CN-119047316-B - Coal-fired received base element carbon content regression system and operation method

CN119047316BCN 119047316 BCN119047316 BCN 119047316BCN-119047316-B

Abstract

The invention discloses a regression calculation method of the carbon content of a received base element of fire coal, which belongs to the technical field of measurement of the carbon content of the received base element of fire coal and comprises the steps of obtaining raw data of coal quality after analysis and determination, carrying out maximum and minimum normalization treatment to obtain normalized raw data, training and authenticating a regression calculation model of the carbon content of the received base element of fire coal by using the normalized raw data, and processing the raw data of the actually measured coal quality by using the trained and authenticated model to calculate the carbon content of the received base element of fire coal. According to the invention, a regression calculation model of the carbon content of the received base element of the MISM coal is designed, important characteristics of coal quality data can be captured, a relation between the coal quality data and the carbon content of the received base element of the coal is established, the carbon content of the received base element of the coal can be obtained through regression calculation only by actually measuring part of original data of the coal quality, the detection cost of the carbon content of the received base element of the coal is reduced, and the accuracy of a calculation result is ensured.

Inventors

WANG YINCHU
Xiong xingchuang
ZHAO ZHENGYI
CHEN FENG
LIU ZILONG

Assignees

中国计量科学研究院

Dates

Publication Date: 20260512
Application Date: 20240822

Claims (8)

1. The regression calculation method for the carbon content of the received base element of the coal is characterized by comprising the following steps: S1, acquiring coal quality raw data after analysis and determination, performing maximum and minimum normalization processing to obtain normalized raw data, and dividing the normalized raw data into training data and verification data; S2, training data are used as input data, and a MISM coal received base element carbon content regression calculation model is utilized to process the training data, so that a trained MISM coal received base element carbon content regression calculation model is obtained; S3, taking the verification data as input data, and processing by using a trained MISM fire coal received base element carbon content regression calculation model to obtain an authenticated MISM fire coal received base element carbon content regression calculation model; s4, obtaining actual measurement coal quality original data, and performing pretreatment operation to obtain pretreated actual measurement coal quality original data; S5, taking the preprocessed actually measured coal quality raw data as input data, and processing by using an authenticated MISM coal-fired received base element carbon content regression calculation model to obtain the coal-fired received base element carbon content; the method utilizes a MISM fire coal received base element carbon content regression calculation model to process, and comprises the following specific steps: A1, taking input data as input features of a Mamba mapping layer, and performing feature mapping by using the Mamba mapping layer to obtain first mapping features; a2, adding the first mapping characteristic and the input data to obtain a first added characteristic; a3, according to the first addition characteristic, utilizing the linear activation layer to perform integration activation to obtain an integration characteristic; A4, taking the integrated features as input features of a Mamba mapping layer, and performing feature mapping by using the Mamba mapping layer to obtain second mapping features; a5, adding the second mapping feature and the first adding feature to obtain a second adding feature; A6, according to the second addition characteristic, utilizing the linear activation layer to perform integrated activation to obtain the carbon content of the base element received by the coal; the mathematical expression of the MISM fire coal received base element carbon content regression calculation model is as follows: Wherein, the For the coal to receive the carbon content of the base element, For the second addition feature, For the second mapping feature, In order to integrate the features of the device, For the first addition feature, For the first mapping feature, In order to input the data it is possible, As a function of the linear activation layer, For the map layer function Mamba, Is that The input of the function is provided by, Is a constant of the setting.
2. The regression calculation method of the carbon content of the received base element of the fire coal according to claim 1, wherein the analyzed and determined raw data of the coal quality comprises total moisture content, air-base moisture, air-base total sulfur, air-base ash, air-base volatile, air-base hydrogen content, air-base fixed carbon, barrel heating value and the carbon content of the received base element of the fire coal; The normalized raw data comprises normalized total moisture content, normalized air-dry basis moisture, normalized air-dry basis total sulfur, normalized air-dry basis ash, normalized air-dry basis volatile, normalized air-dry basis hydrogen content, normalized air-dry basis fixed carbon, normalized cartridge heating value and normalized coal-fired received basis element carbon content; The pretreated actual measurement coal quality raw data comprises pretreated total moisture content, pretreated air-dry basis moisture, pretreated air-dry basis ash, pretreated air-dry basis volatile matters, pretreated air-dry basis fixed carbon and pretreated cartridge calorific value.
3. The regression calculation method of the carbon content of the received base element of the fire coal according to claim 1, wherein the Mamba mapping layer performs feature mapping, and the specific implementation manner is as follows: B1, acquiring input features of Mamba mapping layers; b2, normalizing by utilizing root mean square of the features according to the input features to obtain normalized input features; B3, processing by using a linear layer according to the normalized input characteristics to obtain a third mapping characteristic; b4, processing by using the 1D convolution layer according to the third mapping characteristic to obtain a convolution characteristic; B5, processing by using an activation layer according to the convolution characteristics to obtain first activation characteristics; B6, processing by using a linear layer according to the first activation characteristic to obtain a fourth mapping characteristic; B7, using the fourth mapping feature as an input feature of the ISSM mapping layer, and mapping by using the ISSM mapping layer to obtain a fifth mapping feature; b8, processing by using the activation layer according to the fifth mapping layer to obtain a second activation feature; b9, processing by using a linear layer according to the input characteristics to obtain a sixth mapping characteristic; b10, according to the sixth mapping characteristic, processing by using the activation layer to obtain a third activation characteristic; B11, multiplying the third activated feature and the second activated feature to obtain a new feature; and B12, processing by using a linear layer according to the new characteristics to obtain the mapping characteristics output by the Mamba mapping layer.
4. The method for regression calculation of elemental carbon content received from a coal as claimed in claim 3 wherein the mathematical expression of Mamba map is as follows: Wherein, the For the map feature of Mamba map layer output, As a new feature of the device, it is possible, For the third activation feature to be a feature of the second activation, For the sixth mapping feature, In order for the second activation feature to be a second activation feature, For the fifth mapping layer the mapping layer is used, For the fourth mapping layer of the layer to be mapped, In order for the first activation feature to be a first, In order to be a convolution feature, For the third mapping feature, For the input features to be normalized, The linear element activation function is gated for Sigmoid, As a function of the linear layer(s), For the map layer function ISSM, As a function of the 1D convolution layer, For Mamba the input features of the mapping layer, For Mamba mapping the scaled result of the layer input feature, As a matrix of trainable parameters of dimension 8, For the root mean square of the normalized features, For the set constant value, the control unit, As a result of the measured coal quality parameters, The input of the function is activated for Sigmoid gating linear cells.
5. The regression calculation method of the carbon content of the received base element of the fire coal according to claim 3, wherein the ISSM mapping layer is mapped according to the following specific implementation manner: C1, acquiring input characteristics of a ISSM mapping layer and a last state matrix of the ISSM mapping layer, and respectively calculating to obtain a system matrix of the ISSM mapping layer, a control matrix of the ISSM mapping layer, an output matrix of the ISSM mapping layer and a direct transfer matrix of the ISSM mapping layer; C2, multiplying the input features by a control matrix of the ISSM mapping layer to obtain control features; c3, multiplying the last state matrix of the ISSM mapping layer by the system matrix of the ISSM mapping layer to obtain state characteristics; C4, adding the state characteristics and the control characteristics to obtain a current state matrix of ISSM mapping layers; c5, multiplying the output matrix of the ISSM mapping layer by the current state matrix of the ISSM mapping layer to obtain a first feature; C6, multiplying the direct transfer matrix of ISSM mapping layers with the input matrix to obtain a second characteristic; And C7, adding the first feature and the second feature to obtain a fifth mapping feature output by the ISSM mapping layer.
6. The method for regression calculation of carbon content of received base element of fire coal according to claim 5, wherein the calculation modes of the system matrix of ISSM mapping layer, the control matrix of ISSM mapping layer, the output matrix of ISSM mapping layer and the direct transfer matrix of ISSM mapping layer in C1 are as follows: D1, acquiring input features of ISSM mapping layers; D2, setting a matrix with dimension of MxN, and initializing by using L2 norm normalization to obtain an initialized matrix, wherein M is the number of rows of the matrix, and N is the number of columns of the matrix; D3, processing by using four different linear layers according to the input characteristics of the ISSM mapping layers to respectively obtain seventh mapping characteristics, eighth mapping characteristics, ninth mapping characteristics and output matrixes of the ISSM mapping layers; D4, according to the seventh mapping feature, processing by using the activation layer to obtain a fourth activation feature; D5, multiplying the fourth activation feature by the eighth mapping feature to obtain a control matrix of ISSM mapping layers; D6, multiplying the fourth activation feature by the initialized matrix to obtain a system feature; Carrying out exponential transformation on the system characteristics to obtain a system matrix of ISSM mapping layers; And D8, according to the ninth mapping characteristic, activating by using the activation layer to obtain a direct transfer matrix of the ISSM mapping layer.
7. The regression calculation method of the carbon content of the received base element of the fire coal according to claim 6, wherein the mathematical expressions of the system matrix of the ISSM mapping layer, the control matrix of the ISSM mapping layer, the output matrix of the ISSM mapping layer and the direct transfer matrix of the ISSM mapping layer are as follows: Wherein, the For a ISSM system matrix of mapping layers, For the control matrix of ISSM mapping layers, For the ISSM output matrix of the mapping layer, For the ISSM direct transfer matrix of the mapping layer, For ISSM the input features of the mapping layer, As a feature of the system, For the fourth activation feature, For the ninth mapping feature, For the eighth mapping feature, For the seventh mapping feature, 、、 And For four different linear layers, The linear element activation function is gated for Sigmoid, For the matrix to be initialized, As a matrix with dimensions of 16 x N, Normalized to the L2 norm.
8. The method for regression calculation of elemental carbon content received from coal according to claim 1 wherein the specific steps of S3 are as follows: S301, taking verification data as input data, and calculating through a trained MISM fire coal received base element carbon content regression calculation model to obtain predicted normalized fire coal received base element carbon content; s302, performing inverse normalization operation on the predicted normalized carbon content of the coal-received base element to obtain the predicted carbon content of the coal-received base element; s303, calculating the average relative error between the predicted carbon content of the received base element of the fire coal and the carbon content of the received base element of the fire coal in the raw data of the coal quality; And S304, judging whether the average relative error is smaller than or equal to a set threshold value, if so, obtaining an authenticated MISM fire coal received base element carbon content regression calculation model, otherwise, optimizing the MISM fire coal received base element carbon content regression calculation model, and returning to S2.

Description

Coal-fired received base element carbon content regression system and operation method Technical Field The invention belongs to the technical field of measurement of carbon content of a coal-fired received base element, and particularly relates to a regression calculation method of carbon content of a coal-fired received base element. Background The carbon accounting method is used as an important method for carbon emission calculation, and provides a guarantee for fairness and fairness of transactions in the carbon market in the thermal power industry. The carbon accounting method requires enterprises to collect, count and calculate key data strictly according to relevant standards and methods so as to ensure the accuracy of carbon emission results. According to the accounting requirements in the enterprise greenhouse gas emission accounting and report guide power generation facility, in the coal quality analysis of thermal power enterprises, the carbon content of the base element received by the fire coal is a key actual measurement parameter in carbon accounting. On one hand, the carbon content of the base element received by the coal directly influences the accuracy of the carbon accounting of an enterprise, and the method is a guarantee for the enterprise to bear reasonable performance cost. On the other hand, the carbon content of the element of the received base of the coal is calculated by other coal analysis parameters, and auxiliary parameters are needed for data reliability verification, and besides the measured element carbon content, the parameters to be measured comprise 8 parameters including total moisture, air-dry base total sulfur, air-dry base ash, air-dry base volatile matters, air-dry base hydrogen content, air-dry base fixed carbon and heat productivity of the cartridge. The measurement of these parameters requires a plurality of corresponding instruments, the measurement flow is complex and consumes a lot of manpower, and enterprises bear a lot of detection cost. If the enterprises do not actually measure the elemental carbon, the carbon emission calculation is needed to be performed by adopting a default value, and the calculation result is large, so that the implementation cost born by the enterprises is unreasonable. In order to ensure that enterprises bear reasonable performance cost and reduce detection cost of enterprises, it is required to consider how to accurately calculate the carbon content of the base element received by the coal under the condition of reducing measured parameters. On the basis of the existing measurement data, 8 parameters including total moisture, air-based total sulfur, air-based ash, air-based volatile matter, air-based hydrogen content, air-based fixed carbon and heat productivity of a bomb are taken as input, a calculation model is constructed to carry out regression on the carbon content of the base element received by the fire coal, and when actual calculation is carried out, only partial parameters such as 6 parameters including total moisture, air-based ash, air-based volatile matter, air-based fixed carbon and heat productivity of the bomb are measured, and the parameters which are not measured take default values such as the total sulfur of the air-based and the air-based hydrogen content take default values, so that detection parameters and detection instruments can be reduced, and the detection cost of enterprises can be reduced. Therefore, how to construct a regression model of the carbon content of the received base element of the coal based on part of the input parameters is a key problem to be solved. Disclosure of Invention Aiming at the defects in the prior art, the regression calculation method for the carbon content of the coal-fired received base element builds a regression model for the carbon content of the coal-fired received base element and solves the problems of high calculation cost and low calculation efficiency of enterprise carbon accounting. In order to achieve the aim of the invention, the technical scheme adopted by the invention is that the regression calculation method for the carbon content of the base element received by the fire coal comprises the following steps: S1, acquiring coal quality raw data after analysis and determination, performing maximum and minimum normalization processing to obtain normalized raw data, and dividing the normalized raw data into training data and verification data; S2, training data are used as input data, and a MISM coal received base element carbon content regression calculation model is utilized to process the training data, so that a trained MISM coal received base element carbon content regression calculation model is obtained; S3, taking the verification data as input data, and processing by using a trained MISM fire coal received base element carbon content regression calculation model to obtain an authenticated MISM fire coal received base element carbon content regression calc