Colored aluminum casting carbon emission accounting method and system based on hierarchical memory network

Abstract

The invention discloses a colored aluminum casting carbon emission accounting method based on a hierarchical memory network, and the method comprises the steps: firstly, determining the accounting time range and boundary of colored aluminum casting carbon emission, and obtaining the carbon emission time sequence data of different materials according to the carbon emission historical data of different materials in different production states in the accounting boundary; training and fitting carbon emission time sequence data of various materials in an accounting boundary by using a hierarchical memory network, and adjusting through a cruising algorithm to obtain optimal parameters of the hierarchical memory network; carbon emission factors of different materials in the accounting time range and the boundary in the actual production state are predicted according to the optimized hierarchical memory network, direct carbon emission and indirect carbon emission in the casting process of the colored aluminum are calculated through the predicted carbon emission factors, and therefore the total carbon emission amount is obtained. By means of the method, the colored aluminum casting carbon emission within the accounting time range and the boundary can be accurately calculated, and guarantee is provided for accounting of the colored aluminum casting carbon emission.

Description

Technical Field

[0001] The present invention belongs to the field of carbon emission accounting for non-ferrous aluminum smelting and casting, and more specifically, relates to a method and system for carbon emission accounting for non-ferrous aluminum smelting and casting based on a hierarchical memory network. Background Art

[0002] With the rapid development of China's aluminum alloy casting industry, carbon emission accounting methods and strategy analysis are particularly important for the aluminum casting industry. In order to achieve transparency of carbon emissions of aluminum production enterprises, it is necessary to uniformly stipulate the accounting time range and boundaries of aluminum casting, and at the same time, it is necessary to accurately calculate the carbon emissions generated during the non-ferrous aluminum casting process within the accounting time range and boundaries, so as to provide support for the subsequent production strategy adjustment and industrial upgrading of the enterprise. Therefore, in order to solve the above problems, it is of great significance to provide a non-ferrous aluminum casting carbon emission accounting method and system based on hierarchical memory network. Summary of the invention

[0003] In response to the above defects or improvement needs of the prior art, the present invention discloses a method and system for calculating carbon emissions from non-ferrous aluminum smelting based on a hierarchical memory network, thereby solving the problem that it is difficult to accurately calculate the carbon emissions generated in the non-ferrous aluminum smelting process within the calculation time range and boundaries.

[0004] To achieve the above objectives, according to one aspect of the present invention, a method for calculating carbon emissions from non-ferrous aluminum casting based on a hierarchical memory network is provided, and the specific method is as follows:

[0005] A1. Determine the time range and boundary for calculating carbon emissions from non-ferrous aluminum smelting based on the process of non-ferrous aluminum smelting;

[0006] A2. Obtain the carbon emission time series data of various materials by processing the historical data of different materials under different production conditions within the accounting boundary;

[0007] A3. Use hierarchical memory networks to train and fit the carbon emission time series data of various materials within the accounting boundary;

[0008] A4. Use the cruising algorithm to optimize the parameters of the hierarchical memory network;

[0009] A5. Predict the carbon emission factors of different materials under actual production conditions within the accounting time range and boundaries based on the trained hierarchical memory network;

[0010] A6. Calculate the direct and indirect carbon emissions in the non-ferrous aluminum smelting process based on the predicted carbon emission factors, and finally obtain the total carbon emissions.

[0011] Preferably, the accounting time range and boundary of carbon emissions from non-ferrous aluminum casting is determined according to the process of non-ferrous aluminum casting, and the historical data of different materials under different production conditions within the accounting boundary are processed to obtain the carbon emission time series data of various materials. The specific process is as follows:

[0012] Step 1: Determine the smelting and casting process of non-ferrous aluminum, including the process from aluminum concentrate to refined aluminum products and the process from scrap aluminum to refined aluminum products, and identify the carbon emission categories in the aluminum smelting and casting process. The carbon emission categories include direct carbon emissions and indirect carbon emissions. Direct carbon emissions include: fuel combustion, energy use as raw materials and industrial processes, and indirect carbon emissions include: net purchased electricity, heat and production emissions;

[0013] Step 2: According to the two aluminum smelting and casting processes and carbon emission categories, determine the emission sources that may generate carbon emissions in the actual smelting and casting process, and determine the accounting boundary of carbon emissions. At the same time, collect historical data of different materials under different production conditions, including: time, material name, carbon emission category, temperature, oxygen concentration, material input, carbon output and other information;

[0014] Step 3: Filter the acquired information and retain the required data, such as time, material input, actual carbon output, temperature, oxygen concentration, etc., to construct a carbon emission time series data set for different materials, and then perform maximum and minimum normalization on the obtained data, scale the data to [0,1], and perform maximum and minimum normalization on the data using the following formula:

[0015]

[0016] Among them, x is the actual value of the data to be normalized, x' is the normalized value, and x max 、x min are the maximum and minimum values ​​of the data.

[0017] Preferably, a hierarchical memory network is used to train and fit the carbon emission time series data of various materials within the accounting boundary. The specific process is as follows:

[0018] Step 1: For the number of hidden layers m and the delay memory coefficient δ of each hidden layer i , the number of neurons in each hidden layer The sparsity of the internal connection matrix of each hidden layer ξ i , the spectral radius ρ of the internal connection matrix of each hidden layer i , the Euclidean norm regularization constraint coefficient ε is initialized, where the number of hidden layers should not be too large, otherwise it will greatly increase the complexity of the calculation, and the delay memory coefficient should select an appropriate value according to the time span of the actual required long-term memory information;

[0019] Step 2: Initialize the input weight matrix and internal connection matrix of each hidden layer, and update the internal connection matrix according to the internal connection matrix sparsity, the maximum eigenvalue of the internal connection matrix and the spectral radius. The specific steps are as follows:

[0020] Step 2-1: Input weight matrix for each hidden layer Perform random assignment, where i = 1, 2, 3, ..., m. When i = 1, enter the weight matrix Each weight value in is within [-0.5, 0.5]. When i≠1, enter the weight matrix The weight values ​​in are within [-0.5, 0.5], where is the number of neurons in hidden layer i, D is the dimension of input layer parameters;

[0021] Step 2-2: Internal connection matrix of each hidden layer Random assignment, where i = 1, 2, 3, ..., m, internal connection matrix The weight values ​​in are within [-0.5, 0.5], where is the number of neurons in the hidden layer i. The number of non-zero elements in the internal connection matrix is ​​calculated according to the sparsity of the internal connection matrix of each hidden layer. The calculation formula for the number of non-zero elements in the internal connection matrix of each hidden layer is as follows:

[0022]

[0023] Among them, n i is the number of non-zero elements in the internal connection matrix of hidden layer i, ξ i is the internal connection matrix sparsity of hidden layer i, is the number of neurons in hidden layer i;

[0024] Step 2-3: Calculate the maximum eigenvalue and spectral radius of the internal connection matrix of each hidden layer, and update the internal connection matrix of each hidden layer according to the maximum eigenvalue and spectral radius of the internal connection matrix. The updating formula of the internal connection matrix of each hidden layer is as follows:

[0025]

[0026] in, is the internal connection matrix of hidden layer i, ρ i is the spectral radius of the internal connection matrix of hidden layer i, and W is Random matrices of the same dimension, where the values ​​of each element in the matrix are within [-0.5, 0.5], λ max (W) is the maximum eigenvalue of the internal connection matrix of the hidden layer;

[0027] Step 3: Establish a collection matrix, add the normalized carbon emission time series data to the collection matrix, calculate the time series data of each hidden layer neuron and add it to the collection matrix, and then obtain the final collection matrix. The specific process is as follows:

[0028] Step 3-1: Create an empty collection matrix The normalized carbon emission time series data u∈R D×T Add to the collection matrix, where D is the input layer parameter dimension and L is the output layer parameter dimension. is the number of neurons in hidden layer i, T is the length of the time series;

[0029] Step 3-2: Calculate the time series data of the first layer of neurons in the hidden layer. The calculation formula for the time series data of the first layer of neurons in the hidden layer is as follows:

[0030]

[0031] in, is the time series data of the first layer of neurons in the hidden layer, and f(·) is a nonlinear activation function, The internal connection matrix of the first hidden layer, is the data of the first layer of neurons in the hidden layer at time t-1, is the input weight matrix of the first hidden layer, and u(t) is the input carbon emission time series data;

[0032] Step 3-3: Calculate the time series data of the i-th (i=2, 3, 4, ..., m) layer of neurons in the hidden layer. The calculation formula for the time series data of the i-th layer of neurons in the hidden layer is as follows:

[0033]

[0034] Among them, δ j is the delayed memory coefficient of hidden layer j, θ i is the total delayed memory time of hidden layer i, is the input of hidden layer i at time t, is t-θ i The values ​​of each neuron in the hidden layer i-1 at the moment, is the time series data of the i-th layer neurons in the hidden layer, and f(·) is a nonlinear activation function, is the internal connection matrix of the i-th layer in the hidden layer, is the value of each neuron in hidden layer i at time ti, is the input weight matrix of hidden layer i;

[0035] Step 4: Calculate the output weight matrix using the collection matrix, actual output, and Euclidean norm regularization constraint coefficient. The output weight matrix calculation formula is as follows:

[0036]

[0037] in, is the output weight matrix, Y is the actual output value, X is the collection matrix, and ε is the Euclidean norm regularization constraint coefficient, which is usually a very small positive number.

[0038] Preferably, the cruising algorithm is used to optimize the parameters of the hierarchical memory network, and the optimized neural network is used to predict the carbon emission factors of different materials in the actual production state within the accounting time range and boundary. The specific process is as follows:

[0039] Step 1: Initialize the position and fitness of the fish school, and initialize the initial values ​​of parameters such as M, N, EN, PN, SN, and DN, where M is the maximum number of iterations, N is the size of the fish school, EN is the number of fish exploring and foraging in the fish school, PN is the number of fish sensing danger, SN is the safety value, and DN is the warning value (generated by a random number). The output weight matrix in the hierarchical memory network is used as a variable in the search space, and then a loop iteration is performed, and the fish school is sorted to obtain the individual position of the fish currently in the optimal position, as well as the optimal fitness value;

[0040] Step 2: Update the position of the exploratory fish according to the action characteristics of the exploratory fish. The position of the exploratory fish is updated by the following formula:

[0041]

[0042] Among them, t represents the current iteration number, L i,j represents the position information of the i-th fish in the j-th dimension, arctan is the inverse tangent function, c is a random number between [0,1], M is the maximum number of iterations, Q is a random number that follows a Gaussian distribution, R is a unit row vector, DN is the warning value, and SN is the safety value;

[0043] Step 3: Update the position of the foraging fish according to the position of the exploring foraging fish. Use the following formula to update the position of the foraging fish, that is,

[0044]

[0045] Among them, t represents the current iteration number, L i,j represents the position information of the i-th fish in the j-th dimension, arctan is the inverse tangent function, L p is a random position of the fish near the current optimal position, L worstis the worst position occupied by the fish, N is the size of the fish school, Q is a random number that follows a Gaussian distribution, R is a unit row vector, and A + is a row vector containing only random 1 and -1;

[0046] Step 4: Update the position of the fish group according to the reaction law of the fish group to the natural enemy. The position of the fish group is updated by the following formula:

[0047]

[0048] Among them, t represents the current iteration number, L i,j represents the position information of the i-th fish in the j-th dimension, α is the step size control parameter, K is a random parameter of [-1,1], and L best , L worst are the best and worst positions for fish, f i represents the fitness value of the current individual, f g Indicates the current maximum fitness value, f w Indicates the current minimum fitness value, γ is the smallest constant to avoid the denominator being 0;

[0049] Step 5: Update the historical optimal fitness. When the maximum number of iterations is reached, output the optimal individual position and fitness value. Use the global optimal solution as the output weight matrix of the hierarchical memory network to obtain the optimized hierarchical memory network. Input the carbon emission time series data of different materials within the accounting boundary into the optimized network to obtain the carbon emission factors of different materials in the actual production status within the accounting time range and boundary.

[0050] Preferably, the direct carbon emissions and indirect carbon emissions in the non-ferrous aluminum smelting process are calculated according to the predicted carbon emission factors, and the total carbon emissions are finally obtained. The specific process is as follows:

[0051] Step 1: Calculate the carbon emissions generated by fuel combustion during aluminum smelting and casting based on the predicted carbon emission factor. The carbon emissions generated by fuel combustion are calculated using the following formula:

[0052]

[0053] Among them, C 燃烧 To consume the carbon emissions from burning fossil fuels, is the activity level of the ith fossil fuel, is the carbon emission factor of the i-th fossil fuel; where the activity level of the i-th fossil fuel It is calculated by the following formula:

[0054]

[0055] Among them, NCV iis the average lower calorific value of the i-th fossil fuel, FC i is the net consumption of the ith fossil fuel;

[0056] Step 2: Calculate the carbon emissions generated by the use of energy as raw materials in the aluminum smelting process based on the predicted carbon emission factor. The carbon emissions generated by the use of energy as raw materials are calculated using the following formula:

[0057]

[0058] Among them, C 原材料 Carbon emissions from the use of energy as raw materials, is the activity level, i.e. the consumption of energy as a reducing agent, is the carbon emission factor when the jth energy is used as raw material;

[0059] Step 3: Calculate the carbon emissions generated by the industrial production process of aluminum smelting and casting based on the predicted carbon emission factor. The carbon emissions generated by the industrial production process are calculated using the following formula:

[0060] C 过程 =C 熔剂 +C 原料

[0061] Among them, the carbon emissions generated by flux consumption are calculated as follows:

[0062]

[0063] In the formula, C 熔剂 is the carbon emission generated by flux consumption, P k is the net consumption of the kth flux during the accounting period, Carbon emission factor of the kth flux;

[0064] Among them, the carbon emissions generated by the consumption of carbon-containing raw materials are calculated as follows:

[0065]

[0066] In the formula, C 原料 is the carbon emissions from the consumption of carbon-containing raw materials, M l To calculate the purchase amount of the first carbon-containing raw material within the time range and boundary, is the carbon emission factor of the lth purchased raw material, where l is the type of carbon-containing raw material;

[0067] Step 4: Calculate the carbon emissions generated by the net purchased electricity and heat used in the aluminum smelting process based on the predicted carbon emission factor. The carbon emissions generated by the net purchased electricity and heat are calculated using the following formula:

[0068] C电力、热力 =AD 电力 ×EF 电力 +AD 热力 ×EF 热力

[0069] Among them, C 电力、热力 To net purchase carbon emissions from electricity and heat, AD 电力 To calculate the net purchased electricity within the time range and boundary, AD 热力 The net purchase of heat within the accounting time range and boundary, EF 电力 is the carbon emission factor of electricity, EF 热力 is the carbon emission factor for heat;

[0070] Step 5: Sum the carbon emissions from fuel combustion, carbon emissions from energy used as raw materials, carbon emissions from industrial production processes, and carbon emissions from net purchased electricity and heat within the accounting time range to obtain the sum of direct and indirect carbon emissions, that is, the total carbon emissions.

[0071] According to another aspect of the present invention, a non-ferrous aluminum casting carbon emission accounting system based on a hierarchical memory network is provided, comprising:

[0072] An accounting time range and boundary determination and data processing device is used to determine the accounting time range and boundary of carbon emissions from non-ferrous aluminum smelting according to the process of non-ferrous aluminum smelting, and obtain carbon emission time series data of various materials by processing historical data of different materials under different production conditions within the accounting boundary;

[0073] A data training and fitting device, used to train and fit the carbon emission time series data of various materials within the accounting boundary using a hierarchical memory network;

[0074] A hierarchical memory network optimization and carbon emission factor prediction device is used to optimize the parameters of the hierarchical memory network using a cruising algorithm, and to predict the carbon emission factors of different materials in the actual production state within the accounting time range and boundary based on the trained network;

[0075] The carbon emission calculation device is used to calculate the direct carbon emissions and indirect carbon emissions in the non-ferrous aluminum smelting process according to the predicted carbon emission factors, and finally obtain the total carbon emissions.

[0076] Preferably, the accounting time range and boundary of carbon emissions from non-ferrous aluminum casting is determined according to the process of non-ferrous aluminum casting, and the historical data of different materials under different production conditions within the accounting boundary are processed to obtain the carbon emission time series data of various materials. The specific process is as follows:

[0077] Step 1: Determine the smelting and casting process of non-ferrous aluminum, including the process from aluminum concentrate to refined aluminum products and the process from scrap aluminum to refined aluminum products, and identify the carbon emission categories in the aluminum smelting and casting process. The carbon emission categories include direct carbon emissions and indirect carbon emissions. Direct carbon emissions include: fuel combustion, energy use as raw materials and industrial processes, and indirect carbon emissions include: net purchased electricity, heat and production emissions;

[0078] Step 2: According to the two aluminum smelting and casting processes and carbon emission categories, determine the emission sources that may generate carbon emissions in the actual smelting and casting process, and determine the accounting boundary of carbon emissions. At the same time, collect historical data of different materials under different production conditions, including: time, material name, carbon emission category, temperature, oxygen concentration, material input, carbon output and other information;

[0079] Step 3: Filter the acquired information and retain the required data, such as time, material input, actual carbon output, temperature, oxygen concentration, etc., to construct a carbon emission time series data set for different materials, and then perform maximum and minimum normalization on the obtained data, scale the data to [0,1], and perform maximum and minimum normalization on the data using the following formula:

[0080]

[0081] Among them, x is the actual value of the data to be normalized, x' is the normalized value, and x max 、x min are the maximum and minimum values ​​of the data.

[0082] Preferably, a hierarchical memory network is used to train and fit the carbon emission time series data of various materials within the accounting boundary. The specific process is as follows:

[0083] Step 1: For the number of hidden layers m and the delay memory coefficient δ of each hidden layer i , the number of neurons in each hidden layer The sparsity of the internal connection matrix of each hidden layer ξ i , the spectral radius ρ of the internal connection matrix of each hidden layer i , the Euclidean norm regularization constraint coefficient ε is initialized, where the number of hidden layers should not be too large, otherwise it will greatly increase the complexity of the calculation, and the delay memory coefficient should select an appropriate value according to the time span of the actual required long-term memory information;

[0084] Step 2: Initialize the input weight matrix and internal connection matrix of each hidden layer, and update the internal connection matrix according to the internal connection matrix sparsity, the maximum eigenvalue of the internal connection matrix and the spectral radius. The specific steps are as follows:

[0085] Step 2-1: Input weight matrix for each hidden layer Perform random assignment, where i = 1, 2, 3, ..., m. When i = 1, enter the weight matrix Each weight value in is within [-0.5, 0.5]. When i≠1, enter the weight matrix The weight values ​​in are within [-0.5, 0.5], where is the number of neurons in hidden layer i, D is the dimension of input layer parameters;

[0086] Step 2-2: Internal connection matrix of each hidden layer Random assignment, where i = 1, 2, 3, ..., m, internal connection matrix The weight values ​​in are within [-0.5, 0.5], where is the number of neurons in the hidden layer i. The number of non-zero elements in the internal connection matrix is ​​calculated according to the sparsity of the internal connection matrix of each hidden layer. The calculation formula for the number of non-zero elements in the internal connection matrix of each hidden layer is as follows:

[0087]

[0088] Among them, n i is the number of non-zero elements in the internal connection matrix of hidden layer i, ξ i is the internal connection matrix sparsity of hidden layer i, is the number of neurons in hidden layer i;

[0089] Step 2-3: Calculate the maximum eigenvalue and spectral radius of the internal connection matrix of each hidden layer, and update the internal connection matrix of each hidden layer according to the maximum eigenvalue and spectral radius of the internal connection matrix. The updating formula of the internal connection matrix of each hidden layer is as follows:

[0090]

[0091] in, is the internal connection matrix of hidden layer i, ρ i is the spectral radius of the internal connection matrix of hidden layer i, and W is Random matrices of the same dimension, where the values ​​of each element in the matrix are within [-0.5, 0.5], λ max (W) is the maximum eigenvalue of the internal connection matrix of the hidden layer;

[0092] Step 3: Establish a collection matrix, add the normalized carbon emission time series data to the collection matrix, calculate the time series data of each hidden layer neuron and add it to the collection matrix, and then obtain the final collection matrix. The specific process is as follows:

[0093] Step 3-1: Create an empty collection matrix The normalized carbon emission time series data u∈R D×T Add to the collection matrix, where D is the input layer parameter dimension and L is the output layer parameter dimension. is the number of neurons in hidden layer i, T is the length of the time series;

[0094] Step 3-2: Calculate the time series data of the first layer of neurons in the hidden layer. The calculation formula for the time series data of the first layer of neurons in the hidden layer is as follows:

[0095]

[0096] in, is the time series data of the first layer of neurons in the hidden layer, and f(·) is a nonlinear activation function, The internal connection matrix of the first hidden layer, is the data of the first layer of neurons in the hidden layer at time t-1, is the input weight matrix of the first hidden layer, and u(t) is the input carbon emission time series data;

[0097] Step 3-3: Calculate the time series data of the i-th (i=2, 3, 4, ..., m) layer of neurons in the hidden layer. The calculation formula for the time series data of the i-th layer of neurons in the hidden layer is as follows:

[0098]

[0099] Among them, δ j is the delayed memory coefficient of hidden layer j, θ i is the total delayed memory time of hidden layer i, is the input of hidden layer i at time t, is t-θ i The values ​​of each neuron in the hidden layer i-1 at the moment, is the time series data of the i-th layer neurons in the hidden layer, and f(·) is a nonlinear activation function, is the internal connection matrix of the i-th layer in the hidden layer, is the value of each neuron in hidden layer i at time ti, is the input weight matrix of hidden layer i;

[0100] Step 4: Calculate the output weight matrix using the collection matrix, actual output, and Euclidean norm regularization constraint coefficient. The output weight matrix calculation formula is as follows:

[0101]

[0102] in, is the output weight matrix, Y is the actual output value, X is the collection matrix, and ε is the Euclidean norm regularization constraint coefficient, which is usually a very small positive number.

[0103] Preferably, the cruising algorithm is used to optimize the parameters of the hierarchical memory network, and the optimized neural network is used to predict the carbon emission factors of different materials in the actual production state within the accounting time range and boundary. The specific process is as follows:

[0104] Step 1: Initialize the position and fitness of the fish school, and initialize the initial values ​​of parameters such as M, N, EN, PN, SN, and DN, where M is the maximum number of iterations, N is the size of the fish school, EN is the number of fish exploring and foraging in the fish school, PN is the number of fish sensing danger, SN is the safety value, and DN is the warning value (generated by a random number). The output weight matrix in the hierarchical memory network is used as a variable in the search space, and then a loop iteration is performed, and the fish school is sorted to obtain the individual position of the fish currently in the optimal position, as well as the optimal fitness value;

[0105] Step 2: Update the position of the exploratory fish according to the action characteristics of the exploratory fish. The position of the exploratory fish is updated by the following formula:

[0106]

[0107] Among them, t represents the current iteration number, L i,j represents the position information of the i-th fish in the j-th dimension, arctan is the inverse tangent function, c is a random number between [0,1], M is the maximum number of iterations, Q is a random number that follows a Gaussian distribution, R is a unit row vector, DN is the warning value, and SN is the safety value;

[0108] Step 3: Update the position of the foraging fish according to the position of the exploring foraging fish. Use the following formula to update the position of the foraging fish, that is,

[0109]

[0110] Among them, t represents the current iteration number, L i,j represents the position information of the i-th fish in the j-th dimension, arctan is the inverse tangent function, L p is a random position of the fish near the current optimal position, L worst is the worst position occupied by the fish, N is the size of the fish school, Q is a random number that follows a Gaussian distribution, R is a unit row vector, and A + is a row vector containing only random 1 and -1;

[0111] Step 4: Update the position of the fish group according to the reaction law of the fish group to the natural enemy. The position of the fish group is updated by the following formula:

[0112]

[0113] Among them, t represents the current iteration number, L i,j represents the position information of the i-th fish in the j-th dimension, α is the step size control parameter, K is a random parameter of [-1,1], and L best , L worst are the best and worst positions for fish, f i represents the fitness value of the current individual, f g Indicates the current maximum fitness value, f w Indicates the current minimum fitness value, γ is the smallest constant to avoid the denominator being 0;

[0114] Step 5: Update the historical optimal fitness. When the maximum number of iterations is reached, output the optimal individual position and fitness value. Use the global optimal solution as the output weight matrix of the hierarchical memory network to obtain the optimized hierarchical memory network. Input the carbon emission time series data of different materials within the accounting boundary into the optimized network to obtain the carbon emission factors of different materials in the actual production status within the accounting time range and boundary.

[0115] Preferably, the direct carbon emissions and indirect carbon emissions in the non-ferrous aluminum smelting process are calculated according to the predicted carbon emission factors, and the total carbon emissions are finally obtained. The specific process is as follows:

[0116] Step 1: Calculate the carbon emissions generated by fuel combustion during aluminum smelting and casting based on the predicted carbon emission factor. The carbon emissions generated by fuel combustion are calculated using the following formula:

[0117]

[0118] Among them, C 燃烧 To consume the carbon emissions from burning fossil fuels, is the activity level of the ith fossil fuel, is the carbon emission factor of the i-th fossil fuel; where the activity level of the i-th fossil fuel It is calculated by the following formula:

[0119]

[0120] Among them, NCV i is the average lower calorific value of the i-th fossil fuel, FC i is the net consumption of the ith fossil fuel;

[0121] Step 2: Calculate the carbon emissions generated by the use of energy as raw materials in the aluminum smelting process based on the predicted carbon emission factor. The carbon emissions generated by the use of energy as raw materials are calculated using the following formula:

[0122]

[0123] Among them, C 原材料 Carbon emissions from the use of energy as raw materials, is the activity level, i.e. the consumption of energy as a reducing agent, is the carbon emission factor when the jth energy is used as raw material;

[0124] Step 3: Calculate the carbon emissions generated by the industrial production process of aluminum smelting and casting based on the predicted carbon emission factor. The carbon emissions generated by the industrial production process are calculated using the following formula:

[0125] C 过程 =C 熔剂 +C 原料

[0126] Among them, the carbon emissions generated by flux consumption are calculated as follows:

[0127]

[0128] In the formula, C 熔剂 is the carbon emission generated by flux consumption, P k is the net consumption of the kth flux during the accounting period, Carbon emission factor of the kth flux;

[0129] Among them, the carbon emissions generated by the consumption of carbon-containing raw materials are calculated as follows:

[0130]

[0131] In the formula, C 原料 is the carbon emissions from the consumption of carbon-containing raw materials, M l To calculate the purchase amount of the first carbon-containing raw material within the time range and boundary, is the carbon emission factor of the lth purchased raw material, where l is the type of carbon-containing raw material;

[0132] Step 4: Calculate the carbon emissions generated by the net purchased electricity and heat used in the aluminum smelting process based on the predicted carbon emission factor. The carbon emissions generated by the net purchased electricity and heat are calculated using the following formula:

[0133] C 电力、热力 =AD 电力 ×EF 电力 +AD 热力 ×EF 热力

[0134] Among them, C 电力、热力 To net purchase carbon emissions from electricity and heat, AD电力 To calculate the net purchased electricity within the time range and boundary, AD 热力 The net purchase of heat within the accounting time range and boundary, EF 电力 is the carbon emission factor of electricity, EF 热力 is the carbon emission factor for heat;

[0135] Step 5: Sum the carbon emissions from fuel combustion, carbon emissions from energy used as raw materials, carbon emissions from industrial production processes, and carbon emissions from net purchased electricity and heat within the accounting time range to obtain the sum of direct and indirect carbon emissions, that is, the total carbon emissions.

[0136] According to another aspect of the present invention, there is provided a computer-readable storage medium having program instructions stored thereon, wherein the program instructions, when executed by a processor, implement any of the above-described methods for calculating carbon emissions from non-ferrous aluminum smelting based on a hierarchical memory network.

[0137] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0138] The present invention provides a method and system for calculating carbon emissions from non-ferrous aluminum smelting based on a hierarchical memory network. The hierarchical memory network optimized by the cruising algorithm can predict the carbon emission factors of different materials in the actual production state within the calculation time range and boundary, and calculate direct carbon emissions and indirect carbon emissions based on the predicted carbon emission factors, thereby obtaining the total carbon emissions, so that the carbon emissions in the aluminum smelting process can be more accurately calculated, which makes up for the defect of large calculation errors in existing carbon emission accounting methods. The hierarchical memory network used in the present accounting method and system can process input data of different lengths, and can remember previous inputs, and has a good delayed memory function. At the same time, multiple hidden layers help to better process the complex relationships between data, making it easier to discover the hidden relationships between data. The present invention can accurately calculate the carbon emissions in the aluminum smelting process, which is of great significance to the popularization and wide application of the carbon emission accounting method for non-ferrous aluminum smelting based on a hierarchical memory network in the future. BRIEF DESCRIPTION OF THE DRAWINGS

[0139] Figure 1 is a device connection diagram provided by an embodiment of the present invention;

[0140] Figure 2 It is a schematic flow chart of a method for calculating carbon emissions from non-ferrous aluminum smelting based on a hierarchical memory network provided by an embodiment of the present invention;

[0141] Figure 3It is a hierarchical memory network architecture diagram of a non-ferrous aluminum smelting carbon emission accounting method based on a hierarchical memory network provided in an embodiment of the present invention;

[0142] Figure 4 It is a calculation boundary and calculation diagram within the carbon emission calculation time range of a non-ferrous aluminum smelting carbon emission calculation method based on a hierarchical memory network provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0143] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0144] Example 1: Figure 1 As shown, an embodiment of the present invention provides a non-ferrous aluminum casting carbon emission accounting system based on a hierarchical memory network, the system comprising:

[0145] An accounting time range and boundary determination and data processing device is used to determine the accounting time range and boundary of carbon emissions from non-ferrous aluminum smelting according to the process of non-ferrous aluminum smelting, and obtain carbon emission time series data of various materials by processing historical data of different materials under different production conditions within the accounting boundary;

[0146] A data training and fitting device, used to train and fit the carbon emission time series data of various materials within the accounting boundary using a hierarchical memory network;

[0147] A hierarchical memory network optimization and carbon emission factor prediction device is used to optimize the parameters of the hierarchical memory network using a cruising algorithm, and to predict the carbon emission factors of different materials in the actual production state within the accounting time range and boundary based on the trained network;

[0148] The carbon emission calculation device is used to calculate the direct carbon emissions and indirect carbon emissions in the non-ferrous aluminum smelting process according to the predicted carbon emission factors, and finally obtain the total carbon emissions.

[0149] Example 2: Figure 2 As shown, a flow chart of a method for calculating carbon emissions from non-ferrous aluminum smelting based on a hierarchical memory network provided by an embodiment of the present invention includes the following steps:

[0150] Step A1: Determine the time range and boundary for calculating carbon emissions from non-ferrous aluminum smelting according to the process of non-ferrous aluminum smelting;

[0151] Step A2: Obtain the carbon emission time series data of various materials by processing the historical data of different materials under different production conditions within the accounting boundary;

[0152] Step A3: Use a hierarchical memory network to train and fit the carbon emission time series data of various materials within the accounting boundary;

[0153] Step A4: Optimize the parameters of the hierarchical memory network using the cruising algorithm;

[0154] Step A5: Predict the carbon emission factors of different materials under actual production conditions within the accounting time range and boundary based on the trained hierarchical memory network;

[0155] Step A6: Calculate the direct carbon emissions and indirect carbon emissions in the non-ferrous aluminum smelting process based on the predicted carbon emission factors, and finally obtain the total carbon emissions.

[0156] Example 3: Figure 3 As shown, a hierarchical memory network architecture diagram of a non-ferrous aluminum casting carbon emission accounting method based on a hierarchical memory network provided by an embodiment of the present invention uses a hierarchical memory network to train and fit the carbon emission time series data of various materials within the accounting boundary. The specific process is as follows:

[0157] Step 1: For the number of hidden layers m and the delay memory coefficient δ of each hidden layer i , the number of neurons in each hidden layer Q r i es , the sparsity of the internal connection matrix of each hidden layer ξ i , the spectral radius ρ of the internal connection matrix of each hidden layer i , the Euclidean norm regularization constraint coefficient ε is initialized, where the number of hidden layers should not be too large, otherwise it will greatly increase the complexity of the calculation, and the delay memory coefficient should select an appropriate value according to the time span of the actual required long-term memory information;

[0158] Step 2: Initialize the input weight matrix and internal connection matrix of each hidden layer, and update the internal connection matrix according to the internal connection matrix sparsity, the maximum eigenvalue of the internal connection matrix and the spectral radius. The specific steps are as follows:

[0159] Step 2-1: Input weight matrix for each hidden layer Perform random assignment, where i = 1, 2, 3, ..., m. When i = 1, enter the weight matrix Each weight value in is within [-0.5, 0.5]. When i≠1, enter the weight matrix The weight values ​​in are within [-0.5, 0.5], where is the number of neurons in hidden layer i, D is the dimension of input layer parameters;

[0160] Step 2-2: Internal connection matrix of each hidden layer Random assignment, where i = 1, 2, 3, ..., m, internal connection matrix The weight values ​​in are within [-0.5, 0.5], where is the number of neurons in the hidden layer i. The number of non-zero elements in the internal connection matrix is ​​calculated according to the sparsity of the internal connection matrix of each hidden layer. The calculation formula for the number of non-zero elements in the internal connection matrix of each hidden layer is as follows:

[0161]

[0162] Among them, n i is the number of non-zero elements in the internal connection matrix of hidden layer i, ξ i is the internal connection matrix sparsity of hidden layer i, is the number of neurons in hidden layer i;

[0163] Step 2-3: Calculate the maximum eigenvalue and spectral radius of the internal connection matrix of each hidden layer, and update the internal connection matrix of each hidden layer according to the maximum eigenvalue and spectral radius of the internal connection matrix. The updating formula of the internal connection matrix of each hidden layer is as follows:

[0164]

[0165] in, is the internal connection matrix of hidden layer i, ρ i is the spectral radius of the internal connection matrix of hidden layer i, and W is Random matrices of the same dimension, where the values ​​of each element in the matrix are within [-0.5, 0.5], λ max (W) is the maximum eigenvalue of the internal connection matrix of the hidden layer;

[0166] Step 3: Establish a collection matrix, add the normalized carbon emission time series data to the collection matrix, calculate the time series data of each hidden layer neuron and add it to the collection matrix, and then obtain the final collection matrix. The specific process is as follows:

[0167] Step 3-1: Create an empty collection matrix The normalized carbon emission time series data u∈R D×T Add to the collection matrix, where D is the input layer parameter dimension and L is the output layer parameter dimension. is the number of neurons in hidden layer i, T is the length of the time series;

[0168] Step 3-2: Calculate the time series data of the first layer of neurons in the hidden layer. The calculation formula for the time series data of the first layer of neurons in the hidden layer is as follows:

[0169]

[0170] in, is the time series data of the first layer of neurons in the hidden layer, and f(·) is a nonlinear activation function, The internal connection matrix of the first hidden layer, is the data of the first layer of neurons in the hidden layer at time t-1, is the input weight matrix of the first hidden layer, and u(t) is the input carbon emission time series data;

[0171] Step 3-3: Calculate the time series data of the i-th (i=2, 3, 4, ..., m) layer of neurons in the hidden layer. The calculation formula for the time series data of the i-th layer of neurons in the hidden layer is as follows:

[0172]

[0173] Among them, δ j is the delayed memory coefficient of hidden layer j, θ i is the total delayed memory time of hidden layer i, is the input of hidden layer i at time t, is t-θ i The values ​​of each neuron in the hidden layer i-1 at the moment, is the time series data of the i-th layer neurons in the hidden layer, and f(·) is a nonlinear activation function, is the internal connection matrix of the i-th layer in the hidden layer, is the value of each neuron in hidden layer i at time ti, is the input weight matrix of hidden layer i;

[0174] Step 4: Calculate the output weight matrix using the collection matrix, actual output, and Euclidean norm regularization constraint coefficient. The output weight matrix calculation formula is as follows:

[0175]

[0176] in, is the output weight matrix, Y is the actual output value, X is the collection matrix, and ε is the Euclidean norm regularization constraint coefficient, which is usually a very small positive number.

[0177] Example 4: Figure 4As shown, the carbon emission accounting boundary and calculation diagram within the carbon emission accounting time range of a non-ferrous aluminum smelting carbon emission accounting method based on a hierarchical memory network provided by an embodiment of the present invention, the calculation process of the carbon emissions within the accounting time range and boundary is:

[0178] Step 1: Calculate the carbon emissions generated by fuel combustion during aluminum smelting and casting based on the predicted carbon emission factor. The carbon emissions generated by fuel combustion are calculated using the following formula:

[0179]

[0180] Among them, C 燃烧 To consume the carbon emissions from burning fossil fuels, is the activity level of the ith fossil fuel, is the carbon emission factor of the i-th fossil fuel; where the activity level of the i-th fossil fuel It is calculated by the following formula:

[0181]

[0182] Among them, NCV i is the average lower calorific value of the i-th fossil fuel, FC i is the net consumption of the ith fossil fuel;

[0183] Step 2: Calculate the carbon emissions generated by the use of energy as raw materials in the aluminum smelting process based on the predicted carbon emission factor. The carbon emissions generated by the use of energy as raw materials are calculated using the following formula:

[0184]

[0185] Among them, C 原材料 Carbon emissions from the use of energy as raw materials, is the activity level, i.e. the consumption of energy as a reducing agent, is the carbon emission factor when the jth energy is used as raw material;

[0186] Step 3: Calculate the carbon emissions generated by the industrial production process of aluminum smelting and casting based on the predicted carbon emission factor. The carbon emissions generated by the industrial production process are calculated using the following formula:

[0187] C 过程 =C 熔剂 +C 原料

[0188] Among them, the carbon emissions generated by flux consumption are calculated as follows:

[0189]

[0190] In the formula, C 熔剂 is the carbon emission generated by flux consumption, P k is the net consumption of the kth flux during the accounting period, Carbon emission factor of the kth flux;

[0191] Among them, the carbon emissions generated by the consumption of carbon-containing raw materials are calculated as follows:

[0192]

[0193] In the formula, C 原料 is the carbon emissions from the consumption of carbon-containing raw materials, M l To calculate the purchase amount of the first carbon-containing raw material within the time range and boundary, is the carbon emission factor of the lth purchased raw material, where l is the type of carbon-containing raw material;

[0194] Step 4: Calculate the carbon emissions generated by the net purchased electricity and heat used in the aluminum smelting process based on the predicted carbon emission factor. The carbon emissions generated by the net purchased electricity and heat are calculated using the following formula:

[0195] C 电力、热力 =AD 电力 ×EF 电力 +AD 热力 ×EF 热力

[0196] Among them, C 电力、热力 To net purchase carbon emissions from electricity and heat, AD 电力 To calculate the net purchased electricity within the time range and boundary, AD 热力 The net purchase of heat within the accounting time range and boundary, EF 电力 is the carbon emission factor of electricity, EF 热力 is the carbon emission factor for heat;

[0197] Step 5: Sum the carbon emissions from fuel combustion, carbon emissions from energy used as raw materials, carbon emissions from industrial production processes, and carbon emissions from net purchased electricity and heat within the accounting time range to obtain the sum of direct and indirect carbon emissions, that is, the total carbon emissions.

[0198] In another embodiment of the present invention, a computer-readable storage medium is provided, on which program instructions are stored. When the program instructions are executed by a processor, a method and system for calculating carbon emissions from non-ferrous aluminum casting based on a hierarchical memory network as described above is implemented.

[0199] It should be pointed out that, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, and two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.

[0200] The method according to the present invention described above can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CD ROM, RAM, floppy disk, hard disk or magneto-optical disk), or as a computer code that is originally stored in a remote recording medium or a non-temporary machine-readable medium downloaded through a network and will be stored in a local recording medium, so that the method described herein can be stored in such software processing on a recording medium using a general-purpose computer, a special-purpose processor or programmable or special-purpose hardware (such as an ASIC or FPGA). It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code, and when the software or computer code is accessed and executed by a computer, a processor or hardware, the processing method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the processing shown here, the execution of the code converts the general-purpose computer into a special-purpose computer for executing the processing shown here.

[0201] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for calculating carbon emissions from non-ferrous aluminum casting based on a hierarchical memory network, characterized in that: include: The method comprises the following steps: A1. Determine the time range and boundary for calculating carbon emissions from non-ferrous aluminum smelting based on the process of non-ferrous aluminum smelting; A2. Obtain the carbon emission time series data of various materials by processing the historical data of different materials under different production conditions within the accounting boundary; A3. Use hierarchical memory networks to train and fit the carbon emission time series data of various materials within the accounting boundary; A4. Use the cruising algorithm to optimize the parameters of the hierarchical memory network; A5. Predict the carbon emission factors of different materials under actual production conditions within the accounting time range and boundaries based on the trained hierarchical memory network; A6. Calculate the direct and indirect carbon emissions in the non-ferrous aluminum smelting process based on the predicted carbon emission factors, and finally obtain the total carbon emissions.

2. The method for calculating carbon emissions from non-ferrous aluminum casting based on a hierarchical memory network according to claim 1 is characterized in that: In steps A1 and A2, the accounting time range and boundary of carbon emissions from non-ferrous aluminum casting are determined according to the process of non-ferrous aluminum casting, and the historical data of different materials under different production conditions within the accounting boundary are processed to obtain the carbon emission time series data of various materials. The process is as follows: Step 1: Determine the smelting and casting process of non-ferrous aluminum, including the process from aluminum concentrate to refined aluminum products and the process from scrap aluminum to refined aluminum products, and identify the carbon emission categories in the aluminum smelting and casting process. The carbon emission categories include direct carbon emissions and indirect carbon emissions. Direct carbon emissions include: fuel combustion, energy use as raw materials and industrial processes, and indirect carbon emissions include: net purchased electricity, heat and production emissions; Step 2: According to the two aluminum smelting and casting processes and carbon emission categories, determine the emission sources that may generate carbon emissions in the actual smelting and casting process, and determine the accounting boundary of carbon emissions. At the same time, collect historical data of different materials under different production conditions, including: time, material name, carbon emission category, temperature, oxygen concentration, material input, and carbon output information; Step 3: Filter the acquired information and retain the required data, including time, material input, actual carbon output, temperature, and oxygen concentration, to construct a carbon emission time series data set for different materials. Then perform maximum and minimum normalization on the obtained data, scale the data to [0,1], and perform maximum and minimum normalization on the data using the following formula: Among them, x is the actual value of the data to be normalized, x' is the normalized value, and x max 、x min are the maximum and minimum values ​​of the data.

3. The method for calculating carbon emissions from non-ferrous aluminum casting based on a hierarchical memory network according to claim 1 is characterized in that: In step A3, the hierarchical memory network is used to train and fit the carbon emission time series data of various materials within the accounting boundary. The process is as follows: Step 1: For the number of hidden layers m and the delay memory coefficient δ of each hidden layer i , the number of neurons in each hidden layer The sparsity of the internal connection matrix of each hidden layer ξ i , the spectral radius ρ of the internal connection matrix of each hidden layer i , the Euclidean norm regularization constraint coefficient ε is initialized; Step 2: Initialize the input weight matrix and internal connection matrix of each hidden layer, and update the internal connection matrix according to the internal connection matrix sparsity, the maximum eigenvalue of the internal connection matrix and the spectral radius. The specific steps are as follows: Step 2-1: Input weight matrix for each hidden layer Perform random assignment, where i = 1, 2, 3, ..., m. When i = 1, enter the weight matrix Each weight value in is within [-0.5, 0.5]. When i≠1, enter the weight matrix The weight values ​​in are within [-0.5, 0.5], where is the number of neurons in hidden layer i, D is the dimension of input layer parameters; Step 2-2: Internal connection matrix of each hidden layer Randomly assign values, where i = 1, 2, 3, ..., m, and the weight values ​​in the internal connection matrix are within [-0.5, 0.5], where is the number of neurons in the hidden layer i. The number of non-zero elements in the internal connection matrix is ​​calculated according to the sparsity of the internal connection matrix of each hidden layer. The calculation formula for the number of non-zero elements in the internal connection matrix of each hidden layer is as follows: Among them, n i is the number of non-zero elements in the internal connection matrix of hidden layer i, ξ i is the internal connection matrix sparsity of hidden layer i, is the number of neurons in hidden layer i; Step 2-3: Calculate the maximum eigenvalue and spectral radius of the internal connection matrix of each hidden layer, and update the internal connection matrix of each hidden layer according to the maximum eigenvalue and spectral radius of the internal connection matrix. The updating formula of the internal connection matrix of each hidden layer is as follows: in, is the internal connection matrix of hidden layer i, ρ i is the spectral radius of the internal connection matrix of hidden layer i, and W is Random matrices of the same dimension, where the values ​​of each element in the matrix are within [-0.5, 0.5], λ max (W) is the maximum eigenvalue of the internal connection matrix of the hidden layer; Step 3: Establish a collection matrix, add the normalized carbon emission time series data to the collection matrix, calculate the time series data of each hidden layer neuron and add it to the collection matrix, and then obtain the final collection matrix. The specific process is as follows: Step 3-1: Create an empty collection matrix The normalized carbon emission time series data u∈R D×T Add to the collection matrix, where D is the input layer parameter dimension and L is the output layer parameter dimension. is the number of neurons in hidden layer i, T is the length of the time series; Step 3-2: Calculate the time series data of the first layer of neurons in the hidden layer. The calculation formula for the time series data of the first layer of neurons in the hidden layer is as follows: in, is the time series data of the first layer of neurons in the hidden layer, and f(·) is a nonlinear activation function, The internal connection matrix of the first hidden layer, is the data of the first layer of neurons in the hidden layer at time t-1, is the input weight matrix of the first hidden layer, and u(t) is the input carbon emission time series data; Step 3-3: Calculate the time series data of the i-th (i=2, 3, 4, ..., m) layer of neurons in the hidden layer. The calculation formula for the time series data of the i-th layer of neurons in the hidden layer is as follows: Among them, δ j is the delayed memory coefficient of hidden layer j, θ i is the total delayed memory time of hidden layer i, is the input of hidden layer i at time t, is t-θ i The values ​​of each neuron in the hidden layer i-1 at the moment, is the time series data of the i-th layer neurons in the hidden layer, and f(·) is a nonlinear activation function, is the internal connection matrix of the i-th layer in the hidden layer, is the value of each neuron in hidden layer i at time ti, is the input weight matrix of hidden layer i; Step 4: Calculate the output weight matrix using the collection matrix, actual output, and Euclidean norm regularization constraint coefficient. The output weight matrix calculation formula is as follows: in, is the output weight matrix, Y is the actual output value, X is the collection matrix, and ε is the Euclidean norm regularization constraint coefficient.

4. The method for calculating carbon emissions from non-ferrous aluminum casting based on a hierarchical memory network according to claim 1 is characterized in that: In steps A4 and A5, the cruising algorithm is used to optimize the parameters of the hierarchical memory network, and the optimized neural network is used to predict the carbon emission factors of different materials in the actual production state within the accounting time range and boundary. The process is as follows: Step 1: Initialize the position and fitness of the fish school, and initialize the initial values ​​of parameters such as M, N, EN, PN, SN, and DN, where M is the maximum number of iterations, N is the size of the fish school, EN is the number of fish exploring and foraging in the fish school, PN is the number of fish sensing danger, SN is the safety value, and DN is the warning value (generated by a random number). The output weight matrix in the hierarchical memory network is used as a variable in the search space, and then a loop iteration is performed, and the fish school is sorted to obtain the individual position of the fish currently in the optimal position, as well as the optimal fitness value; Step 2: Update the position of the exploratory fish according to the action characteristics of the exploratory fish. The position of the exploratory fish is updated by the following formula: Among them, t represents the current iteration number, L i,j represents the position information of the i-th fish in the j-th dimension, arctan is the inverse tangent function, c is a random number between [0,1], M is the maximum number of iterations, Q is a random number that follows a Gaussian distribution, R is a unit row vector, DN is the warning value, and SN is the safety value; Step 3: Update the position of the foraging fish according to the position of the exploring foraging fish. Use the following formula to update the position of the foraging fish, that is, Among them, t represents the current iteration number, L i,j represents the position information of the i-th fish in the j-th dimension, arctan is the inverse tangent function, L p is a random position of the fish near the current optimal position, L worst is the worst position occupied by the fish, N is the size of the fish school, Q is a random number that follows a Gaussian distribution, R is a unit row vector, and A + is a row vector containing only random 1 and -1; Step 4: Update the position of the fish group according to the reaction law of the fish group to the natural enemy. The position of the fish group is updated by the following formula: Among them, t represents the current iteration number, L i,j represents the position information of the i-th fish in the j-th dimension, α is the step size control parameter, K is a random parameter of [-1,1], and L best , L worst are the best and worst positions for fish, f i represents the fitness value of the current individual, f g Indicates the current maximum fitness value, f w Indicates the current minimum fitness value, γ is the smallest constant to avoid the denominator being 0; Step 5: Update the historical optimal fitness. When the maximum number of iterations is reached, output the optimal individual position and fitness value. Use the global optimal solution as the output weight matrix of the hierarchical memory network to obtain the optimized hierarchical memory network. Input the carbon emission time series data of different materials within the accounting boundary into the optimized network to obtain the carbon emission factors of different materials in the actual production status within the accounting time range and boundary.

5. The method for calculating carbon emissions from non-ferrous aluminum casting based on a hierarchical memory network according to any one of claims 1 to 4, characterized in that: In step A6, the direct carbon emissions and indirect carbon emissions in the non-ferrous aluminum smelting process are calculated based on the predicted carbon emission factors, and the total carbon emissions are finally obtained. The process is as follows: Step 1: Calculate the carbon emissions generated by fuel combustion during aluminum smelting and casting based on the predicted carbon emission factor. The carbon emissions generated by fuel combustion are calculated using the following formula: Among them, C 燃烧 To consume the carbon emissions from burning fossil fuels, is the activity level of the ith fossil fuel, is the carbon emission factor of the i-th fossil fuel; where the activity level of the i-th fossil fuel It is calculated by the following formula: Among them, NCV i is the average lower calorific value of the i-th fossil fuel, FC i is the net consumption of the ith fossil fuel; Step 2: Calculate the carbon emissions generated by the use of energy as raw materials in the aluminum smelting process based on the predicted carbon emission factor. The carbon emissions generated by the use of energy as raw materials are calculated using the following formula: Among them, C 原材料 Carbon emissions from the use of energy as raw materials, is the activity level, i.e. the consumption of energy as a reducing agent, is the carbon emission factor when the jth energy is used as raw material; Step 3: Calculate the carbon emissions generated by the industrial production process of aluminum smelting and casting based on the predicted carbon emission factor. The carbon emissions generated by the industrial production process are calculated using the following formula: C 过程 =C 熔剂 +C 原料 Among them, the carbon emissions generated by flux consumption are calculated as follows: In the formula, C 熔剂 is the carbon emission generated by flux consumption, P k is the net consumption of the kth flux during the accounting period, Carbon emission factor of the kth flux; Among them, the carbon emissions generated by the consumption of carbon-containing raw materials are calculated as follows: In the formula, C 原料 is the carbon emissions from the consumption of carbon-containing raw materials, M l To calculate the purchase amount of the first carbon-containing raw material within the time range and boundary, is the carbon emission factor of the lth purchased raw material, where l is the type of carbon-containing raw material; Step 4: Calculate the carbon emissions generated by the net purchased electricity and heat used in the aluminum smelting process based on the predicted carbon emission factor. The carbon emissions generated by the net purchased electricity and heat are calculated using the following formula: C 电力、热力 =AD 电力 ×EF 电力 +AD 热力 ×EF 热力 Among them, C 电力、热力 To net purchase carbon emissions from electricity and heat, AD 电力 To calculate the net purchased electricity within the time range and boundary, AD 热力 The net purchase of heat within the accounting time range and boundary, EF 电力 is the carbon emission factor of electricity, EF 热力 is the carbon emission factor for heat; Step 5: Sum the carbon emissions from fuel combustion, carbon emissions from energy used as raw materials, carbon emissions from industrial production processes, and carbon emissions from net purchased electricity and heat within the accounting time range to obtain the sum of direct and indirect carbon emissions, that is, the total carbon emissions.

6. A non-ferrous aluminum casting carbon emission accounting system based on a hierarchical memory network, characterized in that: include: An accounting time range and boundary determination and data processing device is used to determine the accounting time range and boundary of carbon emissions from non-ferrous aluminum smelting according to the process of non-ferrous aluminum smelting, and obtain carbon emission time series data of various materials by processing historical data of different materials under different production conditions within the accounting boundary; Data training and fitting device, used to learn and train the carbon emission time series data of various materials within the accounting boundary using a hierarchical memory network; A hierarchical memory network optimization and carbon emission factor prediction device is used to optimize the parameters of the hierarchical memory network using a cruising algorithm, and to predict the carbon emission factors of different materials in the actual production state within the accounting time range and boundary based on the trained network; The carbon emission calculation device is used to calculate the direct carbon emissions and indirect carbon emissions in the non-ferrous aluminum smelting process according to the predicted carbon emission factors, and finally obtain the total carbon emissions.

7. The non-ferrous aluminum casting carbon emission accounting system based on hierarchical memory network according to claim 6 is characterized in that: The accounting time range and boundary of carbon emissions from non-ferrous aluminum casting are determined according to the process of non-ferrous aluminum casting, and the historical data of different materials under different production conditions within the accounting boundary are processed to obtain the carbon emission time series data of various materials. The process is as follows: Step 1: Determine the smelting and casting process of non-ferrous aluminum, including the process from aluminum concentrate to refined aluminum products and the process from scrap aluminum to refined aluminum products, and identify the carbon emission categories in the aluminum smelting and casting process. The carbon emission categories include direct carbon emissions and indirect carbon emissions. Direct carbon emissions include: fuel combustion, energy use as raw materials and industrial processes, and indirect carbon emissions include: net purchased electricity, heat and production emissions; Step 2: According to the two aluminum smelting processes and carbon emission categories, determine the emission sources that may generate carbon emissions in the actual smelting process, and determine the accounting boundary of carbon emissions. At the same time, collect historical data of different materials under different production conditions, including: time, material name, carbon emission category, temperature, oxygen concentration, material input, and carbon output information; Step 3: Filter the acquired information and retain the required data, such as time, material input, actual carbon output, temperature, and oxygen concentration, to construct a carbon emission time series data set for different materials. Then perform maximum and minimum normalization on the obtained data, scale the data to [0,1], and perform maximum and minimum normalization on the data using the following formula: Among them, x is the actual value of the data to be normalized, x' is the normalized value, and x max 、x min are the maximum and minimum values ​​of the data.

8. The non-ferrous aluminum smelting and casting carbon emission accounting system based on hierarchical memory network according to claim 6 is characterized in that: The hierarchical memory network is used to train and fit the carbon emission time series data of various materials within the accounting boundary. The process is as follows: Step 1: For the number of hidden layers m and the delay memory coefficient δ of each hidden layer i , the number of neurons in each hidden layer The sparsity of the internal connection matrix of each hidden layer ξ i , the spectral radius ρ of the internal connection matrix of each hidden layer i , the Euclidean norm regularization constraint coefficient ε is initialized, where the number of hidden layers should not be too large, otherwise it will greatly increase the complexity of the calculation, and the delay memory coefficient should select an appropriate value according to the time span of the actual required long-term memory information; Step 2: Initialize the input weight matrix and internal connection matrix of each hidden layer, and update the internal connection matrix according to the internal connection matrix sparsity, the maximum eigenvalue of the internal connection matrix and the spectral radius. The specific steps are as follows: Step 2-1: Input weight matrix for each hidden layer Perform random assignment, where i = 1, 2, 3, ..., m. When i = 1, enter the weight matrix Each weight value in is within [-0.5, 0.5]. When i≠1, enter the weight matrix The weight values ​​in are within [-0.5, 0.5], where is the number of neurons in hidden layer i, D is the dimension of input layer parameters; Step 2-2: Internal connection matrix of each hidden layer Random assignment, where i = 1, 2, 3, ..., m, internal connection matrix The weight values ​​in are within [-0.5, 0.5], where is the number of neurons in the hidden layer i. The number of non-zero elements in the internal connection matrix is ​​calculated according to the sparsity of the internal connection matrix of each hidden layer. The calculation formula for the number of non-zero elements in the internal connection matrix of each hidden layer is as follows: Among them, n i is the number of non-zero elements in the internal connection matrix of hidden layer i, ξ i is the internal connection matrix sparsity of hidden layer i, is the number of neurons in hidden layer i; Step 2-3: Calculate the maximum eigenvalue and spectral radius of the internal connection matrix of each hidden layer, and update the internal connection matrix of each hidden layer according to the maximum eigenvalue and spectral radius of the internal connection matrix. The updating formula of the internal connection matrix of each hidden layer is as follows: in, is the internal connection matrix of hidden layer i, ρ i is the spectral radius of the internal connection matrix of hidden layer i, and W is Random matrices of the same dimension, where the values ​​of each element in the matrix are within [-0.5, 0.5], λ max (W) is the maximum eigenvalue of the internal connection matrix of the hidden layer; Step 3: Establish a collection matrix, add the normalized carbon emission time series data to the collection matrix, calculate the time series data of each hidden layer neuron and add it to the collection matrix, and then obtain the final collection matrix. The specific process is as follows: Step 3-1: Create an empty collection matrix The normalized carbon emission time series data u∈R D×T Add to the collection matrix, where D is the input layer parameter dimension and L is the output layer parameter dimension. is the number of neurons in hidden layer i, T is the length of the time series; Step 3-2: Calculate the time series data of the first layer of neurons in the hidden layer. The calculation formula for the time series data of the first layer of neurons in the hidden layer is as follows: in, is the time series data of the first layer of neurons in the hidden layer, and f(·) is a nonlinear activation function, The internal connection matrix of the first hidden layer, is the data of the first layer of neurons in the hidden layer at time t-1, is the input weight matrix of the first hidden layer, and u(t) is the input carbon emission time series data; Step 3-3: Calculate the time series data of the i-th (i=2, 3, 4, ..., m) layer of neurons in the hidden layer. The calculation formula for the time series data of the i-th layer of neurons in the hidden layer is as follows: Among them, δ j is the delayed memory coefficient of hidden layer j, θ i is the total delayed memory time of hidden layer i, is the input of hidden layer i at time t, is t-θ i The values ​​of each neuron in the hidden layer i-1 at the moment, is the time series data of the i-th layer neurons in the hidden layer, and f(·) is a nonlinear activation function, is the internal connection matrix of the i-th layer in the hidden layer, is the value of each neuron in hidden layer i at time ti, is the input weight matrix of hidden layer i; Step 4: Calculate the output weight matrix using the collection matrix, actual output, and Euclidean norm regularization constraint coefficient. The output weight matrix calculation formula is as follows: in, is the output weight matrix, Y is the actual output value, X is the collection matrix, and ε is the Euclidean norm regularization constraint coefficient, which is usually a very small positive number.

9. The non-ferrous aluminum smelting and casting carbon emission accounting system based on hierarchical memory network according to claim 6 is characterized in that: The cruising algorithm is used to optimize the parameters of the hierarchical memory network, and the optimized neural network is used to predict the carbon emission factors of different materials in the actual production state within the accounting time range and boundary. The process is as follows: Step 1: Initialize the position and fitness of the fish school, and initialize the initial values ​​of the parameters M, N, EN, PN, SN, and DN, where M is the maximum number of iterations, N is the size of the fish school, EN is the number of fish exploring and foraging in the fish school, PN is the number of fish sensing danger, SN is the safety value, and DN is the warning value (generated by a random number). The output weight matrix in the hierarchical memory network is used as a variable in the search space, and then a loop iteration is performed, and the fish school is sorted to obtain the individual position of the fish currently in the optimal position, as well as the optimal fitness value; Step 2: Update the position of the exploratory fish according to the action characteristics of the exploratory fish. The position of the exploratory fish is updated by the following formula: Among them, t represents the current iteration number, L i,j represents the position information of the i-th fish in the j-th dimension, arctan is the inverse tangent function, c is a random number between [0,1], M is the maximum number of iterations, Q is a random number that follows a Gaussian distribution, R is a unit row vector, DN is the warning value, and SN is the safety value; Step 3: Update the position of the foraging fish according to the position of the exploring foraging fish. Use the following formula to update the position of the foraging fish, that is, Among them, t represents the current iteration number, L i,j represents the position information of the i-th fish in the j-th dimension, arctan is the inverse tangent function, L p is a random position of the fish near the current optimal position, L worst is the worst position occupied by the fish, N is the size of the fish school, Q is a random number that follows a Gaussian distribution, R is a unit row vector, and A + is a row vector containing only random 1 and -1; Step 4: Update the position of the fish group according to the reaction law of the fish group to the natural enemy. The position of the fish group is updated by the following formula: Among them, t represents the current iteration number, L i,j represents the position information of the i-th fish in the j-th dimension, α is the step size control parameter, K is a random parameter of [-1,1], and L best , L worst are the best and worst positions for fish, f i represents the fitness value of the current individual, f g represents the current maximum fitness value, f w Indicates the current minimum fitness value, γ is the smallest constant to avoid the denominator being 0; Step 5: Update the historical optimal fitness. When the maximum number of iterations is reached, output the optimal individual position and fitness value. Use the global optimal solution as the output weight matrix of the hierarchical memory network to obtain the optimized hierarchical memory network. Input the carbon emission time series data of different materials within the accounting boundary into the optimized network to obtain the carbon emission factors of different materials in the actual production status within the accounting time range and boundary.

10. The non-ferrous aluminum smelting and casting carbon emission accounting system based on hierarchical memory network according to any one of claims 6 to 9, characterized in that: The direct and indirect carbon emissions in the non-ferrous aluminum smelting process are calculated based on the predicted carbon emission factors, and the total carbon emissions are finally obtained. The process is as follows: Step 1: Calculate the carbon emissions generated by fuel combustion during aluminum smelting and casting based on the predicted carbon emission factor. The carbon emissions generated by fuel combustion are calculated using the following formula: Among them, C 燃烧 To consume the carbon emissions from burning fossil fuels, is the activity level of the ith fossil fuel, is the carbon emission factor of the i-th fossil fuel; where the activity level of the i-th fossil fuel It is calculated by the following formula: Among them, NCV i is the average lower calorific value of the i-th fossil fuel, FC i is the net consumption of the ith fossil fuel; Step 2: Calculate the carbon emissions generated by the use of energy as raw materials in the aluminum smelting process based on the predicted carbon emission factor. The carbon emissions generated by the use of energy as raw materials are calculated using the following formula: Among them, C 原材料 Carbon emissions from the use of energy as raw materials, is the activity level, i.e. the consumption of energy as a reducing agent, is the carbon emission factor when the jth energy is used as raw material; Step 3: Calculate the carbon emissions generated by the industrial production process of aluminum smelting and casting based on the predicted carbon emission factor. The carbon emissions generated by the industrial production process are calculated using the following formula: C 过程 =C 熔剂 +C 原料 Among them, the carbon emissions generated by flux consumption are calculated as follows: In the formula, C 熔剂 is the carbon emission generated by flux consumption, P k is the net consumption of the kth flux during the accounting period, Carbon emission factor of the kth flux; Among them, the carbon emissions generated by the consumption of carbon-containing raw materials are calculated as follows: In the formula, C 原料 is the carbon emissions from the consumption of carbon-containing raw materials, M l To calculate the purchase amount of the first carbon-containing raw material within the time range and boundary, is the carbon emission factor of the lth purchased raw material, where l is the type of carbon-containing raw material; Step 4: Calculate the carbon emissions generated by the net purchased electricity and heat used in the aluminum smelting process based on the predicted carbon emission factor. The carbon emissions generated by the net purchased electricity and heat are calculated using the following formula: C 电力、热力 =AD 电力 ×EF 电力 +AD 热力 ×EF 热力 Among them, C 电力、热力 To net purchase carbon emissions from electricity and heat, AD 电力 To calculate the net purchased electricity within the time range and boundary, AD 热力 The net purchase of heat within the accounting time range and boundary, EF 电力 is the carbon emission factor of electricity, EF 热力 is the carbon emission factor for heat; Step 5: Sum the carbon emissions from fuel combustion, carbon emissions from energy used as raw materials, carbon emissions from industrial production processes, and carbon emissions from net purchased electricity and heat within the accounting time range to obtain the sum of direct and indirect carbon emissions, that is, the total carbon emissions.

Images