CN111257306B - Online dynamic prediction method and system for alkali metal element content of biomass fuel - Google Patents
Online dynamic prediction method and system for alkali metal element content of biomass fuel Download PDFInfo
- Publication number
- CN111257306B CN111257306B CN202010046096.0A CN202010046096A CN111257306B CN 111257306 B CN111257306 B CN 111257306B CN 202010046096 A CN202010046096 A CN 202010046096A CN 111257306 B CN111257306 B CN 111257306B
- Authority
- CN
- China
- Prior art keywords
- flame
- alkali metal
- metal element
- biomass fuel
- signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
- G01N21/71—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light thermally excited
- G01N21/72—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light thermally excited using flame burners
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Regulation And Control Of Combustion (AREA)
Abstract
The invention discloses an online dynamic prediction method and system for alkali metal element content of biomass fuel. The method comprises the steps of carrying out data preprocessing on the original flame radiation signal to generate a flame full spectrum signal and an alkali metal element spectrum signal after preprocessing; and then, determining flame combustion characteristic parameters and flame temperature according to the full spectrum signals of the flame, extracting the spectrum characteristic values of alkali metal elements according to the spectrum signals of the alkali metal elements, jointly forming a characteristic parameter matrix, training the established recurrent neural network model, and generating a trained biomass fuel alkali metal element content online prediction model. The online prediction model for the alkali metal element content of the biomass fuel can be used for dynamically predicting the alkali metal element content of the biomass fuel combusted in the biomass boiler on line, so that the real-time performance and the accuracy of the prediction of the alkali metal element content of the biomass fuel are improved.
Description
Technical Field
The invention relates to the technical field of spectral analysis and detection, in particular to an online dynamic prediction method and system for alkali metal element content of biomass fuel.
Background
In the combustion of a boiler adopting biomass fuel, if the combustion is insufficient or the temperature of a hearth is too high, the slag bonding phenomenon of the biomass boiler is very serious. The content of the low-melting-point alkali metal compound in the biomass fuel is larger than that in the coal, so that the ash melting point is lower. If the temperature of the hearth is relatively high, slag can be formed on the water-cooled wall when biomass ash in a molten or semi-molten state is attached to the water-cooled wall without cooling. And the molten alkali metal compound can react with the inner wall of the hearth, so that the hearth is corroded and potential harm is caused.
Regarding biomass fuel slagging analysis, the prior art is essentially off-line measurement analysis. The content composition and the proportion of elements in the biomass fuel are detected by a chemical method or a spectrum, and the slagging condition is presumed by the content and the composition proportion of alkali metal compounds; or directly measuring the ash melting point of the biomass fuel in a laboratory, and further estimating the slagging condition of the biomass fuel. When the fuel is changed or the mixture ratio is not uniform, the off-line prediction method has deviation.
Disclosure of Invention
The invention aims to provide an online dynamic prediction method and system for the content of alkali metal elements in biomass fuel, and aims to solve the problem of large deviation of prediction results of the existing offline prediction method for the content of alkali metal elements in biomass fuel.
In order to achieve the purpose, the invention provides the following scheme:
a method for online dynamic prediction of alkali metal element content of biomass fuel, comprising:
acquiring an original flame radiation signal when the biomass fuel is combusted;
carrying out data preprocessing on the original flame radiation signal to generate a preprocessed flame radiation signal; the pretreated flame radiation signals comprise pretreated flame full spectrum signals and alkali metal element spectrum signals;
determining flame combustion characteristic parameters and flame temperature according to the flame full-spectrum signal;
extracting a spectral characteristic value of the alkali metal element according to the alkali metal element spectral signal;
generating a characteristic parameter matrix according to the flame combustion characteristic parameters, the flame temperature and the spectral characteristic values;
adopting the cyclic neural network model established by the characteristic parameter matrix training to generate a trained biomass fuel alkali metal element content online prediction model;
and predicting the alkali metal element content of the biomass fuel combusted in the biomass boiler by adopting the biomass fuel alkali metal element content online prediction model.
Optionally, the performing data preprocessing on the original flame radiation signal to generate a preprocessed flame radiation signal specifically includes:
and removing outliers in the original flame radiation signal by adopting a k-means clustering algorithm, performing smooth denoising treatment, simultaneously avoiding the influence of system errors on the original flame radiation signal, and generating a preprocessed flame radiation signal.
Optionally, determining the flame combustion characteristic parameter and the flame temperature according to the flame full spectrum signal specifically includes:
extracting flame combustion characteristic parameters according to the flame full spectrum signal; the flame combustion characteristic parameters comprise flame flicker frequency and radiation energy;
and measuring the flame temperature at the same moment as the flame full-spectrum signal based on a bicolor method.
Optionally, the extracting a spectral feature value of the alkali metal element according to the alkali metal element spectral signal specifically includes:
eliminating the influence of black body radiation in the alkali metal element spectral signal on the spectral intensity to generate an alkali metal element spectral signal with the influence eliminated;
extracting the spectral characteristic value of the alkali metal element spectral signal after the influence is eliminated; the spectral characteristic values comprise characteristics of a mean value, a standard deviation, a kurtosis coefficient and a skewness coefficient.
Optionally, the generating a trained biomass fuel alkali metal element content online prediction model by using the recurrent neural network model established by the characteristic parameter matrix training specifically includes:
acquiring an established recurrent neural network model; the recurrent neural network model comprises an input layer, a recurrent layer, a full connection layer and an output layer;
and training the cyclic neural network model by adopting the characteristic parameter matrix, and optimizing the model parameters of the cyclic neural network model based on a batch gradient descent method to generate a trained biomass fuel alkali metal element content online prediction model.
An online dynamic prediction system for alkali metal element content of biomass fuel, the system comprising:
the flame radiation signal acquisition module is used for acquiring an original flame radiation signal when the biomass fuel is combusted;
the data preprocessing module is used for preprocessing the original flame radiation signal to generate a preprocessed flame radiation signal; the pretreated flame radiation signals comprise pretreated flame full spectrum signals and alkali metal element spectrum signals;
the flame combustion characteristic parameter and flame temperature extraction module is used for determining the flame combustion characteristic parameter and the flame temperature according to the full spectrum signal of the flame;
the spectral feature extraction module is used for extracting a spectral feature value of the alkali metal element according to the alkali metal element spectral signal;
the characteristic parameter matrix generating module is used for generating a characteristic parameter matrix according to the flame combustion characteristic parameters, the flame temperature and the spectral characteristic values;
the model training module is used for training the established recurrent neural network model by adopting the characteristic parameter matrix to generate a trained biomass fuel alkali metal element content online prediction model;
and the online prediction module of the alkali metal element content is used for predicting the alkali metal element content of the biomass fuel combusted in the biomass boiler by adopting the online prediction model of the alkali metal element content of the biomass fuel.
Optionally, the data preprocessing module specifically includes:
and the data preprocessing unit is used for removing outliers in the original flame radiation signal by adopting a k-means clustering algorithm, performing smooth denoising treatment, simultaneously avoiding the influence of system errors on the original flame radiation signal, and generating a preprocessed flame radiation signal.
Optionally, the flame combustion characteristic parameter and flame temperature extraction module specifically includes:
the flame combustion characteristic parameter extraction unit is used for extracting flame combustion characteristic parameters according to the flame full spectrum signal; the flame combustion characteristic parameters comprise flame flicker frequency and radiation energy;
and the flame temperature measuring unit is used for measuring the flame temperature at the same moment as the full spectrum signal of the flame based on a bicolor method.
Optionally, the spectral feature extraction module specifically includes:
the black body radiation eliminating unit is used for eliminating the influence of black body radiation in the alkali metal element spectral signal on the spectral intensity and generating an alkali metal element spectral signal after the influence is eliminated;
the spectral feature extraction unit is used for extracting the spectral feature value of the alkali metal element spectral signal after the influence is eliminated; the spectral characteristic values comprise characteristics of a mean value, a standard deviation, a kurtosis coefficient and a skewness coefficient.
Optionally, the model training module specifically includes:
the model establishing unit is used for acquiring the established recurrent neural network model; the recurrent neural network model comprises an input layer, a recurrent layer, a full connection layer and an output layer;
and the model training optimization unit is used for training the cyclic neural network model by adopting the characteristic parameter matrix, optimizing the model parameters of the cyclic neural network model based on a batch gradient descent method and generating a trained biomass fuel alkali metal element content online prediction model.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides an online dynamic prediction method and system for alkali metal element content of biomass fuel, wherein the method comprises the steps of firstly, obtaining an original flame radiation signal when the biomass fuel is combusted; carrying out data preprocessing on the original flame radiation signal to generate a flame full spectrum signal and an alkali metal element spectrum signal after preprocessing; determining flame combustion characteristic parameters and flame temperature according to the flame full-spectrum signal; extracting a spectral characteristic value of the alkali metal element according to the alkali metal element spectral signal; generating a characteristic parameter matrix according to the flame combustion characteristic parameters, the flame temperature and the spectral characteristic values; and adopting the cyclic neural network model established by the characteristic parameter matrix training to generate a trained biomass fuel alkali metal element content online prediction model. The online prediction model for the alkali metal element content of the biomass fuel can be used for dynamically predicting the alkali metal element content of the biomass fuel combusted in the biomass boiler on line, so that the real-time performance and the accuracy of the prediction of the alkali metal element content of the biomass fuel are improved. The biomass fuel obtained by the online dynamic prediction of the method of the invention has the content of alkali metal elements to guide the combustion, and has important significance for preventing boiler slagging and high-efficiency safe operation.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flow chart of a method for online dynamic prediction of alkali metal content in biomass fuel provided by the invention;
FIG. 2 is a process schematic diagram of an online dynamic prediction method for alkali metal element content of biomass fuel provided by the invention;
FIG. 3 is a graph of the single-acquired spectrum signal spectrum result of the alkali metal element provided by the present invention;
fig. 4 is a schematic diagram of a network structure of a recurrent neural network model provided by the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide an online dynamic prediction method and system for the content of alkali metal elements in biomass fuel, and aims to solve the problem of large deviation of prediction results of the existing offline prediction method for the content of alkali metal elements in biomass fuel.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Fig. 1 is a flow chart of a method for online dynamic prediction of alkali metal element content of a biomass fuel provided by the present invention, and fig. 2 is a process schematic diagram of the method for online dynamic prediction of alkali metal element content of a biomass fuel provided by the present invention. Referring to fig. 1 and fig. 2, in the online dynamic prediction method for alkali metal element content of biomass fuel provided by the invention, a radiation signal of a biomass fuel combustion flame is obtained through an optical fiber spectrum sensor, a flame combustion characteristic parameter is obtained through data analysis and pretreatment, a spectral characteristic value of an alkali metal element is extracted, and a flame temperature at the same moment is measured by adopting a bicolor method, so that an online dynamic prediction model for alkali metal element content of biomass fuel based on a Recurrent Neural Network (RNN) is established, and thus, the online dynamic prediction of the alkali metal element content of biomass fuel is carried out in real time according to the model.
Therefore, the method specifically comprises the following steps:
step 101: and acquiring an original flame radiation signal when the biomass fuel is combusted.
Under the same combustion working condition, namely different biomass fuels are combusted under the same combustion condition, the flame radiation signal of each biomass fuel is obtained by the optical fiber spectrum sensor based on the flame data acquisition system and serves as an original flame radiation signal, and the flame radiation signal comprises a flame full spectrum signal and an alkali metal element spectrum signal. The alkali metal is K (potassium) and Na (sodium). The melting point of K and Na compounds is low, the content of the K and Na compounds in biomass is high, a large amount of low-melting-point compounds are easily formed to reduce the ash melting point, and slagging is generated.
Step 102: and carrying out data preprocessing on the original flame radiation signal to generate a preprocessed flame radiation signal.
The invention carries out data preprocessing on the acquired spectral data, removes outliers and simultaneously avoids the influence of system errors on results. The data preprocessing process specifically comprises the following steps:
and removing outliers in the original flame radiation signal by adopting a k-means clustering algorithm, performing smooth denoising treatment, simultaneously avoiding the influence of system errors on the original flame radiation signal, and generating a preprocessed flame radiation signal. The pretreated flame radiation signals comprise pretreated flame full spectrum signals and alkali metal element spectrum signals.
Wherein the effect of systematic errors on the measurement results comprises the effect of dark noise and background spectra on the measurement results. The invention not only ensures that the spectrometer of the flame data acquisition system is under the normal use environment, but also carries out multiple measurements on the flame spectrum (flame radiation signal) in the experiment and takes the statistical average value to eliminate the dark noise. Meanwhile, the method also carries out multiple spectrum collection aiming at the background without flame, and takes the average value as the background spectrum, so the flame spectrum used by the method is obtained by subtracting the background spectrum, thereby avoiding the influence of the background spectrum on the measurement result.
Step 103: and determining flame combustion characteristic parameters and flame temperature according to the flame full spectrum signal.
Extracting flame combustion characteristic parameters according to the full spectrum signal of the flame, wherein the flame combustion characteristic parameters comprise flame flicker frequency and radiation energy; and measuring the flame temperature at the same time with the full spectrum signal of the flame based on a bicolor method.
The flame full spectrum signal spectrogram acquired by the optical fiber spectrometer (optical fiber spectrum sensor) has the abscissa as wavelength and the ordinate as spectrum intensity. The spectral intensity is a time domain signal, and after discrete fourier transform, a corresponding representation of the spectral intensity in the frequency domain can be obtained:
wherein, x (N) is the spectrum intensity of the nth sampling moment in the flame full spectrum signal spectrogram, and N is the number of the sampling moments; x (k) is a corresponding complex form of x (n).
X (k) is a complex number which can be converted to a corresponding power spectrum according to the formula:
P(k)＝|X(k)|2/N (2)
the radiant energy E of the band is t1To t2The definite integral of the radiation intensity of a specific wave band in a time period has the following calculation formula:
wherein X is the abbreviation for X (k) and E is the radiant energy.
The flame flicker frequency F is calculated as:
wherein, p (i) represents the power spectrum value of the ith sampling moment, f (i) represents the frequency of the ith sampling moment, and N is the number of sampling moments.
The flame flicker frequency F and the radiation energy E reflect the characteristics of the biomass combustion flame to a certain extent; the combustion temperature is directly related to the slagging degree, the higher the temperature is, the more violent the combustion is, and the higher the possibility of slagging is, so that the parameters are taken as characteristic values for biomass slagging prediction.
Step 104: and extracting the spectral characteristic value of the alkali metal element according to the alkali metal element spectral signal.
The invention aims at the alkali metal element spectrum signal, extracts the spectrum characteristic value of the alkali metal element, and eliminates the influence of black body radiation on the spectrum intensity, and concretely comprises the following steps:
and eliminating the influence of black body radiation in the alkali metal element spectral signal on the spectral intensity to generate the alkali metal element spectral signal with the influence eliminated.
FIG. 3 is a single-collected spectrum result chart of alkali metal element spectrum signals provided by the present invention. Blackbody radiation occurs in all bands and is related only to flame temperature. As shown in fig. 3, the wavelength corresponding to K, Na element is the peak part, the bottom is black body radiation, and the peak part is intercepted as the emission spectrum corresponding to K, Na element by subtracting the influence of black body radiation during data processing, that is, the spectrum signal of alkali metal element after influence elimination is obtained.
Extracting the spectral characteristic value of the alkali metal element spectral signal after the influence is eliminated; the spectral characteristic values comprise characteristics of a mean value, a standard deviation, a kurtosis coefficient and a skewness coefficient.
Wherein the mean value calculation formula is:
where μ represents the mean of the spectral intensities at N successive instants, xiRepresenting the spectral intensity at the ith time instant, it is preferred in the present invention that N be taken to be 30.
The standard deviation calculation formula is as follows:
where σ represents the sample standard deviation of the spectral intensity at N consecutive time instances.
The skewness coefficient is a third-order standardized matrix of the sample, is a digital characteristic of the asymmetric degree of statistical data distribution, can represent the asymmetric degree of the probability density function of the spectral radiation intensity of the alkali metal element, and has the following calculation formula:
where S represents the skewness factor of the spectral intensity at N consecutive time instants, and σ represents the standard deviation of the sample.
The kurtosis coefficient is a ratio of a fourth-order central moment of a variable to a square of a variance, is a description of whether a distribution peak value of statistical data is sharp or flat, and can represent the steepness of a probability density function of spectral radiation intensity of an alkali metal element, and a calculation formula is as follows:
wherein G represents the kurtosis coefficient of the spectral intensities at N successive time instants, σ represents the standard deviation of the sample, μ represents the mean of the spectral intensities at the N successive time instants, xiRepresenting the spectral intensity at the ith time instant.
The method extracts the mean value, the standard deviation, the kurtosis coefficient and the skewness coefficient of the alkali metal element spectral data as the spectral characteristic values.
Step 105: and generating a characteristic parameter matrix according to the flame combustion characteristic parameters, the flame temperature and the spectral characteristic values.
The invention adopts 7 eigenvalues of flame flicker frequency, radiation energy, flame temperature and alkali metal element spectral data, mean value, standard deviation, kurtosis coefficient and skewness coefficient to form a characteristic parameter matrix which is used as sample input of model training.
For example, taking K element as an example, every 30 continuous spectrum data are taken as a group, including flame full spectrum data and spectrum data of a waveband where the K element is located, flame flicker frequency, radiation energy and flame temperature are calculated based on the formula and the flame full spectrum data, a mean value, a standard deviation, a kurtosis coefficient and a skewness coefficient are calculated based on the formula and the spectrum data of the waveband where the K element is located, 7 characteristic values are used as sample inputs, and the actual K element content of the biomass fuel is used as sample outputs to jointly form a sample set. In all sample sets, 90% of them were selected as training set and 10% as testing set. Na works in the same way.
Step 106: and adopting the cyclic neural network model established by the characteristic parameter matrix training to generate a trained biomass fuel alkali metal element content online prediction model.
Firstly, an established Recurrent Neural Network (RNN) model is obtained, a training set and a testing set in the guarantee 105 are adopted to train the network, model parameters are optimized based on a batch gradient descent method, and the trained biomass fuel alkali metal element content online prediction model is generated to predict the content of the alkali metal element in the biomass fuel online.
Fig. 4 is a schematic diagram of a network structure of a recurrent neural network model provided by the present invention. Referring to fig. 4, the recurrent neural network model established by the present invention includes an input layer X, a recurrent layer H, a fully-connected layer F, and an output layer Y. Wherein the input layer X normalizes and serializes the input signal, using seven-dimensional data of mean value, standard deviation, kurtosis coefficient, skewness coefficient, flicker frequency, radiation energy and flame temperature; the circulation layer H calculates the serialized data and is characterized in that the output is not only related to the current sequence, but also related to the output of the previous sequence to form a dynamic prediction model; the full-connection layer F transmits data antecedents and errors reversely during training; and finally, calculating and outputting the result obtained by calculation through the Y function of the output layer.
The circulation layer H has 7 neurons in each layer, and has 3 layers. The circulation layer formula is as follows:
H(t)＝σ(wH(H(t-1))+wx(x(t))) (9)
wherein H (t) is the cycle layer output at the time t, H (t-1) is the cycle layer output at the time t-1, w is the weight, sigma is the activation function, and x (t) is the characteristic parameter input at the time t. w is aHThe coefficient matrix of the cycle layer is used for calculating data from the cycle layer at the t-1 moment to the cycle layer at the t moment. w is axThe coefficient matrix of the input layer is used for calculating data from the input layer to the cycle layer at the time t.
And the full connection layer F is used for data antecedent propagation and error back propagation during training. The total connecting layer is 6 layers, the first 4 layers are 14 neurons in each layer, and the second 2 layers are 5 neurons in each layer.
And finally, outputting the prediction result Y (t) by adopting a softmax function through an output layer Y according to the result obtained by calculation.
The input of the recurrent neural network model is a characteristic parameter matrix consisting of 7 characteristic values of flame flicker frequency, radiation energy, flame temperature and K (or Na) element spectral data, standard deviation, kurtosis coefficient and skewness coefficient, and the output of the recurrent neural network model is the content of alkali metal element K (or Na) in different biomass fuels. And inputting the obtained spectral characteristic value sequence signal (characteristic parameter matrix) into an RNN model, training the network, and optimizing model parameters by using a batch gradient descent method to meet the requirement of online dynamic prediction of the alkali metal content of the biomass fuel.
Specifically, the formula of each layer in the recurrent neural network model is as follows:
St＝WHt-1+UXt (10)
Ht＝f(St) (11)
Ft＝g(VHt) (12)
Yt＝j(PFt) (13)
wherein, XtRepresenting the input eigen-parameter matrix, X (t) being the eigen-parameter matrix XtA certain characteristic parameter of. U represents weight matrix of input channel of circulation layer and characteristic parameter matrix X of inputtMultiplying to perform operation; ht-1Represented by the calculation of the loop layer at the time t-1, W is the weight matrix of the loop path of the loop layer, and Ht-1Multiplying to perform operation; jointly obtain an intermediate output result St。HtRepresents the calculation result of the loop layer at the time t, V represents the weight matrix of the output channel of the loop layer, and HtMultiplying; ftRepresenting the output of the loop layer to the fully connected layer. P represents the output channel weight matrix from the full connection layer to the output layer, and FtMultiplication. Y istRepresents the output result of the output layer, as y (t). f. g and j both represent activation functions. The initial time will generally output the result S intermediately0And setting zero, initializing weight parameters randomly, and performing iterative computation. The activation function is generally a function of Tanh (hyperbolic tangent function), Relu (Rectified linear unit), Sigmoid, or the like.
The cyclic neural network model is iteratively trained by adopting the characteristic parameter matrix, and the model parameters of the cyclic neural network model are circularly optimized based on a batch gradient descent method, wherein the ending conditions of model iterative training and cyclic optimization are that a loss function is smaller than a preset value (0.05) or reaches a preset training frequency (2000 cycles). And after the model training is finished, testing the prediction accuracy of the model by using the test set sample, and if the prediction accuracy is higher than a preset accuracy threshold, using the circulating neural network model at the moment as a trained biomass fuel alkali metal element content online prediction model. And if the prediction accuracy of the model is lower than the preset accuracy threshold, returning to the step 101, and training the recurrent neural network model again until the prediction accuracy of the model is higher than the preset accuracy threshold, so as to obtain the trained biomass fuel alkali metal element content online prediction model.
Step 107: and predicting the alkali metal element content of the biomass fuel combusted in the biomass boiler by adopting the biomass fuel alkali metal element content online prediction model.
After the model training is completed, when the model is actually used, an optical fiber spectrum sensor is needed to collect a flame full spectrum signal and an alkali metal element spectrum signal of a combustion flame of the biomass fuel currently combusted in the biomass boiler in real time, then a current characteristic parameter matrix consisting of 7 characteristic values (flame flicker frequency, radiation energy, flame temperature, mean value, standard deviation, kurtosis coefficient and skewness coefficient) of the flame full spectrum signal and the alkali metal element spectrum signal is extracted and input into the trained biomass fuel alkali metal element content online prediction model, and the alkali metal element content in the biomass fuel can be output in real time.
The content of alkali metal elements in the biomass fuel is a main factor of slag formation of the biomass boiler and is related to other factors such as combustion temperature, and the content of the alkali metal elements predicted on line according to the method can assist in predicting the real-time slag formation condition of the biomass boiler. The method aims at the problems that predicted alkali metal elements are K and Na, the content of the alkali metal elements in the biomass fuel is high, and the influence on slagging and corrosion is large.
In the prior art, the content of alkali metal elements of a certain fuel is measured off line, and the slagging tendency of the fuel is judged by combining other factors; the method of the invention online predicts the content of alkali metal elements in the fuel in the hearth at the current moment in real time according to the flame spectrum. The biomass fuel used for combustion of the biomass boiler is often mixed, the content of alkali metal elements is measured on line in real time, the combustion condition is controlled in time, and the slagging tendency is reduced. Because the melting point of the compound formed by the alkali metal element is low and is a main factor for generating slag bonding, the higher the content of the alkali metal element is, the greater the tendency of the fuel or combustion flame in a hearth to generate slag bonding is, and currently, a unified discriminant formula does not exist internationally, and the three conditions are only divided into three conditions of easy slag bonding, moderate slag bonding degree and difficult slag bonding. The method provided by the invention can be used for acquiring the flame combustion condition in real time based on the spectrum information of the biomass combustion flame, dynamically predicting the content of alkali metal elements in the biomass fuel on line, facilitating the analysis of the slagging condition and the corrosion condition of the biomass boiler, and improving the safety and the economical efficiency of boiler combustion.
The method comprises the steps of obtaining a radiation signal of biomass fuel combustion flame through an optical fiber spectrum sensor, obtaining flame combustion characteristic parameters through data analysis and pretreatment, extracting a spectrum characteristic value of an alkali metal element, measuring the flame temperature at the same moment by adopting a two-color method, and establishing a biomass fuel alkali metal element content online dynamic prediction model based on a Recurrent Neural Network (RNN) so as to guide the combustion to be carried out. The method has important significance for preventing boiler slagging and efficient and safe operation, and is suitable for online dynamic prediction of alkali metal element content of the mixed solid fuel.
Based on the online dynamic prediction method for the alkali metal element content of the biomass fuel provided by the invention, the invention also provides an online dynamic prediction system for the alkali metal element content of the biomass fuel, and the system comprises:
the flame radiation signal acquisition module is used for acquiring an original flame radiation signal when the biomass fuel is combusted;
the data preprocessing module is used for preprocessing the original flame radiation signal to generate a preprocessed flame radiation signal; the pretreated flame radiation signals comprise pretreated flame full spectrum signals and alkali metal element spectrum signals;
the flame combustion characteristic parameter and flame temperature extraction module is used for determining the flame combustion characteristic parameter and the flame temperature according to the full spectrum signal of the flame;
the spectral feature extraction module is used for extracting a spectral feature value of the alkali metal element according to the alkali metal element spectral signal;
the characteristic parameter matrix generating module is used for generating a characteristic parameter matrix according to the flame combustion characteristic parameters, the flame temperature and the spectral characteristic values;
the model training module is used for training the established recurrent neural network model by adopting the characteristic parameter matrix to generate a trained biomass fuel alkali metal element content online prediction model;
and the online prediction module of the alkali metal element content is used for predicting the alkali metal element content of the biomass fuel combusted in the biomass boiler by adopting the online prediction model of the alkali metal element content of the biomass fuel.
The data preprocessing module specifically comprises:
and the data preprocessing unit is used for removing outliers in the original flame radiation signal by adopting a k-means clustering algorithm, performing smooth denoising treatment, simultaneously avoiding the influence of system errors on the original flame radiation signal, and generating a preprocessed flame radiation signal.
The flame combustion characteristic parameter and flame temperature extraction module specifically comprises:
the flame combustion characteristic parameter extraction unit is used for extracting flame combustion characteristic parameters according to the flame full spectrum signal; the flame combustion characteristic parameters comprise flame flicker frequency and radiation energy;
and the flame temperature measuring unit is used for measuring the flame temperature at the same moment as the full spectrum signal of the flame based on a bicolor method.
The spectral feature extraction module specifically comprises:
the black body radiation eliminating unit is used for eliminating the influence of black body radiation in the alkali metal element spectral signal on the spectral intensity and generating an alkali metal element spectral signal after the influence is eliminated;
the spectral feature extraction unit is used for extracting the spectral feature value of the alkali metal element spectral signal after the influence is eliminated; the spectral characteristic values comprise characteristics of a mean value, a standard deviation, a kurtosis coefficient and a skewness coefficient.
The model training module specifically comprises:
the model establishing unit is used for acquiring the established recurrent neural network model; the recurrent neural network model comprises an input layer, a recurrent layer, a full connection layer and an output layer;
and the model training optimization unit is used for training the cyclic neural network model by adopting the characteristic parameter matrix, optimizing the model parameters of the cyclic neural network model based on a batch gradient descent method and generating a trained biomass fuel alkali metal element content online prediction model.
The prior art is to use chemical analysis means to measure the alkali metal element content of a certain fuel off line in a laboratory, which has high accuracy but poor timeliness and needs to be re-measured once the fuel is changed or the mixed burning occurs. The method of the invention predicts the alkali metal element content of the fuel in the hearth at the current moment in real time according to the flame spectrum, has strong timeliness compared with the traditional method, has better adaptability to the conditions of biomass boiler fuel change or mixed combustion and the like, is beneficial to adjusting combustion control in time and reducing slagging tendency.
The method and the system of the invention burn different biomass fuels under the same burning condition, obtain the radiation signal of the burning flame of the biomass fuel through the optical fiber spectrum sensor, and obtain the characteristic parameters of the burning flame through data analysis and pretreatment, wherein the characteristic data of the burning flame comprises the flicker frequency and the radiation energy, and the flame temperature at the same moment is measured by a bicolor method; and extracting K, Na element wave band spectral characteristic values, eliminating the influence caused by black body radiation, wherein the extracted characteristic parameters comprise a mean value, a standard deviation, a kurtosis coefficient and a skewness coefficient, and establishing a circulating neural network (RNN) -based biomass fuel K, Na element content online dynamic prediction model. The method and the system are suitable for online dynamic prediction of the content of alkali metal elements in the biofuel, and can be used for guiding the combustion economy and stable operation. When the content of the alkali metal elements of the biomass fuel is predicted to be high, the slagging is easy to generate, the temperature of a hearth can be properly controlled, and the probability of slagging generation caused by too high temperature is avoided. In addition, aiming at the biomass fuel which is easy to generate slag bonding, the biomass fuel which is difficult to generate slag bonding is additionally combusted during combustion, the proportion of the fuel is changed, and the slag bonding can be reduced. If the furnace temperature and fuel ratio cannot be changed too much to maintain combustion efficiency or combustion stability, a slagging inhibitor may be added to the fuel to reduce slagging.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (8)
1. An online dynamic prediction method for alkali metal element content of biomass fuel is characterized by comprising the following steps:
acquiring an original flame radiation signal when the biomass fuel is combusted;
carrying out data preprocessing on the original flame radiation signal to generate a preprocessed flame radiation signal; the pretreated flame radiation signals comprise pretreated flame full spectrum signals and alkali metal element spectrum signals;
the data preprocessing is performed on the original flame radiation signal to generate a preprocessed flame radiation signal, and the method specifically includes:
removing outliers in the original flame radiation signal by adopting a k-means clustering algorithm, performing smooth denoising treatment, simultaneously avoiding the influence of system errors on the original flame radiation signal, and generating a preprocessed flame radiation signal;
the effect of the systematic error on the raw flame radiation signal includes the effect of dark noise and background spectra on the raw flame radiation signal; the dark noise is mainly caused by an instrument, in order to avoid the influence of the dark noise on the original flame radiation signal, the original flame radiation signal is measured for multiple times in an experiment besides ensuring that a spectrometer of a flame data acquisition system is in a normal use environment, and a statistical average value is taken so as to eliminate the dark noise; meanwhile, multiple spectrum collection is carried out on the background without flame, and the average value of the spectrum is taken as the background spectrum, so that the used flame spectrum is obtained by subtracting the background spectrum, and the influence of the background spectrum on the original flame radiation signal is avoided;
determining flame combustion characteristic parameters and flame temperature according to the flame full-spectrum signal;
extracting a spectral characteristic value of the alkali metal element according to the alkali metal element spectral signal;
generating a characteristic parameter matrix according to the flame combustion characteristic parameters, the flame temperature and the spectral characteristic values;
adopting the cyclic neural network model established by the characteristic parameter matrix training to generate a trained biomass fuel alkali metal element content online prediction model;
iteratively training the recurrent neural network model by adopting the characteristic parameter matrix, and circularly optimizing the model parameters of the recurrent neural network model based on a batch gradient descent method, wherein the ending conditions of model iterative training and cyclic optimization are that a loss function is less than a preset value of 0.05 or reaches a preset training time of 2000 cycles; after the iterative training of the model is finished, testing the prediction accuracy of the model by using a test set sample, and if the prediction accuracy is higher than a preset accuracy threshold, using the circulating neural network model at the moment as a trained biomass fuel alkali metal element content online prediction model; if the prediction accuracy is lower than the preset accuracy threshold, returning to the original flame radiation signal obtained when the biomass fuel is combusted, and re-training the circulating neural network model until the prediction accuracy is higher than the preset accuracy threshold, so as to obtain the trained biomass fuel alkali metal element content online prediction model;
and predicting the alkali metal element content of the biomass fuel combusted in the biomass boiler by adopting the biomass fuel alkali metal element content online prediction model.
2. The method for online dynamic prediction of alkali metal element content according to claim 1, wherein the determining of the flame combustion characteristic parameter and the flame temperature according to the flame full spectrum signal specifically comprises:
extracting flame combustion characteristic parameters according to the flame full spectrum signal; the flame combustion characteristic parameters comprise flame flicker frequency and radiation energy;
and measuring the flame temperature at the same moment as the flame full-spectrum signal based on a bicolor method.
3. The method for online dynamic prediction of alkali metal element content according to claim 2, wherein the extracting a spectral feature value of an alkali metal element according to the alkali metal element spectral signal specifically comprises:
eliminating the influence of black body radiation in the alkali metal element spectral signal on the spectral intensity to generate an alkali metal element spectral signal with the influence eliminated;
extracting the spectral characteristic value of the alkali metal element spectral signal after the influence is eliminated; the spectral characteristic values comprise characteristics of a mean value, a standard deviation, a kurtosis coefficient and a skewness coefficient.
4. The method for online dynamic prediction of alkali metal element content according to claim 3, wherein the generating of the trained biomass fuel online prediction model of alkali metal element content by using the recurrent neural network model established by the characteristic parameter matrix training specifically comprises:
acquiring an established recurrent neural network model; the recurrent neural network model comprises an input layer, a recurrent layer, a full connection layer and an output layer;
and training the cyclic neural network model by adopting the characteristic parameter matrix, and optimizing the model parameters of the cyclic neural network model based on a batch gradient descent method to generate a trained biomass fuel alkali metal element content online prediction model.
5. An online dynamic prediction system for alkali metal element content of biomass fuel, which is characterized by comprising:
the flame radiation signal acquisition module is used for acquiring an original flame radiation signal when the biomass fuel is combusted;
the data preprocessing module is used for preprocessing the original flame radiation signal to generate a preprocessed flame radiation signal; the pretreated flame radiation signals comprise pretreated flame full spectrum signals and alkali metal element spectrum signals;
the data preprocessing is performed on the original flame radiation signal to generate a preprocessed flame radiation signal, and the method specifically includes:
removing outliers in the original flame radiation signal by adopting a k-means clustering algorithm, performing smooth denoising treatment, simultaneously avoiding the influence of system errors on the original flame radiation signal, and generating a preprocessed flame radiation signal;
the effect of the systematic error on the raw flame radiation signal includes the effect of dark noise and background spectra on the raw flame radiation signal; the dark noise is mainly caused by an instrument, in order to avoid the influence of the dark noise on the original flame radiation signal, the original flame radiation signal is measured for multiple times in an experiment besides ensuring that a spectrometer of a flame data acquisition system is in a normal use environment, and a statistical average value is taken so as to eliminate the dark noise; meanwhile, multiple spectrum collection is carried out on the background without flame, and the average value of the spectrum is taken as the background spectrum, so that the used flame spectrum is obtained by subtracting the background spectrum, and the influence of the background spectrum on the original flame radiation signal is avoided;
the flame combustion characteristic parameter and flame temperature extraction module is used for determining the flame combustion characteristic parameter and the flame temperature according to the full spectrum signal of the flame;
the spectral feature extraction module is used for extracting a spectral feature value of the alkali metal element according to the alkali metal element spectral signal;
the characteristic parameter matrix generating module is used for generating a characteristic parameter matrix according to the flame combustion characteristic parameters, the flame temperature and the spectral characteristic values;
the model training module is used for training the established recurrent neural network model by adopting the characteristic parameter matrix to generate a trained biomass fuel alkali metal element content online prediction model;
iteratively training the recurrent neural network model by adopting the characteristic parameter matrix, and circularly optimizing the model parameters of the recurrent neural network model based on a batch gradient descent method, wherein the ending conditions of model iterative training and cyclic optimization are that a loss function is less than a preset value of 0.05 or reaches a preset training time of 2000 cycles; after the iterative training of the model is finished, testing the prediction accuracy of the model by using a test set sample, and if the prediction accuracy is higher than a preset accuracy threshold, using the circulating neural network model at the moment as a trained biomass fuel alkali metal element content online prediction model; if the prediction accuracy is lower than the preset accuracy threshold, returning to the original flame radiation signal obtained when the biomass fuel is combusted, and re-training the circulating neural network model until the prediction accuracy is higher than the preset accuracy threshold, so as to obtain the trained biomass fuel alkali metal element content online prediction model;
and the online prediction module of the alkali metal element content is used for predicting the alkali metal element content of the biomass fuel combusted in the biomass boiler by adopting the online prediction model of the alkali metal element content of the biomass fuel.
6. The system for online dynamic prediction of alkali metal element content according to claim 5, wherein the flame combustion characteristic parameter and flame temperature extraction module specifically comprises:
the flame combustion characteristic parameter extraction unit is used for extracting flame combustion characteristic parameters according to the flame full spectrum signal;
the flame combustion characteristic parameters comprise flame flicker frequency and radiation energy;
and the flame temperature measuring unit is used for measuring the flame temperature at the same moment as the full spectrum signal of the flame based on a bicolor method.
7. The system for online dynamic prediction of alkali metal element content according to claim 6, wherein the spectral feature extraction module specifically comprises:
the black body radiation eliminating unit is used for eliminating the influence of black body radiation in the alkali metal element spectral signal on the spectral intensity and generating an alkali metal element spectral signal after the influence is eliminated;
the spectral feature extraction unit is used for extracting the spectral feature value of the alkali metal element spectral signal after the influence is eliminated;
the spectral characteristic values comprise characteristics of a mean value, a standard deviation, a kurtosis coefficient and a skewness coefficient.
8. The system for online dynamic prediction of alkali metal element content according to claim 7, wherein the model training module specifically comprises:
the model establishing unit is used for acquiring the established recurrent neural network model; the recurrent neural network model comprises an input layer, a recurrent layer, a full connection layer and an output layer;
and the model training optimization unit is used for training the cyclic neural network model by adopting the characteristic parameter matrix, optimizing the model parameters of the cyclic neural network model based on a batch gradient descent method and generating a trained biomass fuel alkali metal element content online prediction model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010046096.0A CN111257306B (en) | 2020-01-16 | 2020-01-16 | Online dynamic prediction method and system for alkali metal element content of biomass fuel |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010046096.0A CN111257306B (en) | 2020-01-16 | 2020-01-16 | Online dynamic prediction method and system for alkali metal element content of biomass fuel |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111257306A CN111257306A (en) | 2020-06-09 |
| CN111257306B true CN111257306B (en) | 2021-02-09 |
Family
ID=70947095
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010046096.0A Active CN111257306B (en) | 2020-01-16 | 2020-01-16 | Online dynamic prediction method and system for alkali metal element content of biomass fuel |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111257306B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112013417B (en) * | 2020-08-25 | 2021-07-27 | 华中科技大学 | Combustion optimization adjustment method and system for high-alkali coal boiler |
| CN116448739A (en) * | 2023-03-21 | 2023-07-18 | 华北电力大学 | Method for measuring gas phase alkali metal concentration in furnace based on flame self-emission spectrum |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4896965A (en) * | 1988-09-14 | 1990-01-30 | The United States Of America As Represented By The United States Department Of Energy | Real-time alkali monitoring system |
| US5432090A (en) * | 1992-04-23 | 1995-07-11 | Hitachi, Ltd. | Method for measuring metal ingredients in combustion gas |
| CN104849260A (en) * | 2015-05-18 | 2015-08-19 | 华中科技大学 | Online detection method of concentration of gas-phase alkali metal in boiler combustion flame |
| CN106066314A (en) * | 2016-07-27 | 2016-11-02 | 华中科技大学 | The detection method of gas phase alkali metal concn in a kind of hydrocarbon combustion flame |
| CN110146455A (en) * | 2019-05-27 | 2019-08-20 | 安徽大学 | Laser spectrum gas concentration measuring method based on deep learning |
| CN110567910A (en) * | 2019-09-30 | 2019-12-13 | 华中科技大学 | Method for detecting mass concentration three-dimensional distribution of gas-phase alkali metal in combustion flame |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101701951B (en) * | 2009-10-28 | 2011-07-27 | 北京航空航天大学 | Fuel type on-line identification method based on support vector machine |
| WO2012055753A2 (en) * | 2010-10-27 | 2012-05-03 | Siemens Aktiengesellschaft | Device and method for gasifying carbon-containing fuels |
| CN102393366B (en) * | 2011-09-14 | 2013-04-10 | 上海德诺产品检测有限公司 | Method for indirectly measuring silicon content in defoaming agent through flame atomic absorption |
| US8709815B2 (en) * | 2011-11-04 | 2014-04-29 | Dexsil Corporation | System and method for the analysis of biodiesel |
-
2020
- 2020-01-16 CN CN202010046096.0A patent/CN111257306B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4896965A (en) * | 1988-09-14 | 1990-01-30 | The United States Of America As Represented By The United States Department Of Energy | Real-time alkali monitoring system |
| US5432090A (en) * | 1992-04-23 | 1995-07-11 | Hitachi, Ltd. | Method for measuring metal ingredients in combustion gas |
| CN104849260A (en) * | 2015-05-18 | 2015-08-19 | 华中科技大学 | Online detection method of concentration of gas-phase alkali metal in boiler combustion flame |
| CN106066314A (en) * | 2016-07-27 | 2016-11-02 | 华中科技大学 | The detection method of gas phase alkali metal concn in a kind of hydrocarbon combustion flame |
| CN110146455A (en) * | 2019-05-27 | 2019-08-20 | 安徽大学 | Laser spectrum gas concentration measuring method based on deep learning |
| CN110567910A (en) * | 2019-09-30 | 2019-12-13 | 华中科技大学 | Method for detecting mass concentration three-dimensional distribution of gas-phase alkali metal in combustion flame |
Non-Patent Citations (1)
| Title |
|---|
| 基于火焰检测的煤种识别及发热量预测;阚睿;《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》;20100815(第08期);C039-11 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111257306A (en) | 2020-06-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Application of hybrid model based on empirical mode decomposition, novel recurrent neural networks and the ARIMA to wind speed prediction | |
| CN111257306B (en) | Online dynamic prediction method and system for alkali metal element content of biomass fuel | |
| CN112488145B (en) | NO based on intelligent methodxOnline prediction method and system | |
| CN109670629B (en) | Coal-fired boiler thermal efficiency prediction method based on long-short term memory neural network | |
| CN114077846B (en) | Fault current multi-domain identification method based on RF-LSTM and storage medium | |
| CN111007035B (en) | Method, system and device for detecting concentration of high-temperature gas in secondary combustion chamber based on convolutional neural network and storage medium | |
| CN109978259B (en) | Method and system for predicting residual life of product based on q-type Weibull distribution | |
| Li et al. | Online dynamic prediction of potassium concentration in biomass fuels through flame spectroscopic analysis and recurrent neural network modelling | |
| CN108182553B (en) | Coal-fired boiler combustion efficiency online measurement method | |
| Su et al. | Accuracy improvement on quantitative analysis of the total iron content in branded iron ores by laser-induced breakdown spectroscopy combined with the double back propagation artificial neural network | |
| CN113297742B (en) | Self-energy non-invasive system based on independent component analysis and modeling method thereof | |
| CN117278150B (en) | Indoor wireless network signal measurement and calculation method, equipment and medium | |
| CN114119986A (en) | Spectral characteristic information extraction method for converter steelmaking endpoint prediction | |
| CN111967660B (en) | Ultra-short-term photovoltaic prediction residual error correction method based on SVR | |
| CN106650060B (en) | Photovoltaic cell internal resistance attenuation coefficient prediction method | |
| CN116401948A (en) | Online prediction method and system for generating amount of power station boiler ash based on LSTM | |
| CN114813635B (en) | Method for optimizing combustion parameters of coal stove and electronic equipment | |
| CN110458344A (en) | A kind of super short-period wind power prediction technique of auto-adaptive time resolution ratio | |
| CN115932435A (en) | Resident non-invasive load monitoring method based on low-frequency acquisition signals | |
| CN114896895A (en) | Gate-controlled cyclic neural network-based denitration system inlet nitrogen oxide emission prediction method and system | |
| CN115936236A (en) | Method, system, equipment and medium for predicting energy consumption of cigarette factory | |
| CN115660195A (en) | Carbon emission prediction method and system for industrial park multi-energy complementary system | |
| CN114298300A (en) | Uncertainty prediction method, uncertainty prediction device, electronic device, and storage medium | |
| CN114399024A (en) | Oil gas concentration big data intelligent detection system | |
| CN110633917A (en) | New energy policy implementation effect comprehensive evaluation method based on electric power big data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |