Park planning method for accounting enterprise emission reduction energy efficiency based on graph neural network

Abstract

The invention discloses a park planning method for accounting enterprise emission reduction energy efficiency based on a graph neural network, and belongs to the technical field of low-carbon industrial park planning, and the method comprises the following steps: employing an Internet of Things device and a sensor network to collect and continuously monitor the data information of a park factory in real time; performing standardization processing on the collected data information, and then performing feature extraction on multi-source heterogeneous data by using a graph neural network; accounting the emission reduction energy efficiency of each factory in the park based on a dynamic optimization prediction model by using the extracted feature data; and carrying out benchmarking analysis on the emission reduction energy efficiency data of each factory in the park and a park planning target, and regulating and controlling the structure and management cost of the park factories exceeding a target range. According to the method, the collected data is cleaned and standardized, the multi-source heterogeneous data is subjected to feature extraction and fusion by using the graph neural network, the understanding of the spatial relationship can be enhanced through the graph neural network, and the richness and accuracy of data representation are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of low-carbon industrial park planning, and particularly relates to a park planning method for calculating the emission reduction energy efficiency of enterprises based on graph neural networks. Background Art

[0002] With the intensification of global climate change problems, achieving the carbon neutrality goal has become the common pursuit of governments and enterprises around the world. Existing zero-carbon park planning methods mainly rely on historical data for prediction and determine key control plant units and their improvement measures through static models; however, such methods have limitations in dealing with sudden changes, data acquisition difficulties, and computing resource requirements.

[0003] Chinese Patent CN116843060A discloses a carbon neutrality prediction method based on park planning, including the following steps: forming a carbon flow chain according to the products and production raw materials of each factory building; planning the positions of the factory buildings in the park according to the carbon flow chain to form a park model, and determining the internal logistics path according to the positions of the factory buildings and the carbon flow chain; determining the internal logistics direction of the park model; the patent has the following defects: (1) The disclosed technology in this patent mainly conducts carbon neutrality prediction through a static planning model, lacking the ability to adapt to dynamic changes. For example, changes in market demand, improvements in production processes, etc. may affect the carbon emissions of the park, but this method is difficult to quickly respond to these changes; (2) Some parks may have manufacturing, electronics, and steel industries at the same time. However, due to the large differences in emissions and recovery rates between different industries, different standards need to be divided, but this method lacks planning for multi-project factories. Summary of the Invention

[0004] Technical problems to be solved: Aiming at the technical problems existing in the background art, the present invention provides a park planning method for calculating the emission reduction energy efficiency of enterprises based on graph neural networks, which improves the accuracy and adaptability of the management of each factory in the park.

[0005] Technical solution: A park planning method for calculating the emission reduction energy efficiency of enterprises based on graph neural networks according to the present invention, the park planning method includes the following steps: Step 1: Use Internet of Things devices and sensor networks to collect energy production, energy consumption, social and economic activities, and environmental data of each process link in the whole life cycle of the production process of factories in the park in real time, and continuously monitor and collect the data information of the park factories; Step 2: Standardize the data information collected in Step 1, and then use graph neural networks to extract features from multi-source heterogeneous data; Step 3: Calculate the emission reduction energy efficiency of each factory in the park based on the feature data extracted in Step 2 using a dynamic optimization prediction model; Step 4: Compare the emission reduction energy efficiency data of each factory in the park with the park planning goals, and adjust the structure and management fees of the park factories that exceed the target range.

[0006] Preferably, the data information in Step 1 mainly includes: (1) Install weighing sensors and RFID tag management systems at each waste collection point respectively to record the weight and type of waste at each recycling point; (2) Install RFID tag management systems at the production lines and warehouses of each factory respectively to track the utilization rate data of raw materials, intermediate products and recycled materials; (3) Install electric energy meters and flow meters on the core production equipment of each factory respectively to monitor the energy consumption data in real time; (4) Install greenhouse gas sensors for carbon dioxide and methane at each factory respectively to monitor the emissions in real time, and use a combustion analyzer to monitor the waste gas components during the combustion process, and calculate the total carbon emissions of each factory in the park through the carbon footprint.

[0007] Preferably, the specific steps for data extraction in Step 2: Step 21: Regard each sensor as a node in the graph neural network, and define the weight of the edge according to the physical distance between the sensors; Step 22: The initial feature vector is generated from the sensor data, and the node feature vector is updated through multi-layer graph convolution operations; Step 23: Use the Adam optimizer to update the parameters, and extract the features of the real-time collected data.

[0008] Preferably, in Step 3, according to the waste recovery rate, material recycling utilization rate, and energy use efficiency coefficient of each factory, determine the emission reduction condition coefficient of each factory in the park, and its calculation formula is as follows: ; In the formula: G is the emission reduction condition coefficient of the factory predicted in the park; a is the waste recovery rate of the factory; c is the material recycling utilization rate of the factory; m is the energy use efficiency coefficient of the factory.

[0009] Preferably, in Step 3, the energy use efficiency coefficient is calculated through the energy consumption per unit output value and the relative carbon emission intensity, and its calculation formula is as follows: ; In the formula: m is the energy use efficiency coefficient of the factory; h is the energy consumption per unit output value; s is the relative carbon emission intensity.

[0010] Preferably, in Step 4, when G≥54%, it indicates that the emission reduction energy efficiency of the corresponding factory in the park meets the target range of this quarter, and the management fee for the next quarter will be reduced; When G < 54%, it indicates that the emission reduction energy efficiency of the corresponding factory in the park does not meet the target range of this quarter, and the management fee for the next quarter will be increased, or the factory will be phased out in an orderly manner.

[0011] Preferably, the management fee in step 4 includes administrative management fee and property management fee.

[0012] Compared with the prior art, the present invention has at least the following beneficial effects: 1. The present invention proposes a dynamic analysis method based on graph neural network (GNN), which combines real-time data collection of the Internet of Things, multi-source heterogeneous data fusion and dynamic optimization prediction model. Data is collected in real time through a sensor network, and GNN is used to capture spatial relationships and time series features to achieve dynamic monitoring and adjustment. Multi-industry adaptation is achieved through differential thresholds (such as emission reduction condition coefficient thresholds for manufacturing, electronics, and steel industries); 2. The present invention combines graph neural network with park emission reduction management to construct a lightweight spatio-temporal graph neural network. This model can effectively capture periodic and non-periodic features in time series data, and introduces a reinforcement learning mechanism, enabling the model to automatically adjust parameter settings when the external environment changes, solving the pain points that static models are difficult to handle dynamic data and complex spatial relationships. The combination of the feature extraction ability of GNN and the dynamic optimization model uses graph neural network to extract and fuse features of multi-source heterogeneous data, and through graph neural network, the understanding of spatial relationships can be enhanced, and the richness and accuracy of data representation can be improved; Introduce a real-time data-driven management fee adjustment mechanism, such as adjusting fees based on the emission reduction condition coefficient G, and encourage enterprises to reduce emissions through economic means, enhancing the adaptability and flexibility of the system; 3. According to the waste recovery rate, material recycling utilization rate, unit output value energy consumption and relative carbon emission intensity of the factory, the present invention can accurately and efficiently estimate the emission reduction effect of each factory in the park, and adjust the management fee for the next quarter according to the emission reduction effect of each factory in the park, which can encourage enterprises to optimize the production process, reduce carbon emissions, and can accurately control the entire life cycle of the production of the park factories, meeting the requirements of the carbon neutrality policy, and having significant social and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic diagram of the park planning process of the present invention; Figure 2 It is a structural diagram of the park planning adjustment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0014] For the purpose of making the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the following will be combined with the attached Figures 1 - 2A clear and complete description of the technical solutions of the embodiments of the present invention is given. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art belong to the scope of protection of the present invention.

[0015] As Figures 1 - 2 shown, a park planning method for calculating the enterprise emission reduction energy efficiency based on a graph neural network according to the present invention, the park planning method includes the following steps: (1) Using Internet of Things devices and sensor networks to collect energy production, energy consumption, social and economic activities, and environmental data of each process link in the whole life cycle of the production process of factories in the park in real time, and continuously monitoring and collecting the data information of the park factories. The data information mainly includes the following contents: (1) Install weighing sensors and RFID tag management systems at each waste collection point to record the weight and type of waste at each recycling point; installing weighing sensors at waste collection points can record the weight of waste in real time, and use classification tags to identify different types of waste; by equipping different waste containers with RFID tags, the readers of the RFID tag management system automatically record the amount of recyclable waste and the total amount of collected waste.

[0016] (2) Install RFID tag management systems at the production lines and warehouses of each factory to track the utilization rate data of raw materials, semi-finished products, and recycled materials.

[0017] (3) Install electric energy meters and flow meters on the core production equipment of each factory to monitor energy consumption data in real time, such as monitoring the consumption of other energy sources such as water and natural gas; and combine with the financial module of the park factory to obtain the total output value of the factory.

[0018] (4) Install greenhouse gas sensors for carbon dioxide and methane at each factory to monitor emissions in real time, and use a combustion analyzer to monitor the exhaust gas components during the combustion process, and calculate the total carbon emissions of each factory in the park through the carbon footprint.

[0019] (2) Standardize the collected data information, and then use a graph neural network to extract features from multi-source heterogeneous data. The specific steps of data extraction are as follows: (1) Regard each sensor as a node in the graph neural network, and define the weight of the edge according to the physical distance between the sensors.

[0020] (2) The initial feature vector is generated from sensor data, and the node feature vector is updated through multiple layers of graph convolution operations.

[0021] (3) Use the Adam optimizer for parameter update and extract features from the real-time collected data.

[0022] (III) Based on the dynamic optimization prediction model, calculate the emission reduction energy efficiency of each factory in the park for the extracted feature data: (1) Determine the emission reduction condition coefficient of each factory in the park according to the waste recovery rate, material recycling rate, and energy use efficiency coefficient of each factory. The calculation formula is as follows: ; In the formula: G is the emission reduction condition coefficient of the factory predicted in the park; a is the waste recovery rate of the factory; c is the material recycling rate of the factory; m is the energy use efficiency coefficient of the factory.

[0023] (2) The energy use efficiency coefficient is calculated through the energy consumption per unit output value and the relative carbon emission intensity. The calculation formula is as follows: ; In the formula: m is the energy use efficiency coefficient of the factory; h is the energy consumption per unit output value; s is the relative carbon emission intensity.

[0024] (IV) Compare the emission reduction energy efficiency data of each factory in the park with the park planning target, and adjust the structure and management fees of the park factories that exceed the target range. The management fees include administrative management fees and property management fees. When G≥54%, it indicates that the emission reduction energy efficiency of the corresponding factory in the park meets the target range of this quarter, and the management fee for the next quarter will be reduced; when G<54%, it indicates that the emission reduction energy efficiency of the corresponding factory in the park does not meet the target range of this quarter, and the management fee for the next quarter will be increased, or the factory will be phased out in an orderly manner.

[0025] The method of the present invention is used to calculate the emission reduction energy efficiency of the park factories. The specific example data is shown in 1-9.

[0026] Taking the data of Example 1 as an example, calculate the energy use efficiency coefficient of any factory in the park. The energy consumption per unit output value is taken as h = 0.6 (unit: tce / 10,000 yuan, energy consumption per unit output value = total energy consumption ÷ total output value of the factory). The relative carbon emission intensity is taken as s = 1.2 (unit: tCO2e / 10,000 yuan, relative carbon emission intensity = total carbon emissions ÷ total output value of the factory).

[0027] The calculated value of the energy use efficiency coefficient is: , the larger the value of the energy use efficiency coefficient of the park factory, the greater the carbon emissions per unit output value of the factory.

[0028] When predicting the emission reduction condition coefficient of a factory in an industrial park, the waste recovery rate of the factory is taken as a = 72% (waste recovery rate = recyclable waste volume ÷ total received waste volume × 100%). The material recycling utilization rate of the factory is taken as c = 28% (material recycling utilization rate = recycled material usage ÷ total material usage × 100%).

[0029] The calculated value of the emission reduction condition coefficient of the factory is as follows: , from which it can be seen that G ≥ 61%, and the emission reduction energy efficiency of this factory meets the emission reduction requirements of the electronics industry, so the management fee for the next quarter is reduced.

[0030] Table 1 Parameters and emission reduction energy efficiency results of Examples 1 - 9 of factories in the electronic industrial park:

[0031] .

[0032] From the data in Table 1, it can be seen that when the sample approaches infinity, there will be a dividing line to determine whether the management fee of the factory is reduced based on the emission reduction condition coefficient. That is, when G ≥ 75, it indicates that the emission reduction energy efficiency of the factory in the industrial park meets the emission reduction requirements of the manufacturing industry, and the management fee for the next quarter is reduced at this time; when 75 > G ≥ 61, it indicates that the emission reduction energy efficiency of the factory in the industrial park meets the emission reduction requirements of the electronics industry, and the management fee for the next quarter is reduced at this time; when 61 > G ≥ 54%, it indicates that the emission reduction energy efficiency of the factory in the industrial park meets the emission reduction requirements of the steel industry, and the management fee for the next quarter is reduced at this time.

[0033] The above is the preferred implementation mode of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A park planning method for calculating the energy efficiency of enterprise emission reduction based on graph neural network, characterized in that The park planning method includes the following steps: Step 1: Use Internet of Things devices and sensor networks to collect in real time the energy production, energy consumption, socio-economic activities, and environmental data of each process link in the whole life cycle of the production process in the park, and continuously monitor and collect the data information of the park factories; Step 2: Standardize the data information collected in Step 1, and then use graph neural networks to extract features from multi-source heterogeneous data; Step 3: Based on the dynamic optimization prediction model, calculate the emission reduction energy efficiency of each factory in the park using the feature data extracted in Step 2; Step 4: Compare and analyze the emission reduction energy efficiency data of each factory in each park with the park planning goals, and adjust the structure and management fees of the park factories that exceed the target range.

2. The park planning method for calculating the enterprise emission reduction energy efficiency based on the graph neural network according to claim 1, wherein The data information in Step 1 mainly includes: (1) Install weighing sensors and RFID tag management systems at each waste collection point to record the weight and type of waste at each recycling point; (2) Install RFID tag management systems at the production lines and warehouses of each factory to track the utilization rate data of raw materials, intermediate products, and recycled materials; (3) Install electricity meters and flow meters on the core production equipment of each factory to monitor energy consumption data in real time; (4) Install greenhouse gas sensors for carbon dioxide and methane in each factory to monitor emissions in real time, and use a combustion analyzer to monitor the waste gas components during the combustion process, and calculate the total carbon emissions of each factory in the park through the carbon footprint.

3. The park planning method for calculating the enterprise emission reduction energy efficiency based on the graph neural network according to claim 2, characterized in that, The specific steps for data extraction in Step 2: Step 21: Regard each sensor as a node in the graph neural network, and define the weight of the edge according to the physical distance between the sensors; Step 22: The initial feature vector is generated from sensor data, and the node feature vector is updated through multiple layers of graph convolution operations; Step 23: Use the Adam optimizer to update the parameters and extract features from the real-time collected data.

4. The park planning method based on graph neural network calculation of enterprise emission reduction energy efficiency according to claim 3 is characterized in that: In Step 3, according to the waste recovery rate, material recycling utilization rate, and energy use efficiency coefficient of each factory, determine the emission reduction condition coefficient of each factory in the park. The calculation formula is as follows: ； In the formula: G is the emission reduction condition coefficient of the factory predicted in the park; a is the waste recovery rate of the factory; c is the material recycling utilization rate of the factory; m is the energy use efficiency coefficient of the factory.

5. The park planning method based on graph neural network calculation of enterprise emission reduction energy efficiency according to claim 4 is characterized in that: In Step 3, the energy use efficiency coefficient is calculated through the energy consumption per unit output value and the relative carbon emission intensity. The calculation formula is as follows: ； In the formula: m is the energy use efficiency coefficient of the factory; h is the energy consumption per unit output value; s is the relative carbon emission intensity.

6. The park planning method for calculating the enterprise emission reduction energy efficiency based on the graph neural network according to claim 4, characterized in that, In Step 4, when G≥54%, it indicates that the emission reduction energy efficiency of the corresponding factory in the park meets the target range of this quarter, and the management fee for the next quarter will be reduced; When G<54%, it indicates that the emission reduction energy efficiency of the corresponding factory in the park does not meet the target range of this quarter, and the management fee for the next quarter will be increased, or the factory will be phased out in an orderly manner.

7. The park planning method based on graph neural network calculation of enterprise emission reduction energy efficiency according to claim 6 is characterized in that: The management fee in Step 4 includes administrative management fees and property management fees.

Images