Carbon footprint prediction method and device for mechanical and electrical products and medium

Abstract

The invention discloses a carbon footprint prediction method and device for an electromechanical product and a medium, belongs to the technical field of energy conservation and environmental protection, and aims to solve the technical problems that in the prior art, when carbon emission prediction is carried out at the initial stage of product design, considered influence factors are not comprehensive, and unified data expression of product design information is lacked. The method comprises the following steps: constructing a full-life-cycle low-carbon design model of the mechanical and electrical product; obtaining a design level in the design model; acquiring life cycle scene information corresponding to any design node in the design hierarchy; and determining a product carbon emission prediction result corresponding to the design level based on the life cycle scene information. According to the method, the product design information is expressed through the life cycle scene information, normative expression of the full life cycle information of the design scheme is realized, carbon emission prediction of the product is carried out based on the full life cycle scene information involved in the low-carbon design process of the product, and consideration factors are more comprehensive.

Description

Technical Field

[0001] The present invention relates to the technical field of energy conservation and environmental protection, and in particular to a method, device and medium for predicting the carbon footprint of an electromechanical product. Background Art

[0002] The increase in greenhouse gas emissions, especially carbon dioxide emissions, has led to serious environmental problems and may damage the ecological balance. Data from the Environmental Investigation Agency show that various products cause nearly 84% of energy-related carbon dioxide emissions and 90% of energy consumption, and are considered to be one of the main sources of greenhouse gas emissions. It is worth noting that the design stage affects more than 80% of the environmental impact of the product throughout its life cycle. Therefore, if the carbon emissions of products can be predicted in the early stages of design, it will provide effective support for enterprises to reduce carbon emissions.

[0003] In terms of design information modeling, existing models mainly focus on the structure and topological relationship of products. However, some design information that has a significant impact on carbon emissions has been neglected, such as energy type and geographical location. Fukushige, Umeda and others pointed out that regional differences can lead to changes in environmental regulations, resource requirements and labor costs, which will have a significant impact on product life cycle design. Kong et al. analyzed the environmental impact of design schemes under different energy types, and the results showed that the environmental impact of hydropower generation is 49.62% lower than that of thermal power generation. It can be seen that the difference in energy type has an important impact on the environment. In addition, existing life cycle inventory databases, including SimaPro and Ecoinvent, collect background data of different geography, time and quality. Selecting different background data will affect the results of product environmental assessment. In summary, the status of the product and the activities and interactions between the product and the external environment vary greatly throughout the life cycle stage. Corresponding to different energy and resource inputs and outputs, there are great differences in carbon emission results. Changes and assumptions in product life cycle activities have a profound impact on the carbon emissions of the product. However, the current low-carbon design theory has not yet systematically considered the mechanisms of these activities, which affects the accuracy of product carbon emissions.

[0004] In terms of carbon footprint prediction methods, knowledge reuse methods using artificial neural networks have become a hot topic and trend in current research. This method can solve new problems based on historical cases, thus providing valuable help for carbon emission prediction of various design schemes. However, the intricate interconnections and diversity of product design information during the life cycle, and the current lack of a unified data expression and management framework, pose challenges to the accurate prediction of product carbon emissions.

[0005] To sum up, when predicting product carbon emissions in the early stage of design, the existing technology mainly considers the impact of product structure and topological relationship on product carbon footprint, that is, it mainly considers the impact of product function layer and structural layer on carbon footprint, but lacks the impact of product design feature layer and processing technology layer on product carbon footprint, that is, the influencing factors considered are not comprehensive, which affects the accuracy of product carbon emission prediction; and there is a lack of unified data expression for product design information such as product function layer, structural layer, design feature layer and processing technology layer, which is not conducive to accurate prediction of product carbon emissions in the early stage of product design. Summary of the invention

[0006] The present invention provides a carbon footprint prediction method, device and medium for electromechanical products, which are used to solve the technical problems that the existing technology does not consider all the influencing factors when predicting the carbon emissions of products in the early stage of design, and lacks a unified data expression for product design information, thereby reducing the accuracy of product carbon emission prediction.

[0007] In a first aspect, the present invention provides a method for predicting the carbon footprint of an electromechanical product, comprising:

[0008] Constructing a low-carbon design model for the entire life cycle of the electromechanical product;

[0009] Obtaining a design level in the design model;

[0010] Obtaining life cycle scenario information corresponding to any design node in the design hierarchy;

[0011] Based on the life cycle scenario information, a product carbon emission prediction result corresponding to the design level is determined.

[0012] In a feasible implementation manner, the design hierarchy in the design model includes a design feature layer, and obtaining life cycle scenario information corresponding to any design node in the design hierarchy includes:

[0013] Obtain scene information in a raw material acquisition scene corresponding to any design node in the design feature layer, wherein the scene information in the raw material acquisition scene includes at least one of material type, part size, material quality, part shape, geographic location and energy type.

[0014] In a feasible implementation manner, the design hierarchy in the design model includes a processing technology layer, and obtaining life cycle scenario information corresponding to any design node in the design hierarchy includes:

[0015] Obtain scene information in a processing and manufacturing scene corresponding to any design node in a processing technology layer, wherein the scene information in the processing and manufacturing scene includes at least one of a process type, processing power, processing time, energy type, processing equipment type, and geographic location.

[0016] In a feasible implementation manner, the design hierarchy in the design model includes a functional layer, and obtaining life cycle scenario information corresponding to any design node in the design hierarchy includes:

[0017] Obtain scenario information in a usage scenario corresponding to any design node in the functional layer, wherein the scenario information in the usage scenario includes at least one of function type, service life, usage condition, energy type, operating power and geographical location.

[0018] In a feasible implementation manner, the design hierarchy in the design model includes a structure layer, and obtaining life cycle scenario information corresponding to any design node in the design hierarchy includes:

[0019] The scene information in the recycling processing scene corresponding to any design node in the structure layer is obtained, wherein the scene information in the recycling processing scene includes at least one of structure type, recycling quality, recycling time, recycling power, energy type and geographical location.

[0020] In a feasible implementation manner, after determining the product carbon emission prediction result corresponding to the design level based on the life cycle scenario information, the method further includes:

[0021] The sum of the product carbon emission prediction results corresponding to all design levels in the design model is calculated, and the sum of the product carbon emission prediction results is determined as the carbon emission prediction result corresponding to the entire life cycle of the product.

[0022] In a feasible implementation, based on the life cycle scenario information, determining the product carbon emission prediction result corresponding to the design level includes:

[0023] The product carbon emission prediction result corresponding to the design level is determined based on the life cycle scenario information through a convolutional neural network model.

[0024] In a feasible implementation manner, before determining the product carbon emission prediction result corresponding to the design level based on the life cycle scenario information through a convolutional neural network model, the method further includes:

[0025] Training the convolutional neural network model using sample data to obtain initial values ​​of parameters of the convolutional neural network model;

[0026] If the initial values ​​of the parameters of the convolutional neural network model meet the preset performance indicators of the convolutional neural network model, the initial values ​​of the parameters of the convolutional neural network model are determined as the values ​​of the optimal parameters to obtain an optimized convolutional neural network model;

[0027] If the initial values ​​of the parameters of the convolutional neural network model do not meet the preset performance indicators of the convolutional neural network model, updating the parameters of the convolutional neural network model by a differential evolution algorithm until the parameters meet the termination conditions of the differential evolution algorithm;

[0028] If the parameters meet the preset performance indicators of the convolutional neural network model, the parameters are determined as optimal parameters to obtain an optimized convolutional neural network model;

[0029] If the parameters do not meet the preset performance indicators of the convolutional neural network model, the parameters of the convolutional neural network model are repeatedly updated through the differential evolution algorithm until the parameters meet the preset performance indicators of the convolutional neural network model.

[0030] In a second aspect, the present invention provides a carbon footprint prediction device for an electromechanical product, the device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor so that the at least one processor can execute a carbon footprint prediction method for an electromechanical product as described in any of the above-mentioned embodiments.

[0031] In a third aspect, the present invention provides a non-volatile computer storage medium, which is a non-volatile computer-readable storage medium, and the non-volatile computer-readable storage medium stores at least one program, each of which includes instructions, and when the instructions are executed by a terminal, the terminal executes a carbon footprint prediction method for an electromechanical product described in any of the above embodiments.

[0032] The carbon footprint prediction method, device and medium of an electromechanical product provided by the present invention have the following beneficial technical effects compared with the prior art:

[0033] (1) In view of the problems of low-carbon information extraction and design scheme expression caused by the multi-source heterogeneity of information and different activities in the low-carbon design process of electromechanical products, the present invention proposes a low-carbon design multi-level modeling method integrating life cycle scenarios: identifying and defining life cycle scenarios to systematically describe the carbon footprint related activities of the design scheme, and constructing a multi-level low-carbon product design model based on life cycle scenarios. Each design level in the design model corresponds to each stage in the entire life cycle of the product. When predicting the carbon emissions of the product, considering any design level in the design model also takes into account any life cycle stage in the entire life cycle involved in the product low-carbon design process, and the factors affecting the carbon emissions of the product considered are more comprehensive. Furthermore, the design nodes in the design level of the design model are associated with the life cycle scenario information. The product design information is expressed through the life cycle scenario information, thereby realizing the standardized expression of the entire life cycle information of the design scheme, and predicting the carbon emissions of the product based on the entire life cycle scenario information involved in the product low-carbon design process.

[0034] (2) The method proposed in the present invention includes optimizing the CNN model through the DE algorithm, and obtaining a multi-layer perceptron training network model integrating DE and CNN. DE has significant efficiency and accuracy in guiding the training of the CNN model, thereby ensuring the construction of an efficient and reasonable prediction model. Driven by life cycle scenarios, efficient and accurate estimation of product carbon footprint is achieved, the efficiency and accuracy of carbon footprint estimation are improved, and strong support is provided for early judgment of the pros and cons of the scheme and exploration of the potential for scheme improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0036] Figure 1 A flow chart of a method for predicting the carbon footprint of an electromechanical product provided by the present invention;

[0037] Figure 2 A schematic diagram of the corresponding relationship between a product design model and life cycle stages provided by the present invention;

[0038] Figure 3 A schematic diagram of a multi-level low-carbon design model for products based on life cycle scenarios provided by the present invention;

[0039] Figure 4 The present invention provides a flowchart of a method for optimizing CNN based on DE;

[0040] Figure 5A schematic diagram of the mapping relationship between the design level and life cycle scenario of a gearbox provided by the present invention;

[0041] Figure 6 A schematic diagram of the input and output of a training set and the correlation coefficient R provided by the present invention;

[0042] Figure 7 A schematic diagram of the input and output of a verification set and the correlation coefficient R provided by the present invention;

[0043] Figure 8 A schematic diagram of the input and output of a test set and the correlation coefficient R provided by the present invention;

[0044] Fig. 9 A schematic diagram of the input and output of all sample data and the correlation coefficient R provided by the present invention;

[0045] Fig.10 A schematic diagram of an iterative curve fitting CNN-DE provided by the present invention;

[0046] Fig.11 A schematic diagram of an iteration curve of a single CNN provided by the present invention;

[0047] Fig.12 A schematic structural diagram of a carbon footprint prediction device for an electromechanical product provided by the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0049] The increase in greenhouse gas emissions, especially carbon dioxide emissions, has led to serious environmental problems and may damage the ecological balance. Data from the Environmental Investigation Agency show that various products cause nearly 84% of energy-related carbon dioxide emissions and 90% of energy consumption, and are considered to be one of the main sources of greenhouse gas emissions. Due to the implementation of the Kyoto Protocol and the Copenhagen Protocol, companies have begun to take measures to reduce greenhouse gas emissions from their products. It is worth noting that the design stage affects more than 80% of the environmental impact of the product throughout its life cycle. Therefore, if the carbon emissions of products can be predicted in the early stages of design, it will provide effective support for companies to reduce carbon emissions.

[0050] Generally speaking, the basic prerequisites for predicting product carbon emissions in the early design stage are: (1) a unified design model that standardizes the design information of the product throughout its life cycle; and (2) an effective and reliable carbon emission prediction method.

[0051] Design models can formally represent product design information. Generally, design modeling methods include two categories: unit process-oriented models and life cycle-oriented models. Unit process-oriented models mainly focus on specific design stages or specific design goals in the life cycle. Such models are prone to form information islands. Therefore, life cycle models containing product life cycle information have received widespread attention. Zhang et al. used extensional matter-element to represent the knowledge of green design units, and constructed ordered triples through design unit types, green design unit characteristics and their sizes, effectively supporting product green design. He et al. constructed a product life cycle decision space representing alternative life cycle paths through a five-layer weighted directed graph, and searched for low-carbon design solutions through the established decision space. In addition, Fu et al. proposed a constraint-driven product concept design model based on functional, behavioral and structural models, and considered constraint evaluation criteria in the model reasoning process to obtain the optimal design solution.

[0052] In the absence of design information in the early stage of design, an effective product carbon emission prediction method is urgently needed. As an ISO standardized method, the Life Cycle Assessment (LCA) method is widely used to determine the greenhouse gas emissions of a product's life cycle. This method is a quantitative evaluation process, which is based on the collection of complete life cycle inventory data. However, due to the lack of specific information in the design stage, it is difficult to obtain inventory data directly. To address this problem, two methods, Streamlined Life Cycle Assessment (SLCA) and data allocation methods, are proposed to simplify the life cycle assessment process. Specifically, SLCA implements LCA by using existing life cycle data and tools, simplifying life cycle stages, environmental impact types or data, etc. For example, Danilecki et al. simplified the data source and obtained test data from certified dismantling stations or the latest literature data to analyze the environmental performance of automobiles. In addition, the data allocation method mainly allocates the existing design information according to certain rules or criteria to obtain the inventory data required for LCA. For example, Zhang et al. believed that the product is an organic combination of connected units. Based on the analysis of the connection characteristics of each component, a recursive allocation method for carbon emissions of connected units was proposed based on the analytic hierarchy process to calculate the greenhouse gas emissions of connected units. The above methods have the advantages of being easy to use, cost-saving and time-saving, but the simplified information often affects the accuracy and reliability of the results. To solve this problem, a knowledge reuse method has been proposed, which can support the carbon emission prediction of product design schemes by capturing, verifying, storing and retrieving design information based on existing design knowledge. Knowledge reuse mainly includes case-based reasoning, rule-based reasoning, artificial neural network and other methods. Among them, artificial neural network is a data-driven supervised learning technology with nonlinear adaptive information processing capabilities. It can analyze association models by training and learning sample data to support efficient prediction of results. For example, considering the significant impact of buildings on the economy, society and environment, Tatari and Kucukvar constructed an artificial neural network model to predict the cost premium of green buildings that have obtained the Leadership in Energy and Environmental Design certification. Kuo et al. proposed a green supplier selection method that combines artificial neural network with two multi-attribute decision analysis methods, data envelopment analysis and network analysis method.

[0053] In terms of design information modeling, existing models mainly focus on the structure and topological relationship of products. However, some design information that has a significant impact on carbon emissions has been neglected, such as energy type and geographical location. Fukushige, Umeda and others pointed out that regional differences can lead to changes in environmental regulations, resource requirements and labor costs, which will have a significant impact on product life cycle design. Kong et al. analyzed the environmental impact of design schemes under different energy types, and the results showed that the environmental impact of hydropower generation is 49.62% lower than that of thermal power generation. It can be seen that the difference in energy type has an important impact on the environment. In addition, existing life cycle inventory databases, including SimaPro and Ecoinvent, collect background data of different geography, time and quality. Selecting different background data will affect the results of product environmental assessment. In summary, the status of the product and the activities and interactions between the product and the external environment vary greatly throughout the life cycle stage. Corresponding to different energy and resource inputs and outputs, there are great differences in carbon emission results. Changes and assumptions in product life cycle activities have a profound impact on the carbon emissions of the product. However, the current low-carbon design theory has not yet systematically considered the mechanisms of these activities, which affects the accuracy of product carbon emissions.

[0054] In terms of carbon footprint prediction methods, knowledge reuse methods using artificial neural networks have become a hot topic and trend in current research. This method can solve new problems based on historical cases, thus providing valuable help for carbon emission prediction of various design schemes. However, the intricate interconnections and diversity of product design information during the life cycle, and the current lack of a unified data expression and management framework, pose challenges to the accurate prediction of product carbon emissions.

[0055] In addition, neural network models are prone to local minima and slow convergence, so the performance of artificial neural networks needs to be enhanced to ensure accurate prediction of product carbon emissions.

[0056] Estimating carbon emissions in the early design stage of a product is of great significance for enterprises to reduce carbon emissions. However, the information in the design stage is multi-source and heterogeneous, and the inventory data is lacking and difficult to collect, which limits the expression of the design scheme and the accurate assessment of carbon emissions. Therefore, this application is aimed at the early design stage, and based on the life cycle scenario drive, a product carbon emission prediction method integrating convolutional neural network and differential evolution is proposed to solve this challenge. Specifically, the concept of life cycle scenario is proposed to represent all activities in each stage of the product design life cycle, and a multi-level low-carbon design model based on life cycle scenario is established to comprehensively express the design information of the entire life cycle of the product design scheme. On this basis, through the life cycle scenario drive, a product carbon emission prediction method combining integrated convolutional neural network and differential evolution algorithm is proposed. Among them, the convolutional neural network model is trained using the life cycle scenario and carbon emissions of the historical design scheme, and the model parameters are iteratively optimized by the differential evolution algorithm. This application provides designers with a method for achieving efficient and accurate carbon emission prediction in the early design stage, which can effectively support the low-carbon design of products.

[0057] The technical solution proposed by the present invention is described in detail below with reference to the accompanying drawings.

[0058] Figure 1 A flow chart of a method for predicting the carbon footprint of an electromechanical product provided by the present invention. Figure 1 As shown, the method includes the following execution steps:

[0059] S101. Build a low-carbon design model for the entire life cycle of electromechanical products.

[0060] The design model in the present invention includes multiple design levels in product design. The design levels in the design model correspond to the stages in the life cycle of the electromechanical product. All design levels in the design model correspond to the entire life cycle of the product, that is, correspond to each life cycle stage in the entire life cycle of the product.

[0061] For example, Figure 2 As shown, the design model includes a function layer, a structure layer, a design feature layer and a processing technology layer. The entire life cycle of a product includes the raw material acquisition stage, the processing and manufacturing stage, the use stage and the recycling and processing stage. The design feature layer in the design model corresponds to the raw material acquisition stage in the product life cycle, the processing technology layer in the design model corresponds to the processing and manufacturing stage in the product life cycle, the function layer in the design model corresponds to the use stage in the product life cycle, and the structure layer in the design model corresponds to the recycling and processing stage in the product life cycle.

[0062] S102: Obtain the design level in the design model.

[0063] The present invention is applicable to predicting product carbon emissions during the product design stage. The design model includes at least one design level, for example, Figure 2 A schematic diagram of the corresponding relationship between a product design model and a life cycle stage provided by the present invention, such as Figure 2 As shown, the product design model includes four design levels: function layer, structure layer, design feature layer and processing technology layer.

[0064] S103. Obtain life cycle scenario information corresponding to any design node in the design hierarchy.

[0065] Each design level in the product design model of the present invention includes at least one design node, for example, Figure 2 As shown, the functional layer includes four design nodes: F1, F2, F3 and F4; the structural layer includes six design nodes: S1, S2, S3, S4, S5 and S6; the design feature layer includes six design nodes: DF1, DF2, DF3, DF4, DF5 and DF6; and the processing technology layer includes six design nodes: MP1, MP2, MP3, MP4, MP5 and MP6.

[0066] like Figure 2 As shown, the product design model constructed by the present invention describes the framework of the product life cycle model, integrates product design information from the system association mechanism between function, structure, design features and processing technology, and maps the design information of each layer of the model to the life cycle stage: the function layer corresponds to the use stage, the structure layer corresponds to the recycling and processing stage, the design feature layer corresponds to the raw material acquisition stage, and the processing technology layer corresponds to the processing and manufacturing stage.

[0067] For each design node in the design hierarchy in the product design model, each design node corresponds to life cycle scenario information of a corresponding life cycle stage.

[0068] As a possible implementation method, the design hierarchy in the design model includes a design feature layer, and the life cycle scenario information corresponding to any design node in the design hierarchy is obtained, including:

[0069] Obtain scene information in a raw material acquisition scene corresponding to any design node in the design feature layer, wherein the scene information in the raw material acquisition scene includes at least one of material type, part size, material quality, part shape, geographic location, and energy type.

[0070] As a possible implementation method, the design hierarchy in the design model includes a processing technology layer, and the life cycle scenario information corresponding to any design node in the design hierarchy is obtained, including:

[0071] The scene information in the processing and manufacturing scene corresponding to any design node in the processing technology layer is obtained, and the scene information in the processing and manufacturing scene includes at least one of process type, processing power, processing time, energy type, processing equipment type and geographical location.

[0072] As a possible implementation method, the design hierarchy in the design model includes a functional layer, and the life cycle scenario information corresponding to any design node in the design hierarchy is obtained, including:

[0073] Obtain scenario information in a usage scenario corresponding to any design node in the functional layer, wherein the scenario information in the usage scenario includes at least one of function type, service life, usage condition, energy type, operating power and geographical location.

[0074] As a possible implementation method, the design hierarchy in the design model includes a structure layer, and the life cycle scenario information corresponding to any design node in the design hierarchy is obtained, including:

[0075] The scene information in the recycling processing scene corresponding to any design node in the structure layer is obtained, and the scene information in the recycling processing scene includes at least one of the structure type, recycling quality, recycling time, recycling power, energy type and geographical location.

[0076] In the complex process of product low-carbon design, the product status and the activities and interactions between the product and the external environment are different. The important information in the activities and interactions includes not only the input and output of energy and materials, but also the characteristics and qualitative description of the process activities. Based on this, considering the differences in life cycle activities, the present invention proposes and defines the concept of life cycle scenario (LS), which includes the activities and interactions between the product and the external environment in the scope of carbon emission assessment. This scenario can systematically describe the full life cycle activity information related to the carbon footprint of the product. It is essentially a set of background data related to product design elements and design processes, as shown in formula (1).

[0077] LS={Lifecycle_S, Entity_I, Node_I, Emission_C} (1)

[0078] Among them, Lifecycle_S represents the product life cycle stage, for example, the raw material acquisition stage, the processing and manufacturing stage, the use stage and the recycling stage; Entity_I represents the product design entity information containing ecological significance, that is, the life cycle scenario information. For example, for the raw material acquisition stage, the scenario information in the raw material acquisition scenario includes at least one of the material type, part size, material quality, part shape, geographical location and energy type; Node_I represents the design node in the design hierarchy corresponding to the life scenario information; Emission_C represents the carbon footprint of the design node under the scenario, that is, the product carbon emission prediction result corresponding to the life scenario information.

[0079] Based on the analysis of information characteristics at each stage, the life cycle scenarios are divided into: raw material acquisition scenario, processing and manufacturing scenario, usage scenario, and recycling and disposal scenario.

[0080] By integrating the interface between life cycle scenarios and multi-level low-carbon design models, a multi-level low-carbon design model based on life cycle scenarios is constructed, which can provide a comprehensive and complete description of the design information of the entire product life cycle stage, such as Figure 3 shown.

[0081] Figure 3 A schematic diagram of a multi-level low-carbon design model for products based on life cycle scenarios provided by the present invention is shown in FIG. Figure 3 As shown, the design model includes four design levels: function layer F, structure layer S, design feature layer DF and processing technology layer MP; the product life cycle activities include four life cycle stages: raw material acquisition, processing and manufacturing, use and recycling. For each life cycle stage, there are four types of life cycle scenario information: raw material acquisition scenario, processing and manufacturing scenario, use scenario and recycling scenario. Among them, the scenario information in the raw material acquisition scenario includes material type, part size, material quality, part shape, geographical location and energy type; the scenario information in the processing and manufacturing scenario includes process type, processing power, processing time, energy type, processing equipment type and geographical location; the scenario information in the use scenario includes function type, service life, operating conditions, energy type, operating power and geographical location; the scenario information in the recycling scenario includes structure type, recycling quality, recycling time, recycling power, energy type and geographical location.

[0082] S104. Determine the product carbon emission prediction result corresponding to the design level based on the life cycle scenario information.

[0083] The life scenario information in each design node in the design hierarchy of the present invention includes product design information related to carbon emissions in the design node. This information can be used to determine the product carbon emission prediction result of the design node, and then determine the product carbon emission prediction result corresponding to the design hierarchy.

[0084] For example, based on the life cycle scenario information of each design node in the design feature layer, the carbon emission prediction results of the products at each design node in the design feature layer can be determined, and then the carbon emission prediction results of the products corresponding to the design feature layer can be determined, that is, the carbon emission prediction results of the products in the raw material acquisition stage. Based on the life cycle scenario information of each design node in the processing technology layer, the carbon emission prediction results of the products at each design node in the processing technology layer can be determined, and then the carbon emission prediction results of the products corresponding to the processing technology layer can be determined, that is, the carbon emission prediction results of the products in the processing and manufacturing stage. Based on the life cycle scenario information of each design node in the functional layer, the carbon emission prediction results of the products at each design node in the functional layer can be determined, and then the carbon emission prediction results of the products corresponding to the functional layers can be determined, that is, the carbon emission prediction results of the products in the use stage. Based on the life cycle scenario information of each design node in the structural layer, the carbon emission prediction results of the products at each design node in the structural layer can be determined, and then the carbon emission prediction results of the products corresponding to the structural layers can be determined, that is, the carbon emission prediction results of the products in the recycling and processing stage.

[0085] As a possible implementation method, after determining the product carbon emission prediction results corresponding to the design level based on the life cycle scenario information in S104, the method also includes: calculating the sum of the product carbon emission prediction results corresponding to all design levels in the design model, and determining the sum of the product carbon emission prediction results as the carbon emission prediction result corresponding to the entire life cycle of the product.

[0086] Exemplarily, as described in the above example, the design model includes four design levels, namely, structure layer, function layer, design feature layer and processing technology layer. After respectively determining the product carbon emission prediction results corresponding to the above structure layer, function layer, design feature layer and processing technology layer, the product carbon emission prediction results of the above four design levels are summed up and calculated. The summed result is the sum of the product carbon emission prediction results of the four life cycle stages, namely, the raw material acquisition stage, the processing and manufacturing stage, the use stage and the recycling and treatment stage, that is, the carbon emission prediction result corresponding to the entire life cycle of the product.

[0087] As a possible implementation method, S104 determines the product carbon emission prediction results corresponding to the design level based on the life cycle scenario information, including:

[0088] Through the convolutional neural network model, based on the life cycle scenario information, the product carbon emission prediction results corresponding to the design level are determined.

[0089] like Figure 4 As shown in the figure, the structure of the Convolutional Neural Networks (CNN) model mainly consists of five layers: input layer, convolution layer, pooling layer, fully connected layer and output layer, which are specifically introduced as follows.

[0090] The input layer represents the attributes of the design solution life cycle scenario, including raw material acquisition scenario, manufacturing scenario, usage scenario and recycling scenario, that is, life cycle scenario information. The input data is divided into two different types: qualitative and quantitative. Qualitative information is mainly conveyed in text format, and the word embedding technology in Word2Vec is used for conversion encoding process to map each word to a high-dimensional vector space. These vectors can capture the semantic relationship between words, enabling them to participate in the training process. The encoded qualitative information is then integrated with the quantitative information into a complete feature vector as the input of the CNN model, and all input data are normalized to the interval [0,1] through the max-min method.

[0091] The convolution layer is mainly responsible for extracting features from the input data (life cycle scenario information). This layer effectively reduces the number of model parameters by sliding the convolution kernel on the input data, performing local connections and weight sharing, while improving the efficiency and generalization of feature extraction. The ReLU activation function is used to perform nonlinear transformation on the output, which can extract local features from the data, thereby improving the expression ability of the model. The calculation process of this layer is shown in formula (2).

[0092]

[0093] In formula (2), relu() represents the activation function, conv() represents the convolution function, and x i represents the input feature map of the convolution kernel, k i represents the convolution kernel itself, b i is the bias parameter, A i represents the feature map generated by each layer, and M represents the total number of feature maps.

[0094] The pooling layer reduces the number of model parameters and computational complexity by reducing the dimension of the data. The maximum pooling method is used to retain the largest eigenvalue in each region, as shown in formula (3).

[0095] P i =max(Ai(t)) (3)

[0096] In formula (3), P i represents the i-th output vector of the pooling layer, and t represents the pooling width.

[0097] The fully connected layer is a bridge between the features extracted by the convolutional and pooling layers. It integrates these features and prepares them for the final stage of the convolutional neural network.

[0098] The output layer represents the output node of the variable, and outputs the product carbon emission prediction result corresponding to the life cycle scenario information, that is, the predicted carbon emission of the design scheme. A single neuron is used to output the predicted carbon emission of the design scheme using a linear activation function, as shown in formula (4).

[0099] Out j =f(w j p j-1 +b j ) (4)

[0100] In formula (4), f() represents the activation function, w j represents the weight between the jth layer and the j+1th layer, P j-1 represents the output of the j-1th layer, b j Indicates the bias parameter of the jth layer, Out j is the output of the jth layer.

[0101] During the training process, the Mean Square Error (MSE) is used to evaluate the performance of the model by comparing the predicted values ​​with the true values, as shown below:

[0102]

[0103] In formula (5), N is the number of samples; y targ is the true value, y pred The network output predicted by the CNN model.

[0104] In summary, the CNN method refers to obtaining a training model based on input information (life cycle scenario information) and output information (corresponding carbon emission prediction results). Its main purpose is to generate a model by adjusting parameters to minimize the MSE between the predicted value and the true value. On this basis, the generated model is used to predict the carbon emissions of the design solution.

[0105] It should be noted that the convolutional neural network needs to adjust parameters according to sample data during training. However, the training process of CNN is slow, its global search capability is limited, and it is prone to problems such as local extreme values ​​and data overfitting. Differential Evolution Algorithm (DE) is a random direct search method with robust global search capability and stability, which makes up for the shortcomings encountered in the CNN training process.

[0106] As a possible implementation method, the present invention proposes a carbon footprint prediction method that integrates CNN and DE. This method not only utilizes the powerful feature extraction capability of CNN, but also has strong convergence and learning capabilities through the iterative operation of DE. CNN can efficiently obtain training models based on historical knowledge to predict product carbon emissions, and use DE to optimize the weights and bias parameters of the training model, thereby avoiding the problem of easily falling into local extreme values ​​and overfitting data. Figure 4 The present invention provides a flowchart of a method for optimizing CNN based on DE, such as Figure 4 As shown, before S104, the following steps are also included.

[0107] S1041: Determine the structure of the convolutional neural network and set the hyperparameters of the CNN, including pooling width, learning rate, etc.

[0108] S1042: Input all sample data into the network, including input data (ie, life cycle scenario information) and output data (ie, carbon emissions).

[0109] S1043: Use CNN to train the sample data to obtain the initial values ​​of the parameters of the convolutional neural network model, that is, the initial values ​​of the weights and bias parameters of the convolutional neural network.

[0110] If the initial values ​​of the parameters of the convolutional neural network model meet the preset performance indicators of the convolutional neural network model, the initial values ​​of the parameters are determined as the values ​​of the optimal parameters to obtain an optimized convolutional neural network model.

[0111] If the initial values ​​of the parameters of the convolutional neural network model do not meet the preset performance indicators of the convolutional neural network model, execute S1044-S1045.

[0112] S1044: Set the initial values ​​of weight and bias parameters as the initial population, and calculate the mean square error (MSE) of the population. The differential mutation operator and the crossover operator are set to F and Cr respectively.

[0113] S1045: The weights and bias parameters of the convolutional neural network model are used as variables, and the neural network model is updated through the differential evolution algorithm. The differential mutation adopts the DE / rand / 1 strategy, the crossover adopts the binomial crossover, and the dominant individuals are selected to enter the next generation through greedy selection. The weights and bias parameters are updated through DE, and the population is iteratively optimized based on the fitness function.

[0114] S1046: Repeat S1044-S1045 until the termination condition of the DE algorithm is met. The planting condition of the DE algorithm may be that the fitness of DE meets a preset error standard, or the number of iterations reaches a maximum number of iterations.

[0115] S1047: Output the optimal weights and bias parameters, i.e., the optimal parameters, to obtain an optimized convolutional neural network model, i.e., the product carbon emission prediction model.

[0116] Taking the carbon footprint prediction of wind turbine gearbox as an example, the effectiveness of this application method is verified. Table 1 shows the design scenario information of each stage of the gearbox life cycle. Figure 5 A schematic diagram of the mapping relationship between the design level and life cycle scenario of a gearbox provided by the present invention, based on Figure 5 The hierarchical structure of the transmission is used to arrange and combine different design scenarios, and 128 (2 to the seventh power) sets of design scenarios for the gearbox are obtained. The sample set is divided into three types: training set, validation set and test set, with a ratio of 70%, 15% and 15%. Carbon footprint prediction is performed using a prediction method based on CNN and DE.

[0117] Table 1 Design scenario information for each stage of the gearbox life cycle

[0118]

[0119] (1) Validity of prediction results

[0120] Through the training of CNN and DE, Figure 6 A schematic diagram of the input and output of a training set and the correlation coefficient R provided by the present invention, Figure 7 A schematic diagram of the input and output of a verification set and the correlation coefficient R provided by the present invention, Figure 8 A schematic diagram of the input and output of a test set and the correlation coefficient R provided by the present invention, Fig. 9 The figure is a schematic diagram of the input and output of all sample data and the correlation coefficient R provided by the present invention, which respectively shows the actual carbon footprint and the estimated carbon footprint of each design scheme in the training set and the test set. Table 2 is the evaluation result of the convolutional neural network model, which lists the mean absolute error (MAE), standard deviation (RMSE), correlation coefficient R and determination coefficient R of the training set, test set and validation set respectively. 2 evaluation results.

[0121] Table 2 Convolutional neural network model evaluation results

[0122]

[0123] The analysis results show that the correlation coefficient R value of the training set is 0.9843, and the corresponding determination coefficient R 2The correlation coefficient and the coefficient of determination are both close to 1, and the degree of fit is good, indicating that the accuracy of the trained DE-CNN model is very high. The correlation coefficient R values ​​of the test set and the validation set are 0.9635 and 0.9652, respectively, indicating that the carbon footprint estimation rates of the test data and the validation data are high. In order to evaluate the goodness of fit generated by CNN-DE, the R value of the overall data must be taken, and its value is 0.9788, indicating that the prediction results of the neural network are well fitted with the actual carbon emission values, which also indicates the accuracy of the estimation results. As shown in Table 2, the root mean square error and mean absolute error of the training model are close to zero, which means that the error between the estimated value and the actual value is small, proving the effectiveness of the neural network carbon footprint estimation model. Although the data set trained in this paper is relatively small, the evaluation indicators of the neural network carbon footprint estimation model are relatively good, verifying the feasibility and accuracy of the proposed method.

[0124] (2) Effectiveness of prediction method

[0125] In the training process of convolutional neural network, DE is introduced to enhance the optimization of convolutional neural network weights and bias parameters. In order to verify the effectiveness of the method proposed in this paper, Fig.10 A schematic diagram of an iterative curve fitting CNN-DE provided by the present invention, Fig.11 A schematic diagram of an iteration curve of a single CNN provided by the present invention.

[0126] like Fig.10 As shown in Figure 2, the CNN-DE algorithm converges to the optimal value through iteration in the 17th generation and tends to be stable, which shows that CNN-DE can quickly and effectively obtain the optimal result and improve the efficiency of neural network training. Fig.11 As shown in the figure, a single CNN stays at the optimal value after about the 38th generation, and there is an obvious local optimal stagnation, and the convergence speed and accuracy are both lower than those of CNN-DE. This result shows that the differential evolution algorithm effectively helps CNN find better weight and bias parameters, that is, the integrated CNN-DE model proposed in this application has certain advantages over individual neural network training, and CNN-DE converges faster and produces better results than individual CNN predictions. This highlights the significant efficiency and accuracy of DE in guiding neural network training, thereby ensuring the construction of an efficient and reasonable prediction model.

[0127] Corresponding to the above-mentioned embodiment, the present invention also provides a carbon footprint prediction device for electromechanical products. Fig.12A structural schematic diagram of a carbon footprint prediction device for an electromechanical product provided in an embodiment of the present invention, the device may include: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor so that the at least one processor can execute a carbon footprint prediction method for an electromechanical product as described in the above embodiment.

[0128] In a possible implementation of the present invention, the aforementioned at least one processor is capable of executing the following steps: constructing a low-carbon design model for the entire life cycle of electromechanical products; obtaining the design hierarchy in the design model; obtaining the life cycle scenario information corresponding to any design node in the design hierarchy; and determining the product carbon emission prediction result corresponding to the design hierarchy based on the life cycle scenario information.

[0129] In a specific implementation, the present invention can also be provided as a non-volatile computer storage medium, wherein the storage medium is a non-volatile computer-readable storage medium, and the non-volatile computer-readable storage medium stores at least one program, each program includes instructions, and when the instructions are executed by a terminal, the terminal executes a carbon footprint prediction method for an electromechanical product as described in the above embodiment.

[0130] In a possible implementation of the present invention, the aforementioned terminal executes to construct a low-carbon design model for the entire life cycle of electromechanical products; obtains the design hierarchy in the design model; obtains the life cycle scenario information corresponding to any design node in the design hierarchy; and determines the product carbon emission prediction result corresponding to the design hierarchy based on the life cycle scenario information.

[0131] The various embodiments of the present invention are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0132] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0133] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting the carbon footprint of electromechanical products, characterized in that: The method comprises: Constructing a low-carbon design model for the entire life cycle of the electromechanical product; Obtaining a design level in the design model; Obtaining life cycle scenario information corresponding to any design node in the design hierarchy; Based on the life cycle scenario information, a product carbon emission prediction result corresponding to the design level is determined.

2. The method for predicting the carbon footprint of electromechanical products according to claim 1, characterized in that: The design hierarchy in the design model includes a design feature layer, and obtaining life cycle scenario information corresponding to any design node in the design hierarchy includes: Obtain scene information in a raw material acquisition scene corresponding to any design node in the design feature layer, wherein the scene information in the raw material acquisition scene includes at least one of material type, part size, material quality, part shape, geographic location and energy type.

3. The method for predicting the carbon footprint of electromechanical products according to claim 1, characterized in that: The design hierarchy in the design model includes a processing technology layer, and obtaining life cycle scenario information corresponding to any design node in the design hierarchy includes: Obtain scene information in a processing and manufacturing scene corresponding to any design node in a processing technology layer, wherein the scene information in the processing and manufacturing scene includes at least one of a process type, processing power, processing time, energy type, processing equipment type, and geographic location.

4. The method for predicting the carbon footprint of electromechanical products according to claim 1, characterized in that: The design hierarchy in the design model includes a functional layer, and obtaining life cycle scenario information corresponding to any design node in the design hierarchy includes: Obtain scenario information in a usage scenario corresponding to any design node in the functional layer, wherein the scenario information in the usage scenario includes at least one of function type, service life, usage condition, energy type, operating power and geographical location.

5. The method for predicting the carbon footprint of electromechanical products according to claim 1, characterized in that: The design hierarchy in the design model includes a structure layer, and obtaining life cycle scenario information corresponding to any design node in the design hierarchy includes: The scene information in the recycling processing scene corresponding to any design node in the structure layer is obtained, wherein the scene information in the recycling processing scene includes at least one of structure type, recycling quality, recycling time, recycling power, energy type and geographical location.

6. The method for predicting the carbon footprint of electromechanical products according to claim 1, characterized in that: After determining the product carbon emission prediction result corresponding to the design level based on the life cycle scenario information, the method further includes: The sum of the product carbon emission prediction results corresponding to all design levels in the design model is calculated, and the sum of the product carbon emission prediction results is determined as the carbon emission prediction result corresponding to the entire life cycle of the product.

7. The method for predicting the carbon footprint of electromechanical products according to claim 1, characterized in that: Based on the life cycle scenario information, determining the product carbon emission prediction result corresponding to the design level, including: The product carbon emission prediction result corresponding to the design level is determined based on the life cycle scenario information through a convolutional neural network model.

8. The method for predicting the carbon footprint of electromechanical products according to claim 7, characterized in that: Before determining the product carbon emission prediction result corresponding to the design level based on the life cycle scenario information through a convolutional neural network model, the method further includes: Training the convolutional neural network model using sample data to obtain initial values ​​of parameters of the convolutional neural network model; If the initial values ​​of the parameters of the convolutional neural network model meet the preset performance indicators of the convolutional neural network model, the initial values ​​of the parameters of the convolutional neural network model are determined as the values ​​of the optimal parameters to obtain an optimized convolutional neural network model; If the initial values ​​of the parameters of the convolutional neural network model do not meet the preset performance indicators of the convolutional neural network model, updating the parameters of the convolutional neural network model by a differential evolution algorithm until the parameters meet the termination conditions of the differential evolution algorithm; If the parameters meet the preset performance indicators of the convolutional neural network model, the parameters are determined as optimal parameters to obtain an optimized convolutional neural network model; If the parameters do not meet the preset performance indicators of the convolutional neural network model, the parameters of the convolutional neural network model are repeatedly updated through the differential evolution algorithm until the parameters meet the preset performance indicators of the convolutional neural network model.

9. A carbon footprint prediction device for electromechanical products, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, so that the at least one processor can execute the carbon footprint prediction method for an electromechanical product according to any one of claims 1-8.

10. A non-volatile computer storage medium, characterized in that: The storage medium is a non-volatile computer-readable storage medium, and the non-volatile computer-readable storage medium stores at least one program, each of which includes instructions, and when the instructions are executed by the terminal, the terminal executes a carbon footprint prediction method for an electromechanical product according to any one of claims 1-8.

Images