale ash returns to the dry distillation chamber through a circulation system as a solid heat carrier to heat the oil shale, forming a cycle, and the other part of the hot shale ash is discharged from the cooling chamber, and the preheating chamber is heated during the discharge process.

5. The oil shale pyrolysis control method according to claim 3, characterized in that: In step three, the volatile organic compounds generated by pyrolysis of oil shale include carbon monoxide, carbon dioxide, methane, and hydrocarbon gases.

6. The oil shale pyrolysis control method according to claim 3, characterized in that: In step 4, the electronic nose is composed of 16 gas sensors. The sensors are in contact with the gas products to generate voltage signals, which are received and converted into digital signals by the acquisition card, and the data is sent to the controller, including the following specific steps: S1. Collect gas composition data and build a structured dataset; S2. Calibrate the collected data values; S3. Normalize the data, the formula is as follows: Where x' is the standardized value, μ is the mean value of the data, and σ is the variance of the data.

7. The method for controlling pyrolysis of oil shale according to claim 3, characterized in that: The step five is to construct an RNN model to process the gas data collected by the electronic nose, thereby identifying the gas composition ratio and determining the pyrolysis maturity of the oil shale, including the following specific steps: S1. Construct an RNN model, use gas composition data for training, and save the identification parameters of the oil shale pyrolysis stage after training until the results converge; S2. Use the saved RNN model to predict the stage of oil shale pyrolysis and obtain the accurate value of the composition ratio of pyrolysis gas products; S3. The RNN model includes parameters related to the number of hidden layers, the number of hidden neural nodes, the activation function, the optimizer, the learning rate, the random inactivation layer and the number of iterations and the loss function training model; The expression of the RNN model is as follows: (1) Forward propagation: a (t) =b+Wh (t-1) +Ux (t) ; h (t) =tanh(a (t) ); o (t) =c+Vh (t) ; and^ (t) =softmax(or (t) ); a (t) represents the hidden state at the tth time step, x (t) represents the input of the tth time step, y^ (t) represents the output of the t-th time step, b and c represent the bias vector and weight matrix respectively, U, V and W correspond to the connections from input to hidden, hidden to output and hidden to hidden respectively, and tanh is the hidden layer activation function; (2) Loss function: ITS t (the t ,y^ t )=-y t logy^ t 4 E(y,y^)=∑ t ITS t (the t ,y^ t )-∑ t the t logy^ t 4 The loss function of an element: called the cross entropy loss function; (3) Reverse calculation: Parameter optimization method: Gradient descent, same as traditional neural networks; Calculate the derivative of the loss function with respect to the parameters; Each output has an effect on the parameters; The derivatives of the parameters are the outputs The sum of derivatives with respect to the parameters; Solutions to gradient disappearance and gradient explosion: 1) Use Leaky ReLU activation function: f(x) = max(αx, x); α is a constant; 2) Gradient clipping and regularization: This is mainly proposed to address gradient explosion. The idea is to set a gradient clipping threshold, and then when updating the gradient, if the gradient exceeds this threshold, it will be forced to be limited within this range.

8. The oil shale pyrolysis control method according to claim 3, characterized in that: In step six, the steps of controlling the pyrolysis temperature by learning the complex nonlinear relationship between gas data and temperature through the SVM model include: Construct an SVM model and use known gas data and temperature data to train the SVM model. The SVM model learns the complex nonlinear relationship between gas data and temperature. The SVM model is as follows: Given a training set D containing N samples, D = {(x1, y1), (x2, y2), ..., (x N ,y N ),}, where x i represents the feature vector of the i-th sample, y i Represents the category label of the i-th sample; feature vector x i Has n dimensions, namely x i =(x 1i ,x 2i ,…,x ni ); The hyperplane is expressed as follows: w·x+b=0; w represents the normal vector of the hyperplane, which determines the direction of the hyperplane, b represents the intercept of the hyperplane, and w·x represents the inner product of w and x; The goal is to find w and b so that all sample points satisfy: y i (w·x i +b)≥1; At the same time, the boundary should be maximized, so the optimization problem is expressed as: min1 / 2"in" 2 ; In order to solve the outliers, the slack variable (x i ,ε i ), so the optimization problem becomes: min1 / 2"in" 2 +C∑ε i ; The constraints are: y i (w·x i +b)≥1-ε i ,i=1,2,…,N; e i ≥0,i=1,2,…,N; ‖w‖ represents the norm of vector w, y i Represents the category label of the i-th sample point, and C is a regularization parameter; For linearly inseparable data, the kernel function is introduced to map the sample points from the original feature space to a high-dimensional feature space. The kernel function is expressed as: x i , x j is a sample point in the input space, It is a function that maps sample points to feature space.

Images