CN-115879635-B - Sludge pyrolysis prediction method and device, electronic equipment and storage medium

CN115879635BCN 115879635 BCN115879635 BCN 115879635BCN-115879635-B

Abstract

The application provides a sludge pyrolysis prediction method and device, electronic equipment and a storage medium, and relates to the technical field of sludge treatment. The method comprises the steps of establishing an activation energy predictor model, a yield predictor model of pyrolysis three-phase products and a yield predictor model of each component of pyrolysis gas, respectively training the activation energy predictor model, the yield predictor model and the yield predictor model by utilizing corresponding training sample sets through a learning rate self-adaptive algorithm based on momentum training until minimum error conditions are met, obtaining a trained activation energy predictor model, a trained yield predictor model and a trained yield predictor model, and finally inputting the corresponding verification sample sets into the trained activation energy predictor model, the trained yield predictor model and the trained yield predictor model to accurately predict the sludge pyrolysis activation energy, the yield of pyrolysis three-phase products and the yield of main components of pyrolysis gas.

Inventors

LIU YANJUN
WANG SHENG
LIU YANTING
CHEN JINGYAO
Tang Qiange

Assignees

中国矿业大学(北京)

Dates

Publication Date: 20260512
Application Date: 20221221

Claims (10)

1. The sludge pyrolysis prediction method is characterized by comprising the following steps of: The method comprises the steps of selecting cellulose, hemicellulose, lignin, protein, soluble sugar and fat in sludge as model substances, and establishing an activation energy prediction model by adopting a BPM artificial neural network, wherein the activation energy prediction model comprises six activation energy prediction sub-models, one model substance corresponds to one activation energy prediction sub-model, and the structure of the activation energy prediction sub-model is that three nodes of an input layer are respectively pyrolysis temperature of the model substance, heating rate of the model substance and conversion rate of the model substance, and one node of an output layer is pyrolysis activation energy; By the formula: Realizing a learning rate self-adaptive algorithm based on momentum training, wherein k is the iteration number, The vector of weights and biases is denoted as momentum constant x, In the form of a random number, E (k) is an error function in the kth iteration, a first training sample set is input into a corresponding activation energy predictor model to perform model training until training accuracy reaches a first set value, model training is completed, and one first training sample set is a first activation energy sample set obtained by performing thermogravimetric analysis and activation energy calculation on one model substance; inputting the first verification sample set into a corresponding trained activation energy predictor model to obtain pyrolysis activation energy of each model substance, wherein one first verification sample set is a second activation energy sample set obtained by performing thermogravimetric analysis and activation energy calculation on one model substance.
2. The method for predicting sludge pyrolysis according to claim 1, further comprising detecting the actual contents of cellulose, hemicellulose, lignin, protein, soluble sugar and fat in the sludge, performing pyrolysis experiments on the sludge, and establishing a pyrolysis product prediction model by using a BPM artificial neural network, wherein the pyrolysis product prediction model comprises a yield predictor model of pyrolysis three-phase products and a yield predictor model of pyrolysis gas components, Wherein the yield predictor model of the pyrolysis three-phase product has the structure that nine nodes of an input layer are respectively pyrolysis final temperature, heating rate, sludge water content, cellulose ratio, hemicellulose ratio, lignin ratio, protein ratio, soluble sugar ratio and fat ratio, three nodes of an output layer are respectively pyrolysis carbon yield, pyrolysis oil yield and pyrolysis gas yield, The pyrolysis gas component yield predictor model has the structure that nine nodes of an input layer are pyrolysis final temperature, heating rate, sludge water content, cellulose duty ratio, hemicellulose duty ratio, lignin duty ratio, protein duty ratio, soluble sugar duty ratio and fat duty ratio respectively, and four nodes of an output layer are the yield of H 2 、CO、CO 2 、CH 4 in pyrolysis gas respectively; By the formula: Realizing a learning rate self-adaptive algorithm based on momentum training, wherein k is the iteration number, The vector of weights and biases is denoted as momentum constant x, In the form of a random number, E (k) is an error function in the kth iteration, a second training sample set is input into a yield predictor model of pyrolysis three-phase products and a yield predictor model of each component of pyrolysis gas for model training until the training precision reaches a second set value, and model training is completed, wherein the second training sample set is a first product sample set obtained by calculating sludge after organic component content detection and pyrolysis experiments; and inputting a second verification sample set into the trained yield predictor model of the pyrolysis three-phase product and the yield predictor model of each component of pyrolysis gas to obtain the yield of the pyrolysis three-phase product and the yield of each component of pyrolysis gas, wherein the second verification sample set is a second product sample set obtained by calculating sludge after organic component content detection and pyrolysis experiments.
3. The method of claim 1, further comprising normalizing the first training sample set and the first validation sample prior to 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.
4. The sludge pyrolysis prediction method according to claim 2, wherein the first training sample set and the second training sample set further comprise a training set and a test set, and the ratio of the training set, the test set, and the first verification sample set, and the second verification sample set is 70% of the training set, 15% of the test set, and 15% of the verification sample set.
5. The method of claim 2, further comprising normalizing the second training sample set and the second validation sample prior to inputting the second training sample set into the yield predictor model of pyrolysis three-phase products and the yield predictor model of pyrolysis gas components and prior to inputting the second validation sample set into the yield predictor model of pyrolysis three-phase products and the yield predictor model of pyrolysis gas components.
6. The sludge pyrolysis prediction method according to claim 2, wherein the activation energy predictor model, the yield predictor model of pyrolysis three-phase products, and the input layer and hidden layer activation functions of the pyrolysis gas component yield predictor model are set as nonlinear hyperbolic tangent Sigmoid functions, and the output layer activation functions are set as linear functions.
7. The sludge pyrolysis prediction method according to claim 2, wherein the number of hidden layers of the activation energy predictor model, the yield predictor model of the pyrolysis three-phase product and the pyrolysis gas component yield predictor model is 1-2, and the number of neurons of each hidden layer is 3-13.
8. The sludge pyrolysis prediction device is characterized by comprising the following modules: The model generation module is configured to select cellulose, hemicellulose, lignin, protein, soluble sugar and fat in sludge as model substances, and adopts a BPM artificial neural network structure to establish an activation energy prediction model, wherein the activation energy prediction model comprises six activation energy prediction sub-models, one model substance corresponds to one activation energy prediction sub-model, the structure of the activation energy prediction sub-model is that three nodes of an input layer are respectively pyrolysis temperature of the model substance, heating rate of the model substance and conversion rate of the model substance, one node of an output layer is pyrolysis activation energy, A model training module configured to pass through the formula: Realizing a learning rate self-adaptive algorithm based on momentum training, wherein k is the iteration number, The vector of weights and biases is denoted as momentum constant x, In the form of a random number, E (k) is an error function in the kth iteration, a first training sample set is input into a corresponding activation energy predictor model to perform model training until training accuracy reaches a first set value, model training is completed, and one first training sample set is a first activation energy sample set obtained by performing thermogravimetric analysis and activation energy calculation on one model substance; The model verification module is configured to input a first verification sample set into a corresponding trained activation energy predictor model to obtain pyrolysis activation energy of each model substance, wherein one first verification sample set is a second activation energy sample set obtained by performing thermogravimetric analysis and activation energy calculation on one model substance.
9. An electronic device comprising a processor, a memory for storing processor-executable instructions, wherein the processor is configured to perform the sludge pyrolysis prediction method of 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, the one or more programs are executable by one or more processors to implement the sludge pyrolysis prediction method of any one of claims 1 to 7.

Description

Sludge pyrolysis prediction method and device, electronic equipment and storage medium Technical Field The application relates to the field of solid waste treatment and recycling, in particular to a sludge pyrolysis prediction method and device, electronic equipment and a storage medium. Background In recent years, the annual sludge yield in China is estimated to be 35.4 MT/year, and the sludge is taken as a byproduct of sewage treatment, and contains various macromolecular organic matters, heavy metals, microorganisms and other substances, so that the environment is greatly harmed if proper safety disposal is not carried out. Meanwhile, due to the characteristic of near zero emission of carbon dioxide, sludge is also defined as a special type of biomass, and is a fossil fuel substitute which is very suitable for pyrolysis. The pyrolysis technology can fully utilize organic matters in the sludge to form solid, liquid and gas three-phase products, and effectively avoid secondary pollution possibly caused by other modes. However, due to the numerous sludge components, a series of complex competition and concurrent reactions occur in the pyrolysis process, including a series of dehydration and pyrolysis reactions of high polymer, reactions between pyrolysis three-phase products and the like, which increases the difficulty of researching the pyrolysis mechanism of the sludge. And both the physicochemical properties of the sludge and the operating conditions of pyrolysis affect the pyrolysis kinetic parameters as well as the composition and yield of pyrolysis products. However, without knowledge of the mechanism, the method of predicting sludge pyrolysis based on experimental data alone remains to be perfected. Disclosure of Invention The application aims to solve the technical problems and provides a sludge pyrolysis prediction method and device, electronic equipment and a storage medium. In order to achieve the above object, the present application provides a sludge pyrolysis prediction method, comprising the steps of: selecting cellulose, hemicellulose, lignin, protein, soluble sugar and fat in sludge as model substances, and establishing an activation energy prediction model by adopting a BPM artificial neural network, wherein the activation energy prediction model comprises six activation energy prediction sub-models, and the structure of the activation energy prediction sub-models is that three nodes of an input layer are pyrolysis temperature, heating rate and conversion rate respectively, and one node of an output layer is pyrolysis activation energy; By the formula: x(k+1)=x(k)+Δx(k+1) The learning rate self-adaptive algorithm based on momentum training is realized, wherein k is iteration times, eta is momentum constant, x represents a vector of weight and deviation, rand () is a random number, delta E (k) is an error function in the kth iteration, a first training sample set is input into an activation energy prediction model for model training until training precision reaches a first set value, and model training is completed, wherein the first training sample set is a first activation energy sample set obtained by performing thermogravimetric analysis and activation energy calculation on model materials; And inputting a first verification sample set into a trained activation energy prediction model to obtain pyrolysis activation energy of each model substance, wherein the first verification sample set is a second activation energy sample set obtained by performing thermogravimetric analysis and activation energy calculation on the model substances. Detecting the actual contents of six organic components, namely cellulose, hemicellulose, lignin, protein, soluble sugar and fat, in sludge, carrying out pyrolysis experiments on the sludge, and establishing a pyrolysis product prediction model by adopting a BPM artificial neural network, wherein the pyrolysis product prediction model comprises a pyrolysis three-phase product yield prediction sub-model and a pyrolysis gas component yield prediction sub-model, wherein the pyrolysis three-phase product yield prediction sub-model has the structure that nine nodes of an input layer are pyrolysis final temperature, heating rate, sludge moisture content, cellulose duty ratio, hemicellulose duty ratio, lignin duty ratio, protein duty ratio, soluble sugar duty ratio and fat duty ratio respectively, three nodes of an output layer are pyrolysis carbon yield, pyrolysis oil yield and pyrolysis gas yield respectively, and the nine nodes of the input layer are pyrolysis final temperature, heating rate, sludge moisture content, cellulose duty ratio, hemicellulose duty ratio, lignin duty ratio, protein duty ratio, soluble sugar duty ratio and fat duty ratio respectively, and the output three nodes of the pyrolysis gas component yield prediction sub-model are pyrolysis gas yield 2、CO、CO2、CH4 respectively; By the formula: x(k+1)=x(k)+Δx(k+1) the learning rate self-adaptive