Sludge pyrolysis prediction method and device, electronic equipment and storage medium
By establishing a sludge pyrolysis prediction model based on BPM artificial neural network and utilizing the momentum training learning rate adaptive algorithm, the problem of predicting activation energy and products in the sludge pyrolysis process was solved, and efficient and accurate prediction of sludge pyrolysis parameters and products was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA UNIV OF MINING & TECH (BEIJING)
- Filing Date
- 2022-12-21
- Publication Date
- 2026-05-12
AI Technical Summary
Existing technologies make it difficult to accurately predict the pyrolysis activation energy and products during sludge pyrolysis, which increases the difficulty of studying the sludge pyrolysis mechanism and affects pyrolysis kinetic parameters, product composition, and yield.
A BPM artificial neural network was used to establish an activation energy prediction model and a pyrolysis product prediction model. The sample set was trained by a momentum training adaptive learning rate algorithm until the set accuracy was reached, so as to achieve efficient and accurate prediction of sludge pyrolysis activation energy and products.
It enables efficient and accurate prediction of the activation energy of sludge pyrolysis, as well as accurate prediction of the yield of pyrolysis three-phase products and gas components, thereby improving the prediction accuracy and efficiency of the sludge pyrolysis process.
Smart Images
Figure CN115879635B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of solid waste treatment, disposal and resource utilization, and in particular to a method and apparatus for predicting sludge pyrolysis, electronic equipment and storage medium. Background Technology
[0002] my country's annual sludge production in recent years is estimated to have reached 35.4 MT / year. As a byproduct of wastewater treatment, sludge contains various macromolecular organic matter, heavy metals, microorganisms, and other substances. If not properly and safely disposed of, it will cause significant environmental harm. Simultaneously, due to its near-zero carbon dioxide emissions, sludge is also defined as a special type of biomass and a fossil fuel substitute highly suitable for pyrolysis. Pyrolysis technology can fully utilize the organic matter in sludge, forming a three-phase product of solid, liquid, and gas, effectively avoiding secondary pollution that may result from other methods.
[0003] However, due to the numerous components of sludge, a series of complex competing and concurrent reactions occur during pyrolysis, including a series of dehydration and pyrolysis reactions of polymers and reactions between the three-phase pyrolysis products, which increases the difficulty of studying the sludge pyrolysis mechanism. Furthermore, the physicochemical properties of the sludge and the operating conditions of pyrolysis both affect the pyrolysis kinetic parameters and the composition and yield of the pyrolysis products.
[0004] However, the method of predicting sludge pyrolysis based solely on experimental data without knowing the mechanism still needs improvement. Summary of the Invention
[0005] The purpose of this application is to solve the aforementioned technical problems and to provide a method and apparatus for predicting sludge pyrolysis, an electronic device, and a storage medium.
[0006] To achieve the above objectives, this application provides a method for predicting sludge pyrolysis, comprising the following steps:
[0007] Cellulose, hemicellulose, lignin, protein, soluble sugar, and fat from sludge were selected as model materials. An activation energy prediction model was established using a BPM artificial neural network. The activation energy prediction model includes six activation energy prediction sub-models. The structure of each activation energy prediction sub-model is as follows: the three nodes of the input layer are pyrolysis temperature, heating rate, and conversion rate, respectively, and the one node of the output layer is the pyrolysis activation energy.
[0008] Through the formula:
[0009]
[0010] x(k+1) = x(k) + Δx(k+1)
[0011] An adaptive learning rate algorithm based on momentum training is implemented, where: k is the number of iterations, η is the momentum constant, x represents the vector of weights and biases, rand() is a random number, and ΔE(k) is the error function at the k-th iteration. The first training sample set is input into the activation energy prediction model for model training until the training accuracy reaches the first set value, and the model training is completed. The first training sample set is the first activation energy sample set obtained by thermogravimetric analysis and activation energy calculation of the model material.
[0012] The first verification sample set is input into the trained activation energy prediction model to obtain the pyrolysis activation energy of each model substance. The first verification sample set is the second activation energy sample set obtained by thermogravimetric analysis and activation energy calculation of the model substance.
[0013] Optionally, the method also includes: detecting the actual content of six organic components in the sludge—cellulose, hemicellulose, lignin, protein, soluble sugar, and fat—and conducting pyrolysis experiments on the sludge. A pyrolysis product prediction model is established using a BPM artificial neural network. This pyrolysis product prediction model includes: a sub-model predicting the yield of the three-phase pyrolysis products and a sub-model predicting the yield of each component of the pyrolysis gas. The structure of the sub-model predicting the yield of the three-phase pyrolysis products is as follows: the nine nodes in the input layer are the final pyrolysis temperature, the heating rate, the sludge moisture content, cellulose... The pyrolysis gas yield prediction sub-model consists of the following components: cellulose content, hemicellulose content, lignin content, protein content, soluble sugar content, and fat content. The output layer has three nodes: pyrolysis char yield, pyrolysis oil yield, and pyrolysis gas yield. The sub-model is structured as follows: the input layer has nine nodes: pyrolysis final temperature, heating rate, sludge moisture content, cellulose content, hemicellulose content, lignin content, protein content, soluble sugar content, and fat content. The output layer has four nodes: the yields of H2, CO, CO2, and CH4 in the pyrolysis gas.
[0014] Through the formula:
[0015]
[0016] x(k+1) = x(k) + Δx(k+1)
[0017] An adaptive learning rate algorithm based on momentum training is implemented, where: k is the number of iterations, η is the momentum constant, x represents the vector of weights and biases, rand() is a random number, and ΔE(k) is the error function at the k-th iteration. The second training sample set is input into the yield prediction sub-model of the three-phase products of pyrolysis and the yield prediction sub-model of each component of pyrolysis gas for model training until the training accuracy reaches the second set value, and the model training is completed. The second training sample set is the first product sample set calculated after the sludge undergoes organic component content detection and pyrolysis experiment.
[0018] The second verification sample set is input into the trained pyrolysis three-phase product yield prediction sub-model and the pyrolysis gas component yield prediction sub-model to obtain the pyrolysis three-phase product yield and the pyrolysis gas component yield. The second verification sample set is the second product sample set calculated after the sludge undergoes organic component content detection and pyrolysis experiment.
[0019] Optionally, the first training sample set and the first validation sample set are normalized before being input into the activation energy prediction model and before being input into the trained activation energy prediction model.
[0020] Optionally, the first training sample set and the second training sample set further include a training set and a test set, wherein the ratio of the number of samples in the training set and the test set to the number of samples in the first verification sample set and the second verification sample set is 70% for the training set, 15% for the test set, and 15% for the verification sample set.
[0021] Optionally, before inputting the second training sample set into the yield prediction sub-model of the three-phase pyrolysis products and the yield prediction sub-model of each component of the pyrolysis gas for model training, and before inputting the second verification sample set into the trained yield prediction sub-model of the three-phase pyrolysis products and the yield prediction sub-model of each component of the pyrolysis gas, the second training sample set and the second verification sample set are normalized.
[0022] Optionally, the activation functions of the input and hidden layers of the activation energy prediction sub-model, the yield prediction sub-model of the three-phase pyrolysis products, and the yield prediction sub-model of each component of the pyrolysis gas are set to nonlinear hyperbolic tangent Sigmoid functions, and the activation function of the output layer is set to a linear function.
[0023] Optionally, the number of hidden layers in the activation energy prediction sub-model, the yield prediction sub-model of the three-phase pyrolysis products, and the yield prediction sub-model of each component of the pyrolysis gas is 1 to 2, and the number of neurons in each hidden layer is 3 to 13.
[0024] Accordingly, this application also provides a sludge pyrolysis prediction device, comprising the following modules:
[0025] The model generation module, specifically the activation energy prediction model module, is configured to select cellulose, hemicellulose, lignin, protein, soluble sugar, and fat from sludge as model materials, and to establish an activation energy prediction model using a BPM artificial neural network. The activation energy prediction model includes six activation energy prediction sub-models; wherein the structure of each activation energy prediction sub-model is as follows: the input layer has three nodes representing pyrolysis temperature, heating rate, and conversion rate, respectively, and the output layer has one node representing pyrolysis activation energy.
[0026] The model training module is configured to use the following formula:
[0027]
[0028] x(k+1) = x(k) + Δx(k+1)
[0029] An adaptive learning rate algorithm based on momentum training is implemented, where: k is the number of iterations, η is the momentum constant, x represents the vector of weights and biases, rand() is a random number, and ΔE(k) is the error function at the k-th iteration. The first training sample set is input into the activation energy prediction model for model training until the training accuracy reaches the first set value, and the model training is completed. The first training sample set is the first activation energy sample set obtained by thermogravimetric analysis and activation energy calculation of the model material.
[0030] The model validation module is configured to input a first validation sample set into the trained activation energy prediction model to obtain the pyrolysis activation energy of each model substance. The first validation sample set is a second activation energy sample set obtained by thermogravimetric analysis and activation energy calculation of the model substance.
[0031] Accordingly, this application also provides an electronic device, the electronic device comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute the above-described sludge pyrolysis data prediction method.
[0032] Accordingly, this application also provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the above-described sludge pyrolysis data prediction method.
[0033] The beneficial effects of this application are:
[0034] The sludge pyrolysis prediction method of this application establishes an activation energy prediction sub-model, and then uses a first training sample set to train the activation energy prediction model with a momentum-based adaptive learning rate algorithm until a first set value is reached to obtain the trained activation energy prediction sub-model. Then, the first verification sample set is input into the trained activation energy prediction sub-model to predict the pyrolysis kinetic parameters of unknown sludge samples with known characteristics, which can achieve efficient and accurate prediction of sludge pyrolysis activation energy.
[0035] On the other hand, the sludge pyrolysis prediction method of this application establishes a yield prediction sub-model of the three-phase pyrolysis products and a yield prediction sub-model of each component of the pyrolysis gas. Then, it uses a second training sample set to train the yield prediction sub-model and the yield prediction sub-model using a momentum-based adaptive learning rate algorithm until a second set value is reached, thus obtaining the trained yield prediction sub-model and the yield prediction sub-model. Finally, it inputs a second verification sample set into the trained yield prediction sub-model and the yield prediction sub-model to achieve accurate prediction of the yield of the three-phase pyrolysis products of sludge and the yield of the main components of the pyrolysis gas. Attached Figure Description
[0036] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly described below.
[0037] Figure 1 This is one of the schematic diagrams illustrating the overall concept of the sludge pyrolysis prediction method provided in an embodiment of this application;
[0038] Figure 2 This is one of the flowcharts of a sludge pyrolysis prediction method provided in an embodiment of this application;
[0039] Figure 3 This is a structural diagram of the activation energy prediction sub-model in the sludge pyrolysis prediction method according to an embodiment of this application;
[0040] Figure 4 This is the second schematic diagram of the overall concept of the sludge pyrolysis prediction method provided in one embodiment of this application;
[0041] Figure 5 The second flowchart is a method for predicting sludge pyrolysis provided in an embodiment of this application.
[0042] Figure 6 This is a structural diagram of the yield prediction sub-model of the three-phase pyrolysis products in the sludge pyrolysis prediction method according to an embodiment of this application;
[0043] Figure 7 This is a structural diagram of the pyrolysis gas yield prediction sub-model in a sludge pyrolysis prediction method according to an embodiment of this application;
[0044] Figure 8 This is a block diagram of a sludge pyrolysis prediction device according to an embodiment of this application;
[0045] Figure 9 A structural block diagram of an electronic device is also provided as an embodiment of this application. Detailed Implementation
[0046] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.
[0047] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of this application, but not all embodiments.
[0048] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0049] First, a brief introduction to the design concept of the embodiments of this application will be given.
[0050] One embodiment of this application first proposes a method for predicting the activation energy of pyrolysis.
[0051] Please refer to Figure 1 This is one of the schematic diagrams illustrating the overall concept of an embodiment of this application.
[0052] First, cellulose, hemicellulose, lignin, protein, soluble sugar, and fat from sludge were selected as model materials. Thermogravimetric analysis was performed on these six model materials, and data on pyrolysis temperature, heating rate, and conversion rate were collected. The activation energy data of the six model materials were calculated using the isotransformation kinetic analysis methods FWO (Flynn-Wall-Ozawa) and KAS (Kissinger-Akahira-Sunose).
[0053] Next, the pyrolysis temperature, heating rate, conversion rate data and corresponding activation energy data obtained from thermogravimetric analysis of the model material are normalized and preprocessed, and the dataset is divided to obtain training set, test set and validation set for subsequent network model training.
[0054] An initial activation energy prediction model is then established by inputting training set data and setting the initial weights, thresholds, hidden layers, and number of neurons. By calculating the output values and errors of each layer, the initial activation energy prediction model that does not meet the minimum error condition is adjusted using a momentum-based adaptive learning rate algorithm, and the number of hidden layers and neurons is modified until the minimum error condition is met, thus obtaining an intermediate activation energy prediction model.
[0055] Finally, the pyrolysis activation energy is predicted using the final activation energy prediction model, resulting in efficient and accurate predictions.
[0056] To clearly demonstrate the implementation process of the sludge pyrolysis prediction method according to an embodiment of this application, please refer to... Figure 2Specifically, it includes steps S10 to S30.
[0057] Step S10: Select cellulose, hemicellulose, lignin, protein, soluble sugar and fat in sludge as model materials, and use BPM artificial neural network (back-propagation artificial neural network with amomentum algorithm) to establish an activation energy prediction model.
[0058] The activation energy prediction model includes six activation energy prediction sub-models.
[0059] Please refer to Figure 3 This is a structural diagram of the activation energy prediction sub-model in a sludge pyrolysis prediction method according to an embodiment of this application. The structure of the activation energy prediction sub-model is as follows: the input layer has three nodes, namely pyrolysis temperature, heating rate, and conversion rate; the output layer has one node, representing the pyrolysis activation energy.
[0060] The number of hidden layers in the activation energy prediction sub-model is 1 to 2, and the number of neurons in each hidden layer is 3 to 13. For example, using cellulose data at 5℃ / min, the optimal structure for training was determined to be two hidden layers, each with 3 neurons.
[0061] In this embodiment, the activation functions of the input and hidden layers of the activation energy prediction sub-model are set to nonlinear hyperbolic tangent sigmoid functions, and the activation function of the output layer is set to a linear function.
[0062] In other embodiments, the activation functions of the input and hidden layers and the activation function of the output layer of the activation energy prediction sub-model can be set to other functions, all of which fall within the protection scope of this application.
[0063] Step S20: Using the formula:
[0064]
[0065] x(k+1) = x(k) + Δx(k+1)
[0066] An adaptive learning rate algorithm based on momentum training is implemented, where: k is the number of iterations, η is the momentum constant, x represents the vector of weights and biases, rand() is a random number, and ΔE(k) is the error function at the k-th iteration. The first training sample set is input into the activation energy prediction model for model training until the training accuracy reaches the first set value, thus completing the model training. The first training sample set is the first activation energy sample set obtained by thermogravimetric analysis and activation energy calculation of the model material.
[0067] Specifically, the six main organic components in sludge—cellulose, hemicellulose, lignin, protein, soluble sugar, and fat—were used as model materials. Thermogravimetric analysis was performed on each of them to collect information on pyrolysis temperature, heating rate, and conversion rate. Then, the activation energies of the six model materials were calculated using the kinetic-free model methods FWO (Flynn-Wall-Ozawa) and KAS (Kissinger-Akahira-Sunose) to construct the first training sample set and the first validation sample set.
[0068] In one optional implementation, the first training sample set further includes a training set and a test set, wherein the ratio of the number of samples in the training set, the test set, and the first validation sample set is 70% for the training set, 15% for the test set, and 15% for the first validation sample set. The training set is used to input into the initial activation energy prediction sub-model for training to obtain an intermediate activation energy prediction sub-model, the test set is used to input into the intermediate model for training to obtain the final model, and the validation set is used to input into the trained final model to obtain the final result.
[0069] In one optional implementation, the training parameters in the activation energy prediction sub-model are: momentum constant 0.9, initial learning rate 0.01. During the training process of the activation energy prediction sub-model, adaptive adjustments are made based on the local error surface. That is, when in a flat region of the error surface, if the modification direction is correct and the error tends to decrease towards the target value, the step size is increased, and the learning rate is multiplied by the increment ratio of 1.05 to increase the learning rate; when in a steep part of the error surface, i.e., the error increases beyond the time limit setting value, it indicates that the correction is excessive, and the step size should be adjusted to decrease it, and the learning rate is multiplied by the reduction ratio of 0.7. By continuously adjusting the weights and thresholds of the network until the minimum error requirement is met, a well-trained activation energy prediction sub-model is obtained.
[0070] In one embodiment, before inputting the first training sample set into the activation energy prediction model and before inputting the first verification sample set into the trained activation energy prediction model, the first training sample set and the first verification sample set need to be normalized to avoid the output data, which accounts for a small proportion, having a relatively large error in the total error.
[0071] Step S30: Input the first verification sample set into the trained activation energy prediction model to obtain the pyrolysis activation energy of each model substance. The first verification sample set is the second activation energy sample set obtained by thermogravimetric analysis and activation energy calculation of the model substance.
[0072] In summary, the sludge pyrolysis prediction method of this application establishes an activation energy prediction sub-model, then uses a first training sample set to train the activation energy prediction model using a momentum-based adaptive learning rate algorithm until the minimum error condition is met, thus obtaining the trained activation energy prediction sub-model. The first verification sample set is then input into the trained activation energy prediction sub-model to predict the pyrolysis kinetic parameters of unknown sludge samples with known characteristics, thereby achieving efficient and accurate prediction of sludge pyrolysis activation energy.
[0073] An embodiment of this application then proposes a method for predicting sludge pyrolysis products and yield.
[0074] Please refer to Figure 4 This is the second schematic diagram of the overall concept of an embodiment of this application.
[0075] First, the content of organic components (i.e. model substances) in the actual sludge was tested, and a pyrolysis experiment was conducted to calculate the yield of the three-phase products of sludge pyrolysis and the output of each component of the pyrolysis gas.
[0076] Next, the data obtained from the pyrolysis experiment of sludge will be preprocessed to obtain training set, test set and validation set for subsequent network model training.
[0077] An initial pyrolysis product prediction model is then established by inputting training set data and setting the initial weights, thresholds, hidden layers, and number of neurons. By calculating the output values and errors of each layer, the initial pyrolysis product prediction model that does not meet the minimum error condition is adjusted using a momentum-based adaptive learning rate algorithm, and the number of hidden layers and neurons is modified until the minimum error condition is met, thus obtaining an intermediate pyrolysis product prediction model.
[0078] Finally, the final pyrolysis product prediction model was used to predict the products and yield of sludge pyrolysis, resulting in efficient and accurate predictions.
[0079] To clearly demonstrate the implementation process of the sludge pyrolysis prediction method according to an embodiment of this application, please refer to... Figure 5 Specifically, it includes steps S10 to S60, wherein steps S10 to S30 are the same as those provided in the previous embodiment, and will not be repeated here.
[0080] Step S40: Detect the actual content of six organic components in the sludge: cellulose, hemicellulose, lignin, protein, soluble sugar, and fat. Conduct pyrolysis experiments on the sludge and establish a pyrolysis product prediction model using a BPM artificial neural network.
[0081] The pyrolysis product prediction model includes: a yield prediction sub-model for the three-phase pyrolysis products and a yield prediction sub-model for each component of the pyrolysis gas.
[0082] Please refer to Figure 6This is a model structure diagram of the pyrolysis three-phase product yield prediction sub-model in the sludge pyrolysis prediction method according to an embodiment of this application. The structure of the pyrolysis three-phase product yield prediction sub-model is as follows: the nine nodes in the input layer are pyrolysis final temperature, heating rate, sludge moisture content, cellulose percentage, hemicellulose percentage, lignin percentage, protein percentage, soluble sugar percentage, and fat percentage; the three nodes in the output layer are pyrolysis char yield, pyrolysis oil yield, and pyrolysis gas yield.
[0083] The number of hidden layers in the yield prediction sub-model of the pyrolysis three-phase products is 1 to 2, and the number of neurons in each hidden layer is 3 to 13.
[0084] Please refer to Figure 7 This is a model structure diagram of the pyrolysis gas component yield prediction sub-model in the sludge pyrolysis prediction method according to an embodiment of this application. The structure of the pyrolysis gas component yield prediction sub-model is as follows: the nine nodes of the input layer are the final pyrolysis temperature, heating rate, sludge moisture content, cellulose ratio, hemicellulose ratio, lignin ratio, protein ratio, soluble sugar ratio, and fat ratio, respectively; the four nodes of the output layer are the yields of H2, CO, CO2, and CH4 in the pyrolysis gas, respectively. The number of hidden layers in the pyrolysis gas component yield prediction sub-model is 1 to 2, and the number of neurons in each hidden layer is 3 to 13.
[0085] In this embodiment, the activation functions of the input and hidden layers of the pyrolysis three-phase product yield prediction sub-model and the pyrolysis gas component yield prediction sub-model are set to nonlinear hyperbolic tangent sigmoid functions, and the activation function of the output layer is set to a linear function. In other embodiments, the activation functions of the input and hidden layers and the output layer of the yield prediction sub-model and the product prediction sub-model can be set to other functions, all of which fall within the protection scope of this application.
[0086] Step S50: Both the yield prediction sub-model and the production prediction sub-model are implemented using the following formula:
[0087]
[0088] x(k+1) = x(k) + Δx(k+1)
[0089] Implement a learning rate adaptive algorithm based on momentum training, where: k is the number of iterations, η is the momentum constant, x represents the vector of weights and biases, rand() is a random number, and ΔE(k) is the error function at the k-th iteration. Input the second training sample set into the yield prediction sub-model of the three-phase products of pyrolysis and the yield prediction sub-model of each component of pyrolysis gas for training until the training accuracy reaches the second set value, and complete the model training.
[0090] By detecting the actual content of the six main organic components in sludge—cellulose, hemicellulose, lignin, protein, soluble sugar, and fat—and conducting pyrolysis experiments on the sludge, information on sludge moisture content, the proportion of model material content, final pyrolysis temperature, heating rate, and the yield of three-phase pyrolysis products and the yield of the main components of pyrolysis gas was collected to construct a second training sample set and a second validation sample set.
[0091] In one optional implementation, the second training sample set further includes a training set and a test set, wherein the ratio of the number of samples in the training set, the test set and the second verification sample set is 70% for the training set, 15% for the test set and 15% for the second verification sample set. The training set is used to input into the initial yield prediction sub-model or the production yield prediction sub-model for training to obtain the intermediate yield prediction sub-model and the intermediate production yield prediction sub-model. The test set is used to input into the above intermediate models for training to obtain the final model. The verification set is used to input into the trained final model to obtain the final result.
[0092] In one optional implementation, the momentum constant in the yield prediction sub-model and the output prediction sub-model is 0.9, and the initial learning rate is set to 0.01. During training, adaptive adjustments are made based on the local error surface. That is, when the error surface is flat, if the modification direction is correct and the error tends to decrease towards the target value, the step size is increased, and the learning rate is multiplied by the increment ratio of 1.05 to increase the learning rate; when the error surface is steep, i.e., the error increases beyond the time limit setting, it indicates that the correction is excessive, and the step size should be adjusted to decrease it, and the learning rate is multiplied by the reduction ratio of 0.7. By continuously adjusting the weights and thresholds of the network until the minimum error requirement is met, the trained yield prediction sub-model and output prediction sub-model are obtained.
[0093] In one embodiment, before inputting the second training sample set into the pyrolysis product prediction model and before inputting the second verification sample set into the trained pyrolysis product prediction model, the second training sample set and the second verification sample set are normalized to avoid the problem that the output data, which accounts for a small proportion, has a relatively large error in the total error.
[0094] Step S60: Input the second verification sample set into the trained yield prediction sub-model and production prediction sub-model to obtain the yield of the three-phase pyrolysis products and the production of each component of the pyrolysis gas.
[0095] In summary, the sludge pyrolysis prediction method of one embodiment of this application establishes a yield prediction sub-model for the three-phase pyrolysis products and a yield prediction sub-model for each component of the pyrolysis gas. Then, it uses a second training sample set to train the yield prediction sub-model and the yield prediction sub-model using a momentum-based adaptive learning rate algorithm until the minimum error condition is met, thus obtaining the trained yield prediction sub-model and the yield prediction sub-model. Finally, it inputs a second verification sample set into the trained yield prediction sub-model and the yield prediction sub-model, thereby achieving accurate prediction of the yield of the three-phase pyrolysis products of sludge and the yield of the main components of the pyrolysis gas.
[0096] Based on the above embodiments, one embodiment of this application also provides a sludge pyrolysis prediction device.
[0097] Figure 8 This is a block diagram of a sludge pyrolysis prediction device according to an embodiment of this application, as shown below. Figure 8 As shown, the device may include the following modules:
[0098] The model generation module 100, wherein the activation energy prediction model module is configured to select cellulose, hemicellulose, lignin, protein, soluble sugar, and fat in sludge as model materials, and establish an activation energy prediction model using a BPM artificial neural network. The activation energy prediction model includes six activation energy prediction sub-models; wherein the structure of each activation energy prediction sub-model is as follows:
[0099] The input layer has three nodes: pyrolysis temperature, heating rate, and conversion rate, while the output layer has one node: pyrolysis activation energy.
[0100] Model training module 200, the model training module being configured to use the formula:
[0101]
[0102] x(k+1) = x(k) + Δx(k+1)
[0103] An adaptive learning rate algorithm based on momentum training is implemented, where k is the iteration number, η is the momentum constant, x represents the vector of weights and biases, rand() is a random number, and ΔE(k) is the error function at the k-th iteration. The first training sample set is input into the activation energy prediction model for model training until the training accuracy reaches a first set value, thus completing the model training. The first training sample set is the first activation energy sample set obtained by thermogravimetric analysis and activation energy calculation of the model material.
[0104] The model verification module 300 is configured to input a first verification sample set into a trained activation energy prediction model to obtain the pyrolysis activation energy of each model substance. The first verification sample set is a second activation energy sample set obtained by thermogravimetric analysis and activation energy calculation of the model substance.
[0105] The specific implementation process of the functions and roles of each module in the above-mentioned device can be found in the implementation process of the corresponding steps in the above-mentioned sludge pyrolysis prediction method, and will not be repeated here.
[0106] like Figure 9 As shown, this application embodiment also provides an electronic device 1, including: at least one processor 11 and a memory 12. Figure 9 Taking a processor 11 as an example. The processor 11 and the memory 12 are connected via a bus 10. The memory 12 stores instructions that can be executed by the processor 11. The instructions are executed by the processor 11 so that the electronic device 1 can perform all or part of the sludge pyrolysis prediction method in the above embodiments.
[0107] The memory 12 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable red-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.
[0108] It should be noted that, Figure 9 The electronic device 1 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.
[0109] This application also provides a computer-readable storage medium storing a computer program that can be executed by a processor 11 to perform the sludge pyrolysis prediction method provided in this application.
[0110] The apparatuses and methods disclosed in the several embodiments provided in this application can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatuses, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0111] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.
Claims
1. A method for predicting sludge pyrolysis, characterized in that, Includes the following steps: Cellulose, hemicellulose, lignin, protein, soluble sugar, and fat from sludge were selected as model substances. An activation energy prediction model was established using a BPM artificial neural network. The activation energy prediction model includes six activation energy prediction sub-models, with one activation energy prediction sub-model corresponding to each of the model substances. The structure of each activation energy prediction sub-model is as follows: the three nodes of the input layer are the pyrolysis temperature, the heating rate, and the conversion rate of the model substance, respectively; and the one node of the output layer is the pyrolysis activation energy. Through the formula: Implement a learning rate adaptive algorithm based on momentum training, where k is the number of iterations. Let x be the momentum constant and x be the vector representing the weights and biases. It is a random number. E(k) is the error function at the kth iteration. The first training sample set is input into the corresponding activation energy prediction sub-model for model training until the training accuracy reaches the first set value and the model training is completed. The first training sample set is the first activation energy sample set obtained by thermogravimetric analysis and activation energy calculation of the model material. The first verification sample set is input into the corresponding trained activation energy prediction sub-model to obtain the pyrolysis activation energy of each model substance. Each first verification sample set is a second activation energy sample set obtained by thermogravimetric analysis and activation energy calculation of the model substance.
2. The sludge pyrolysis prediction method according to claim 1, characterized in that, This also includes: detecting the actual content of six organic components in the sludge—cellulose, hemicellulose, lignin, protein, soluble sugar, and fat—and conducting pyrolysis experiments on the sludge. A pyrolysis product prediction model is established using a BPM artificial neural network, which includes a sub-model predicting the yield of the three-phase pyrolysis products and a sub-model predicting the yield of each component of the pyrolysis gas. The structure of the yield prediction sub-model for the three-phase pyrolysis products is as follows: the nine nodes in the input layer are the final pyrolysis temperature, heating rate, sludge moisture content, cellulose content, hemicellulose content, lignin content, protein content, soluble sugar content, and fat content, respectively; the three nodes in the output layer are the pyrolysis char yield, pyrolysis oil yield, and pyrolysis gas yield, respectively. The structure of the pyrolysis gas component yield prediction sub-model is as follows: the nine nodes of the input layer are the final pyrolysis temperature, heating rate, sludge moisture content, cellulose ratio, hemicellulose ratio, lignin ratio, protein ratio, soluble sugar ratio, and fat ratio, respectively; the four nodes of the output layer are the yields of H2, CO, CO2, and CH4 in the pyrolysis gas, respectively. Through the formula: Implement a learning rate adaptive algorithm based on momentum training, where k is the number of iterations. Let x be the momentum constant and x be the vector representing the weights and biases. It is a random number. E(k) is the error function at the kth iteration. The second training sample set is input into the yield prediction sub-model of the three-phase pyrolysis products and the yield prediction sub-model of each component of the pyrolysis gas for model training until the training accuracy reaches the second set value and the model training is completed. The second training sample set is the first product sample set calculated after the sludge undergoes organic component content detection and pyrolysis experiment. The second verification sample set is input into the trained pyrolysis three-phase product yield prediction sub-model and the pyrolysis gas component yield prediction sub-model to obtain the pyrolysis three-phase product yield and the pyrolysis gas component yield. The second verification sample set is the second product sample set calculated after the sludge undergoes organic component content detection and pyrolysis experiment.
3. The sludge pyrolysis prediction method according to claim 1, characterized in that, Also includes: Before inputting the first training sample set into the activation energy prediction model and before inputting the first validation sample set into the trained activation energy prediction model, the first training sample set and the first validation sample set are normalized.
4. The sludge pyrolysis prediction method according to claim 2, characterized in that, The first training sample set and the second training sample set further include a training set and a test set, and the ratio of the number of samples in the training set and the test set to the number of samples in the first verification sample set and the second verification sample set is 70% for the training set, 15% for the test set, and 15% for the verification sample set.
5. The sludge pyrolysis prediction method according to claim 2, characterized in that, Also includes: Before inputting the second training sample set into the yield prediction sub-model of the three-phase pyrolysis products and the yield prediction sub-model of each component of the pyrolysis gas for model training, and before inputting the second verification sample set into the trained yield prediction sub-model of the three-phase pyrolysis products and the yield prediction sub-model of each component of the pyrolysis gas, the second training sample set and the second verification sample set are normalized.
6. The sludge pyrolysis prediction method according to claim 2, characterized in that, The activation functions of the input and hidden layers of the activation energy prediction sub-model, the yield prediction sub-model of the three-phase pyrolysis products, and the yield prediction sub-model of each component of the pyrolysis gas are set to nonlinear hyperbolic tangent Sigmoid functions, and the activation function of the output layer is set to a linear function.
7. The sludge pyrolysis prediction method according to claim 2, characterized in that, The number of hidden layers in the activation energy prediction sub-model, the yield prediction sub-model of the three-phase pyrolysis products, and the yield prediction sub-model of each component of the pyrolysis gas is 1 to 2, and the number of neurons in each hidden layer is 3 to 13.
8. A sludge pyrolysis prediction device, characterized in that, Includes the following modules: The model generation module is configured to select cellulose, hemicellulose, lignin, protein, soluble sugar, and fat from sludge as model materials, and establish an activation energy prediction model using a BPM artificial neural network structure. The activation energy prediction model includes six activation energy prediction sub-models; each model material corresponds to one activation energy prediction sub-model. The structure of each activation energy prediction sub-model is as follows: the input layer has three nodes representing the pyrolysis temperature, heating rate, and conversion rate of the model material, respectively; the output layer has one node representing the pyrolysis activation energy. The model training module is configured to use the following formula: Implement a learning rate adaptive algorithm based on momentum training, where k is the number of iterations. Let x be the momentum constant and x be the vector representing the weights and biases. It is a random number. E(k) is the error function at the kth iteration. The first training sample set is input into the corresponding activation energy prediction sub-model for model training until the training accuracy reaches the first set value and the model training is completed. The first training sample set is the first activation energy sample set obtained by thermogravimetric analysis and activation energy calculation of the model material. The model validation module is configured to input the first validation sample set into the corresponding trained activation energy prediction sub-model to obtain the pyrolysis activation energy of each model substance. Each first validation sample set is a second activation energy sample set obtained by thermogravimetric analysis and activation energy calculation of the model substance.
9. An electronic device, characterized in that, The electronic device includes: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute the sludge pyrolysis prediction method according to any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the sludge pyrolysis prediction method as described in any one of claims 1 to 7.