Portable coal and waste plastic synergistic pyrolysis tar yield prediction method and device based on knowledge graph-graph neural network
Through the knowledge graph-graph neural network method and R-GCN algorithm, a tar yield prediction model for synergistic pyrolysis of coal and waste plastics was established, which solved the problems of high cost and long time in the existing technology, and achieved efficient and accurate tar Yield prediction and experimental simulation.
Patent Information
- Application Number
- CN202411325401.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-23
- Publication Date
- 2025-01-07
AI Technical Summary
In the prior art, the prediction of synergistic pyrolysis tar yields of coal and waste plastics has high cost, long time consumption and fuzzy relationship between reactants and products, making it difficult to obtain time-sensitive data.
Using the method based on knowledge graph-graph neural network, we can establish a multi-layer network model by dividing entity relationships, building graph neural networks, applying graph convolutional neural networks and R-GCN algorithms, predicting the impact relationship of tar yield, and developing a portable type The device performs experimental simulation and prediction.
Complete results prediction of incomplete experiments are achieved, and the modular design can deal with a variety of emergencies, improving the efficiency and accuracy of tar yield prediction, and reducing costs and time expenditures.
Smart Images
Figure CN119272615A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of energy technology, and in particular to a portable method and device for predicting tar yield of synergistic pyrolysis of coal and waste plastics based on a knowledge graph-graph neural network. Background Art
[0002] Coal is the main energy source in my country, and its production and consumption account for the majority of primary energy. The reserves and production of low-rank coal are both over 50%, and its direct use will lead to serious waste of resources and environmental pollution. The graded and quality-based utilization of low-rank coal with pyrolysis technology as the core is one of the effective technical routes to achieve clean and efficient transformation of coal resources. Tar is an important high-value-added product of this technology, but due to the low H / C atomic ratio of low-rank coal, the tar yield of its pyrolysis alone is low. Existing technologies have shown that waste plastics can be used as excellent hydrogen donors for coal pyrolysis. Adding a certain amount of waste plastics during the pyrolysis of low-rank coal can effectively increase the tar yield. There is an obvious positive synergistic effect in the co-pyrolysis of the two. Using tar yield as a quantitative indicator of the synergistic effect can guide the application of this technology.
[0003] However, in the actual experimental process, different combinations of coal and waste plastics will have a significant impact on the tar yield. At the same time, the process from obtaining different waste plastics and coal raw materials to completing the experiment is time-consuming and laborious, and it is impossible to obtain data with timeliness advantages. Summary of the invention
[0004] Purpose of the invention: The purpose of the present invention is to provide a portable coal and waste plastic synergistic pyrolysis tar yield prediction method and device based on knowledge graph-graph neural network, which can solve the problems existing in the prior art such as high cost, long time consumption, and unclear relationship between reactants and products.
[0005] Technical solution: The portable coal and waste plastic synergistic pyrolysis tar yield prediction method based on knowledge graph-graph neural network described in the present invention comprises the following steps:
[0006] (1) Based on the synergistic pyrolysis experiment of coal and waste plastics, the entity relationships of the knowledge graph KG are divided;
[0007] (2) Based on the existing experimental data of the synergistic pyrolysis of coal and waste plastics, the basic nodes of the graph neural network are constructed, and the experimental data are processed using the graph convolutional neural network;
[0008] (3) The graph neural network is applied to the knowledge graph, and the information of the central node is updated by the edge nodes to aggregate the multi-layer network layer by layer; the edge relationship is modeled, and the R-GCN algorithm is used to obtain modeling information with direction and node connection; the Logic Rules method is used to train the overall structure and predict the influence of each node on the tar yield.
[0009] Furthermore, step (1) is specifically as follows: there are a total of 5 nodes as structural units of the main knowledge graph, including: the waste plastic content, i.e. the percentage of the total reactants, the coal content, i.e. the percentage of the total reactants, the reaction time from the start of the reaction to the end of the reaction, the stable temperature after electric heating as the reaction temperature and the real-time cumulative tar yield.
[0010] Furthermore, step (2) is specifically as follows: compiling the associated information in the knowledge graph, and using a graph convolutional neural network to import and evolve the functional relationship; wherein the experimental data includes: mixing waste plastics in a mixing ratio of 80% waste plastics and 20% coal, an experimental reaction with complete time data, a stable temperature in the overall experiment as a reference experimental temperature of the device, and the tar yield is the total tar yield.
[0011] Furthermore, in step (3), the R-GCN algorithm assigns different weight ratios to edge relationships to distinguish different relationships; by adding potential reverse edge relationships, modeling information with direction and node connections is obtained.
[0012] The portable coal and waste plastic synergistic pyrolysis tar yield prediction device based on knowledge graph-graph neural network of the present invention comprises: an analysis and prediction module, a main experimental module, and a raw material processing module; wherein,
[0013] The analysis and prediction module is used in the knowledge graph-graph neural network algorithm to predict the tar yield of coal and waste plastic synergistic pyrolysis. The model analyzes the tar yield trend under different experimental conditions and realizes the final prediction of tar yield;
[0014] The main experimental module is used to obtain the content of each component;
[0015] The raw material processing module is used to moderately grind coal into coal particles suitable for thermal reactions in small devices and to break up waste plastics.
[0016] Furthermore, the main experimental module includes a feed port, a synergistic pyrolysis chamber, a main insulation layer and a component analyzer; the specific process is as follows: the synergistic pyrolysis chamber is electrically heated by the on-board battery, and the post-processing raw materials of coal and waste plastics are input into the feed port to obtain the content of each component in time units in the component analyzer; the main components include: H2, CH4, CO and CO2.
[0017] Furthermore, the raw material processing module includes: a grinder and a crusher.
[0018] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: the coal and waste plastic synergistic pyrolysis tar yield prediction device developed by the knowledge graph-graph neural network algorithm can realize the complete result prediction of an incomplete portable pyrolysis experiment, and the modular free combination can cope with more emergencies. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 is a flow chart of the present invention;
[0020] Figure 2 It is a schematic diagram of the device of the present invention;
[0021] Figure 3 Schematic diagram of the prediction model of the present invention. DETAILED DESCRIPTION
[0022] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.
[0023] like Figure 1 As shown, the embodiment of the present invention provides a portable coal and waste plastic synergistic pyrolysis tar yield prediction method based on knowledge graph-graph neural network, such as Figure 2 As shown, the following steps are included:
[0024] (1) Based on the synergistic pyrolysis experiment of coal and waste plastics, the entity relationships of the knowledge graph KG are divided as follows: There are 5 nodes as the structural units of the main knowledge graph, including: the waste plastic content, i.e. the percentage of the total reactants, the coal content, i.e. the percentage of the total reactants, the reaction time from the start of the reaction to the end of the reaction, the stable temperature after electric heating as the reaction temperature, and the real-time cumulative tar yield. The knowledge graph formula is as follows:
[0025]
[0026] in, Represents the set of adjacent nodes of node i; Indicates lower-level relationship; c i,r represents the regularization constant; and Indicates the node weight of the current layer; Indicates the current layer relationship.
[0027] (2) Based on the existing experimental data on the synergistic pyrolysis of coal and waste plastics, the basic nodes of the graph neural network are constructed, and the graph convolutional neural network is used to process the experimental data; specifically, the associated information in the knowledge graph is compiled, and the functional relationship is imported and evolved using the graph convolutional neural network; wherein the experimental data include: waste plastics mixed in a ratio of 80% waste plastics and 20% coal, experimental reactions with complete time data, the stable temperature in the overall experiment as the reference experimental temperature of the device, and the tar yield is the total tar yield.
[0028] The topological structure of the graph is described by the topological graph G. Each node represents a vertex. The set of each vertex and the set of each edge are uniformly expressed in the graph by the matrix X. The temperature, raw material ratio and time are set as the features of the vertex. Therefore, the prediction problem of tar yield can be regarded as the mapping function between the learning topological graph G and the feature matrix X. The formula is as follows:
[0029] [X t+1 ,…,X t+T ]=f(G,(X t-n+1 ,…,X t ))
[0030] Where n represents the length of the input time series and T represents the length of the output prediction sequence.
[0031] (3) Apply graph neural network to knowledge graph, use edge nodes to update the information of central nodes, and aggregate multi-layer networks layer by layer; model edge relationships, use R-GCN algorithm to obtain modeling information with direction and node connection; use Logic Rules method to train the overall structure and predict the influence of each node on tar yield. Among them, R-GCN algorithm assigns different weight ratios to edge relationships to distinguish different relationships; and obtains modeling information with direction and node connection by adding potential reverse edge relationships.
[0032] Since graphs have two different modes, unchanged and variable, the synergistic pyrolysis of coal and waste plastics determines that its nodes are invariant as a graph (there are no new changes in reactants and reaction conditions) during the reaction process. However, as time goes by, the data of each node will change. Therefore, it is difficult to define a convolutional neural network on the domain of each node. The spectral method uses the convolution theorem to define graph convolution from the spectral domain.
[0033] The spectral method uses the convolution theorem to define the convolution of the graph from the spectral domain. Convolution theorem for graph signal processing: The Fourier transform of the convolution of a signal is equivalent to the product of the Fourier transforms of the signal, and the formula is as follows:
[0034] F(f·g)=F(f)*F(g)
[0035] Among them, f and g represent two original signals, F(f) represents the Fourier transform of f, · represents the product operator, and * represents the convolution operator.
[0036] The inverse Fourier transform of both ends can be obtained:
[0037] f*g=F -1 (F(f)·F(g))
[0038] When the input parameters are encoded in the multi-relation GCN entity association aggregation, the R-GCN transfer formula begins to work, and the entire knowledge graph-graph neural network prediction training process is measured by the evaluation function, which decodes and calculates the edge data between nodes, and finally obtains a prediction model that meets the requirements.
[0039] The main prediction function is realized by predicting the reaction in a short time based on the algorithm developed above. That is, assuming that it takes T time for the reaction to occur completely, the prediction device described in the present invention only needs to complete t time and make a time-related prediction for the remaining Tt time reaction, and finally obtain the total amount of tar yield from the synergistic pyrolysis of coal and waste plastics.
[0040] like Figure 3 As shown, the embodiment of the present invention also provides a portable coal and waste plastic synergistic pyrolysis tar yield prediction device based on knowledge graph-graph neural network, which is characterized by comprising: an analysis and prediction module, a main experimental module, and a raw material processing module; wherein,
[0041] The analysis and prediction module is used in the knowledge graph-graph neural network algorithm to predict the tar yield of coal and waste plastic synergistic pyrolysis. The model analyzes the tar yield trend under different experimental conditions and realizes the final prediction of tar yield;
[0042] The main experimental module is used to obtain the content of each component; from bottom to top, they are the feed port, the synergistic pyrolysis chamber, the main insulation layer and the component analyzer; the specific process is as follows: the synergistic pyrolysis chamber is electrically heated by the on-board battery, and the post-processing raw materials of coal and waste plastics are input into the feed port to obtain the content of each component in the component analyzer in units of time; the main components include: H2, CH4, CO and CO2.
[0043] The raw material processing module includes: a grinder and a crusher for moderately grinding coal into coal particles suitable for thermal reaction in small devices, and crushing waste plastics.
[0044] Working principle: First, coal and waste plastics are fed in through the processing inlet, where they are broken into pieces and ground, and then enter the heating chamber through the feed channel; the entire heating chamber is a vertically structured cavity, which is surrounded by an electric thermal resistor on the outside and can be electrically heated by the on-board power supply to achieve the temperature conditions for the coordinated pyrolysis of coal and waste plastics; the entire device is surrounded by high-efficiency thermal insulation materials such as thermal insulation cotton, and is wrapped in an aluminum alloy shell on the outside; the top is equipped with a built-in tube for a flue gas analyzer to analyze the composition of the products of the coordinated pyrolysis of coal and waste plastics, and store them in the data interaction system. When an external computer prediction model is connected, the tar yield can be automatically predicted.
Claims
1. A portable coal and waste plastic synergistic pyrolysis tar yield prediction method based on knowledge graph-graph neural network, characterized in that: The following steps are involved: (1) Based on the synergistic pyrolysis experiment of coal and waste plastics, the entity relationships of the knowledge graph KG are divided; (2) Based on the existing experimental data of the synergistic pyrolysis of coal and waste plastics, the basic nodes of the graph neural network are constructed, and the experimental data are processed using the graph convolutional neural network; (3) The graph neural network is applied to the knowledge graph, and the information of the central node is updated by the edge nodes to aggregate the multi-layer network layer by layer; the edge relationship is modeled, and the R-GCN algorithm is used to obtain modeling information with direction and node connection; the Logic Rules method is used to train the overall structure and predict the influence of each node on the tar yield.
2. According to claim 1, a portable coal and waste plastic synergistic pyrolysis tar yield prediction method based on knowledge graph-graph neural network is characterized in that: Step (1) is as follows: There are 5 nodes as the structural units of the main knowledge graph, including: the waste plastic content, i.e. the percentage of the total reactants, the coal content, i.e. the percentage of the total reactants, the reaction time from the start of the reaction to the end of the reaction, the stable temperature after electric heating as the reaction temperature and the real-time cumulative tar yield.
3. According to claim 1, a portable coal and waste plastic synergistic pyrolysis tar yield prediction method based on knowledge graph-graph neural network is characterized in that: Step (2) is specifically as follows: compile the associated information in the knowledge graph, and use the graph convolutional neural network to import and evolve the functional relationship; wherein the experimental data includes: mixing waste plastics with a mixing ratio of 80% waste plastics and 20% coal, an experimental reaction with complete time data, a stable temperature in the overall experiment as a reference experimental temperature of the device, and the tar yield is the total tar yield.
4. According to claim 1, a portable coal and waste plastic synergistic pyrolysis tar yield prediction method based on knowledge graph-graph neural network is characterized in that: In step (3), the R-GCN algorithm assigns different weight ratios to edge relationships to distinguish different relationships; by adding potential reverse edge relationships, modeling information with direction and node connections is obtained.
5. A portable coal and waste plastic synergistic pyrolysis tar yield prediction device based on knowledge graph-graph neural network, characterized in that: include: Analysis and prediction module, main experiment module, raw material processing module; among them, The analysis and prediction module is used in the knowledge graph-graph neural network algorithm to predict the tar yield of coal and waste plastic synergistic pyrolysis. The model analyzes the tar yield trend under different experimental conditions and realizes the final prediction of tar yield; The main experimental module is used to obtain the content of each component; The raw material processing module is used to moderately grind coal into coal particles suitable for thermal reactions in small devices and to break up waste plastics.
6. A portable coal and waste plastic synergistic pyrolysis tar yield prediction device based on knowledge graph-graph neural network according to claim 5, characterized in that: The main experimental module includes a feed port, a synergistic pyrolysis chamber, a main insulation layer and a component analyzer. The specific process is as follows: the synergistic pyrolysis chamber is electrically heated by the on-board battery, and the post-processing raw materials of coal and waste plastics are input into the feed port to obtain the content of each component in time units in the component analyzer. The main components include: H2, CH4, CO and CO2.
7. A portable coal and waste plastic synergistic pyrolysis tar yield prediction device based on knowledge graph-graph neural network according to claim 5, characterized in that: The raw material processing module includes: grinder and chunk crusher.