CN113326983B - Data processing system and method for natural gas consumption prediction - Google Patents
Data processing system and method for natural gas consumption prediction Download PDFInfo
- Publication number
- CN113326983B CN113326983B CN202110590350.8A CN202110590350A CN113326983B CN 113326983 B CN113326983 B CN 113326983B CN 202110590350 A CN202110590350 A CN 202110590350A CN 113326983 B CN113326983 B CN 113326983B
- Authority
- CN
- China
- Prior art keywords
- data
- natural gas
- prediction
- gas consumption
- historical
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
Abstract
The invention discloses a data processing system and a data processing method for natural gas consumption prediction.A data acquisition module is used for acquiring historical data of natural gas consumption and data of natural gas consumption influence factors; the data analysis and screening module is used for analyzing the relationship between the data of the historical influence factors acquired by the data acquisition module and the historical data of the natural gas consumption and selecting corresponding covariates; the first prediction module is used for carrying out preliminary prediction according to the historical data of the natural gas consumption, the data of the future influence factors and the corresponding covariates selected by the data analysis and screening module which are acquired by the data acquisition module; and the second prediction module is used for performing final prediction according to the preliminary prediction data obtained by the first prediction module, realizing scientific and accurate prediction of the natural gas consumption, further improving the reliability of the prediction data of the natural gas consumption and further realizing stable gas supply.
Description
Technical Field
The invention relates to the technical field of natural gas management, in particular to a data processing system and method for natural gas consumption prediction.
Background
In recent years, with the rapid development of economy in China, the demand of natural gas is rapidly increased, and in order to reduce the storage cost of the natural gas, reasonably convey the natural gas and realize stable gas supply, the gas consumption of the natural gas needs to be accurately predicted.
At present, the natural gas consumption of the next day is presumed by generally adopting an empirical method according to the natural gas consumption of the day, and the deviation between the predicted natural gas consumption and the actual natural gas consumption is large due to large interference of artificial subjective factors. When the predicted gas consumption is larger than the actual gas consumption, the storage cost of the natural gas is increased, and when the predicted gas consumption is smaller than the actual gas consumption, stable gas supply cannot be realized, so that an objective prediction method is needed to scientifically and accurately predict the natural gas consumption and realize stable gas supply.
Disclosure of Invention
The invention aims to provide a data processing system and a data processing method for predicting natural gas consumption, which can scientifically and accurately predict natural gas consumption and ensure stable natural gas supply.
To solve the above problems, an aspect of the present invention provides a data processing system for natural gas consumption prediction, including:
the data acquisition module is used for acquiring historical data of natural gas consumption and data of natural gas consumption influence factors, wherein the influence factors comprise historical influence factors and future influence factors;
the data analysis and screening module is used for analyzing the relationship between the data of the historical influence factors acquired by the data acquisition module and the historical data of the natural gas consumption and selecting corresponding covariates;
the first prediction module is used for carrying out preliminary prediction according to the historical data of the natural gas consumption, the data of the future influence factors and the corresponding covariates selected by the data analysis and screening module which are acquired by the data acquisition module, and calculating to obtain corresponding preliminary prediction data; and
and the second prediction module is used for carrying out final prediction according to the preliminary prediction data obtained by the first prediction module and calculating to obtain final prediction data of the natural gas consumption.
The system further comprises a data preprocessing module, a data processing module and a data processing module, wherein the data preprocessing module is used for carrying out normalization processing on the data of the influence factors of the natural gas consumption acquired by the data acquisition module and converting the data of each influence factor into dimensionless data; the formula of the normalization processing is as follows:
wherein: k is a time seriesDimension, k is 1,2, …, n, n is the number of time series dimensions; i is an influence factor, i is 1,2, …, m is the number of the influence factors; x is the number of i (k) ' is the data of the ith influence factor in the kth time series dimension, max [ x [ ] i (k)']Is the minimum value of the ith influence factor, min x i (k)']Is the maximum value of the ith influencing factor.
Further, the data analysis screening module comprises:
the first analysis submodule is used for analyzing the correlation degree between the preprocessed data of the historical influence factors and the natural gas consumption, calculating the correlation coefficient of each historical influence factor, and selecting data corresponding to a plurality of historical influence factors with the correlation degree ranking at the front as a first covariate to be input into the first prediction module;
the second analysis submodule is used for carrying out causal relationship and hysteresis order analysis on the preprocessed data of the historical influence factors and the natural gas consumption, and selecting data corresponding to the historical influence factors which have causal relationship with the natural gas consumption and have hysteresis order more than or equal to the prediction period number as second covariates to be directly input into the second prediction module; and
the third analysis submodule is used for carrying out correlation analysis on the preprocessed data of the historical influence factors and the natural gas consumption, calculating correlation coefficients of the historical influence factors, and selecting data corresponding to the historical influence factors with the correlation coefficients in an expected range as third covariates to be input into the first prediction module;
further, the first analysis submodule is configured to perform relevance analysis on the preprocessed data of the historical influence factors and the natural gas consumption through the following formula, so as to obtain a relevance coefficient of each historical influence factor:
wherein: zeta i (k) ' for each historical influence factor, corresponding to a correlation coefficient on each time series;
the third analysis submodule is used for carrying out correlation analysis on the preprocessed data of the historical influence factors and the natural gas consumption through the following formula, so that a correlation coefficient of each historical influence factor is obtained:
wherein: c o v (x) 0 ,x i ) For natural gas consumption historical data x 0 Influence factor x of natural gas consumption i The covariance of (a); σ x 0 For natural gas consumption historical data x 0 Standard deviation of (a), σ x i Influencing factor x for natural gas consumption i Standard deviation of (2).
Further, the first prediction module comprises:
the first calculation submodule is used for inputting the first covariate input by the first analysis submodule into a pre-constructed gray system model for prediction, calculating a first accumulated predicted value, restoring an actual predicted value according to the first accumulated predicted value to obtain multi-dimensional influence factor prediction temporary data, and inputting the multi-dimensional influence factor prediction temporary data into the second prediction module;
the second calculation submodule is used for directly inputting a second covariate selected by the second analysis submodule into the second prediction module, and the second covariate is multi-dimensional influence factor historical data calculated by the second analysis submodule;
the third calculation sub-module is used for inputting the natural gas consumption historical data acquired by the data acquisition module and a third covariate selected by the third analysis unit into a first neural network model which is constructed in advance for fitting to obtain first natural gas consumption prediction temporary data, and inputting the first natural gas consumption prediction temporary data into the second prediction module; and
and the fourth calculation submodule is used for inputting the historical natural gas consumption data acquired by the data acquisition module and the future influence factors into a pre-constructed addition model for fitting to obtain second temporary natural gas consumption prediction data and inputting the second temporary natural gas consumption prediction data into the second prediction module.
Further, the gray system model adopted by the first calculation submodule is:
wherein:
an original array of first covariates; z is a radical of (1) (k) A whitened background value; alpha is a coefficient of development; b is the amount of gray effect;
the first neural network model adopted by the third calculation submodule is as follows:
wherein:
is the output of the kth time series; σ is an activation function; w is a o The weight parameters are obtained by training a GRU (generalized regression Unit) recurrent neural network model; h is k The state of the hidden layer on the kth time sequence;
the fourth computation submodule adopts an addition model as follows:
y(k)＝g(k)+s(k)+h(k)+ε k
wherein: g (k) is a trend term which represents the variation trend of the time series on the non-period; s (k) is a seasonal term, which may be year, quarter, month, week, or day, etc.; h (k) is a cycle period item which can represent holidays, special dates and the like; epsilon k Is an error term.
Further, the second prediction module is further configured to input the temporary multidimensional influence factor prediction data, the historical multidimensional influence factor data, the temporary first natural gas consumption prediction data, and the temporary second natural gas consumption prediction data into a second neural network model that is constructed in advance for fitting, so as to generate final prediction data of natural gas consumption.
In another aspect of the present invention, a data processing method for natural gas consumption prediction is further provided, which includes the following steps:
s1: acquiring historical data of natural gas consumption, historical influence factor data of the natural gas consumption and future influence factor data;
s2: preprocessing the data of the historical influence factors and the data of the future influence factors, and converting the data of each influence factor into dimensionless data;
s3: analyzing the relationship between the data of the preprocessed historical influence factors and the historical data of the natural gas consumption in the step S2, and selecting corresponding covariates;
s4: inputting the historical natural gas consumption data acquired in the step S1, the data of the future influence factors preprocessed in the step S2 and the corresponding covariates selected in the step S3 into a first prediction module for preliminary prediction to obtain preliminary prediction data;
s5: and inputting the preliminary prediction data obtained in the step S4 into a second prediction module for final prediction to obtain final prediction data of the natural gas consumption.
Further, the specific steps of analyzing the relationship between the historical influence factor and the historical data of natural gas consumption in step S3 are as follows:
s301: analyzing the association degree of the data of the historical influence factors and the historical data of the natural gas consumption, calculating the association coefficient of each historical influence factor, sorting each historical influence factor according to the association degree, and selecting a plurality of historical influence factors with the top association degree as first covariates to be input into a first prediction module;
s302: analyzing the causal relationship and the hysteresis order of the data of the historical influence factors and the historical data of the natural gas consumption, and selecting the historical influence factors which have the causal relationship with the natural gas consumption and the hysteresis order of which is more than or equal to the prediction period number as second covariates to be input into the first prediction module;
s303: and performing correlation analysis on the data of the historical influence factors and the historical data of the natural gas consumption, calculating correlation coefficients of the historical influence factors, and selecting the historical influence factors with the correlation coefficients in an expected range as third covariates to be input into the first prediction module.
Further, the specific steps of the step S4 in which the first prediction module predicts that the preliminary prediction data is obtained are:
s401: inputting the first covariate into a pre-constructed gray system model for prediction, calculating a first accumulated predicted value, restoring an actual predicted value according to the first accumulated predicted value to obtain multi-dimensional influence factor prediction temporary data, and inputting the multi-dimensional influence factor prediction temporary data into a second prediction module;
s402: directly inputting the second covariate serving as multi-dimensional influence factor historical data into a second prediction module;
s403: inputting historical data of natural gas consumption and a third covariate into a first neural network model which is constructed in advance for fitting to obtain first temporary natural gas consumption prediction data, and inputting the first temporary natural gas consumption prediction data into a second prediction module;
s404: and inputting the historical data of the natural gas consumption and the data of the future influence factors into a pre-constructed addition model for fitting to obtain second temporary natural gas consumption prediction data, and inputting the second temporary natural gas consumption prediction data into a second prediction module.
According to the method, the historical data of the natural gas consumption and the data of the influence factors of the natural gas consumption are obtained through the data obtaining module, the influence factors influencing the natural gas consumption are selected based on the gray system model, the cyclic neural network model and the addition model, the data of the selected influence factors are preliminarily predicted, and the preliminarily predicted data are finally predicted based on the MLP multilayer perceptron.
Drawings
FIG. 1 is a block diagram of a gas consumption forecast data processing system according to the present invention.
FIG. 2 is a block diagram of the data analysis screening module of FIG. 1.
FIG. 3 is a block diagram of the first prediction module of FIG. 1.
Fig. 4 is a flow chart of a data processing method for natural gas consumption prediction according to the present invention.
Fig. 5 is a flowchart of step S3 in fig. 4.
Fig. 6 is a flowchart of step S4 in fig. 4.
Detailed Description
The invention will be further explained with reference to the drawings.
Example one
Fig. 1 is a block diagram of an embodiment of a data processing system for natural gas consumption prediction according to the present invention. The data processing system for predicting the natural gas consumption comprises a data acquisition module, a data preprocessing module, a data analysis and screening module, a first prediction module and a second prediction module, wherein historical consumption of natural gas is counted, data of influence factors of the natural gas consumption are acquired, a pre-constructed gray system model, a first neural network model, an addition model and a second neural network model are used for analyzing influence weight of each influence factor on the natural gas consumption, scientific and accurate prediction of the trend of the natural gas consumption is achieved, reliability of prediction data of the natural gas consumption is improved, and stable gas supply is achieved.
The data acquisition module is used for acquiring historical data of natural gas consumption and data of influence factors of the natural gas consumption.
The historical data of the natural gas consumption is data corresponding to the historical use condition of the natural gas in a certain area within a certain time, wherein the certain time can be one year, one quarter, one month or several days and the like, and the historical data of the natural gas consumption can be obtained through statistical data of a statistical department or a related energy company. In this embodiment, the obtained historical data of natural gas consumption may be used as an array x 0 (k) ' means:
x 0 (k)'＝{x 0 (1)',x 0 (2)'，…，x 0 (n)'} (1)
wherein: k is the time-series dimension, k is 1,2, …, n is the number of the time-series dimension k, namely x 0 (1) ' represents historical data of natural gas consumption in a first time series dimension, and so on.
The data of the influence factors of the natural gas consumption are data of factors influencing the natural gas consumption in a certain area within a certain time, and the influence factors comprise historical influence factors and future influence factors. In this embodiment, the historical influence factors include, but are not limited to, one or any combination of natural factors (e.g., temperature, air quality, weather, etc.), economic factors (e.g., holidays, urbanization rate, CPI, GDP, second industry production value, population quantity, etc.), natural gas industry factors (e.g., alternative energy price, natural gas production, natural gas import, energy structure, etc.), and policy factors (e.g., price policy, coal-to-gas policy, energy policy, etc.); the future impact factors include, but are not limited to, holidays and/or weather. The influence factors can be divided into coarse-grained data and fine-grained data according to the statistical period of the data of the influence factors, specifically, data counted by year or quarter such as GDP, CPI, second industry production value, population quantity and the like are taken as coarse-grained data, and data counted by month or day such as natural gas production, weather, temperature, air quality and the like are taken as fine-grained data. In this embodiment, the obtained data of the impact factor may be used as the array x i (k) Represents:
wherein: k is the time series dimension, k is 1,2, …, n is the number of the time series dimension k; i is an influence factor, i is 1,2, …, m is the number of the influence factors, namely x 1 (1) ' data representing a first influence factor over a first time series, x 2 (2) ' denotes data of the second influencing factor over the second time series, and so on.
The data preprocessing module is used for preprocessing the data of the influence factors of the natural gas consumption acquired by the data acquisition module. Because the data units of the influence factors acquired by the data acquisition module are different, the data of the influence factors can be compared and weighted conveniently, the prediction accuracy of the gray system model, the first neural network model, the addition model and the second neural network model in the first prediction module and the second prediction module is improved, normalization processing needs to be carried out on the influence factors before analysis and screening, and the data of the influence factors are converted into dimensionless data. In this embodiment, the normalized natural gas influence factor data x i (k) Comprises the following steps:
wherein: k is the time series dimension, k is 1,2, …, n, n is the number of the time series dimension; i is an influence factor, i is 1,2, …, m is the number of the influence factors; x is the number of i (k) ' is the data of the ith influencing factor in the k-th time series dimension, max [ x i (k)']Is the minimum value of the ith influence factor, min x i (k)']Is the maximum value of the ith influencing factor.
The data analysis and screening module is used for analyzing the relation between the data of the historical influence factors and the historical data of the natural gas consumption after the data acquisition module acquires the data and the data are normalized by the data preprocessing module so as to select the corresponding covariates.
Referring to fig. 2, the data analyzing and screening module includes a first analyzing sub-module, a second analyzing sub-module, and a third analyzing sub-module, which respectively perform association degree analysis, causal relationship, hysteresis order analysis, and correlation analysis on the historical influence factors, and determine a corresponding relationship between data of each historical influence factor and historical data of natural gas consumption.
And the first analysis submodule is used for carrying out association degree analysis on the data of the history influence factors after the normalization processing and the history data of the natural gas consumption, determining the association degree between each history influence factor and the natural gas consumption, and selecting the history influence factors meeting the conditions for next prediction. Because each historical influence factor comprises coarse-grained data and fine-grained data, when the historical influence factors are coarse-grained data, because the statistical sample size of the coarse-grained data is small, in order to improve the prediction accuracy of the system and improve the speed of the system for processing the data, in the embodiment, a grey correlation analysis method is adopted to carry out correlation analysis on the coarse-grained historical influence factors.
Specifically, the correlation coefficient ζ of each historical influence factor corresponding to each time sequence is calculated respectively i (k)'：
Wherein: x is the number of 0 (k) Historical data of natural gas consumption on the kth time series; ρ is an adjustment coefficient for adjusting an error range of the output result, and in this embodiment, the value of ρ is preferably 0.5.
The association coefficient ζ of each time series corresponding to each historical influence factor obtained according to the above formula (4) i (k) ' calculating the average value r of the time series correlation coefficient corresponding to each historical influence factor 0i By the average value r 0i The correlation coefficient ζ as each history influence factor i (k)：
Wherein: k is the time-series dimension, k is 1,2, …, and n is the number of the time-series dimension k.
According to the correlation coefficient zeta of the historical influence factors i (k) And sorting the relevance degrees from big to small, and selecting a plurality of items of data of historical influence factors with the top relevance degree as first covariates to be input into the first prediction module according to the precision of the predicted data.
The second analysis submodule is used for analyzing the causal relationship and the hysteresis order of the data of the historical influence factors and the historical data of the natural gas consumption, which are acquired by the data acquisition module, determining the causal relationship existing between the historical influence factors and the natural gas consumption and the hysteresis condition of the historical influence factors influencing the natural gas consumption, and selecting the historical influence factors of which the causal relationship meets the preset conditions for next prediction. In this example, the granular historical impact factors were subjected to causal and hysteresis order analysis using the granger causal test.
Specifically, a linear regression model is established, and for the historical data of natural gas consumption in n time series dimensions, the granger causal relationship between the historical data and other influence factors in the corresponding time series dimensions is analyzed one by one, and the linear regression model can be expressed as:
wherein: x is the number of it Historical data of the consumption amount of the natural gas to be analyzed on the current time sequence is obtained; p is the hysteresis order of the natural gas consumption historical data; x is the number of it-p Is x it 1,2, …, p; alpha is a regression coefficient of the natural gas consumption historical data on the current time series to be analyzed; q is a hysteresis order of an influence factor corresponding to the historical data of the natural gas consumption to be analyzed on the current time sequence; x is the number of 0t-q Q is a lag term of an influence factor corresponding to the gas consumption history data to be analyzed on the current time series, and q is 1,2, …, q; beta is a regression coefficient of an influence factor corresponding to the historical data of the consumption amount of the natural gas to be analyzed on the current time sequence; epsilon t White noise, the prediction error term of the regression model.
According to the equation (6), when β is 0, the linear regression model is a regression model not including other influence factors, and when β is not 0, the linear regression model is a regression model including other influence factors, and the prediction error term ∈ in the regression model when β does not include other influence factors is calculated respectively t Residual sum of squares RSS of r And the prediction error term epsilon in the regression model when other influence factors are included t Residual sum of squares RSS of u 。
Then according to the residual square sum RSS without other influence factors r And residual sum of squares RSS including other influencing factors u Constructing an F statistic, wherein the calculation formula of the F statistic is as follows:
wherein: n is the number of time series dimensions; p is the hysteresis order of the natural gas consumption historical data; q is the hysteresis order of the impact factor corresponding to the historical data of natural gas consumption to be analyzed over the current time series.
And (4) checking whether each influence factor can obviously influence the historical data of the natural gas consumption one by one according to the formula (7). Specifically, the traversal is performed according to the maximum hysteresis order which is three times (preferably) the prediction period number, and in the traversal process, if the calculated F statistic obeys F distribution and shows that the influence factor and the natural gas consumption have a Glangen causal relationship, the data (preferably) with the p value smaller than 0.05 is selected to be directly used as the input parameter of the second prediction module (namely, the second neural network model).
And the third analysis submodule is used for carrying out correlation analysis on the data of the historical influence factors acquired by the data acquisition module and the historical data of the natural gas consumption, determining the correlation between each historical influence factor and the natural gas consumption, and selecting the historical influence factor of which the correlation coefficient is within the range of preset conditions for next prediction. When the historical influence factor is fine-grained data, correlation analysis is performed on the fine-grained data for improving the prediction precision of the system and improving the data processing speed of the system due to the fact that the statistical sample size of the fine-grained data is large.
Specifically, in this embodiment, a correlation coefficient between each historical influence factor and the historical data of natural gas consumption is calculated by using a pearson correlation analysis method. The calculation formula of the Pearson correlation coefficient is as follows:
wherein: c o v (x) 0 ,x i ) For natural gas consumption historical data x 0 Influence factor x of natural gas consumption i The covariance of (a); σ x 0 For natural gas consumption historical data x 0 Standard deviation of (a), σ x i Influencing factor x for natural gas consumption i Standard deviation of (2).
And (4) selecting an influence factor of the correlation coefficient in an expected range as a third covariate to be input into the first prediction module according to the Pearson correlation coefficient calculated by the formula (8). In this embodiment, the influence factor with a correlation coefficient greater than 0.8 or less than-0.8 is selected as the third covariate.
The first prediction module is used for carrying out preliminary prediction according to the historical data of the natural gas consumption, the data of the future influence factors and the corresponding covariates selected by the data analysis and screening module, which are acquired by the data acquisition module and subjected to normalization processing, and calculating to obtain corresponding preliminary prediction data.
Referring to fig. 3, the first prediction module includes a first computation submodule, a second computation submodule, a third computation submodule and a fourth computation submodule, and performs preliminary prediction on the processed historical data of the natural gas consumption, the data of the future influence factors and the first covariates, the second covariates and the third covariates obtained by the data analysis and screening module based on a pre-constructed gray system model, a first neural network model and an addition model, so as to obtain preliminary prediction data corresponding to the natural gas consumption and the data of each influence factor.
The first calculation submodule is used for inputting the first covariates input by the first analysis submodule into a pre-constructed gray system model for prediction, predicting the first covariates corresponding to the influence factors respectively based on the gray system model, calculating to obtain multi-dimensional influence factor prediction temporary data, and inputting the multi-dimensional influence factor prediction temporary data into the second prediction module. In this embodiment, the multidimensional influence factor prediction temporary data is corresponding temporary data obtained by predicting each influence factor based on data in each time series dimension; for example, when the input first covariate is data corresponding to a GDP, CPI, or other coarse-grained indicators, the temporary prediction data is temporary prediction data of the coarse-grained indicators such as the GDP, CPI, etc. predicted based on the gray system model. Because the first covariate is data corresponding to the coarse-grained index, the predicted value of the first covariate can be used for fitting the trend of the natural gas consumption in a subsequent second prediction module.
Specifically, in this embodiment, GM (1,1) gray system models are used to predict the input first covariates, and generate multidimensional impact factor prediction temporary data. First, a GM (1,1) gray system model is built:
wherein:
an original array of first covariates; z is a radical of (1) (k) A whitened background value; alpha is a coefficient of development; b is the amount of gray effect.
Subjecting the above formula (9) to regression analysis if it is to be
The time series t corresponding to each data in the array is considered to be continuous,
viewed as a function of time series t (i.e.
) Then can be connected to
Become a continuous function thereof
Derivative of (2)
At the same time z (1) (k) Corresponding to the derivative
And obtaining the whitening model corresponding to the gray system model by obtaining the estimated values of alpha and b as follows:
solving the formula (10) to obtain the first accumulated predicted value
Wherein:
an estimated value of the development coefficient alpha;
an estimate of the amount of grey contribution b;
first data of a first covariate original array;
can be based on
And
solving by a least square method; e is the base of the natural logarithm and is a constant.
Obtaining an actual predicted value according to the formula (11)
Comprises the following steps:
wherein:
accumulating the predicted value for the t-th time;
is the t-1 th predicted accumulated value.
The second calculation submodule is used for directly inputting a second covariate selected by the second analysis submodule into the second prediction module, and the second covariate is the multi-dimensional influence factor historical data calculated by the second analysis submodule. In this embodiment, the multi-dimensional influence factor historical data is data of each influence factor based on each time series dimension; for example, when the input second covariate is data corresponding to fine-grained indexes such as natural gas yield and temperature, the multi-dimensional influence factor historical data is historical data of the indexes such as natural gas yield and temperature after the normalization processing.
The third calculation sub-module is used for inputting the historical data of the natural gas consumption amount obtained by the data obtaining module and subjected to normalization processing and third covariates selected by the third analysis unit into a first neural network model which is constructed in advance, fitting the historical data of the natural gas consumption amount and each third covariate based on the first neural network model to obtain first temporary data of natural gas consumption amount prediction, and inputting the first temporary data of the natural gas consumption amount prediction into the second prediction module. In this embodiment, the first gas consumption prediction temporary data is prediction data derived based on data of a historical influence factor.
Specifically, in this embodiment, the GRU recurrent neural network is used to fit the historical data of the usage amount of the natural gas to each of the third covariates.
Firstly, a GRU circulation neural network model is constructed based on historical data of natural gas usage and data of a plurality of influence factors which are screened out by the Larson correlation analysis and have strong correlation with the natural gas usage. The GRU circulating neural network model comprises two hidden layers with neuron number of 64 and a full connection layer, and the GRU circulating neural network model takes the historical data of the natural gas usage and the data of a plurality of influence factors which are selected by correlation analysis and strongly related to the natural gas usage as the input. The GRU recurrent neural network model can be expressed as:
wherein:
is the output of the kth time sequence (namely the output of the current output layer of the GRU recurrent neural network model); sigma is an activation function; w is a o The weight parameters are obtained by training a GRU (generalized regression Unit) recurrent neural network model; h is k The state of the hidden layer on the kth time series.
Then, the data sequence x of each influence factor after normalization processing is respectively processed i (k) And a historical data series x of natural gas usage 0 (k) Inputting the third covariate into the GRU cyclic neural network model, and performing iterative training on the GRU cyclic neural network model by respectively adopting 12 months, 365 days, 30 days and 7 days as backtracking time windows according to the fine-grained data index because the third covariate corresponds to the data of the historical influence factor of the fine-grained index, thereby optimizing the model parameters.
Finally, historical data of natural gas usage on the k time series is obtained
Inputting the data into a trained GRU cyclic neural network model to obtain the natural gas usage data of k +1, k +2, … and k + n time series
Data on the natural gas usage amount of the k +1, k +2, …, k + n time series
Namely predicting temporary data of the first natural gas consumption.
The fourth calculation submodule is used for inputting the natural gas consumption historical data and the data of the future influence factors, which are acquired by the data acquisition module and subjected to normalization processing, into a pre-constructed addition model, fitting the natural gas consumption historical data and the data of the future influence factors based on the addition model to obtain second natural gas consumption prediction temporary data, and inputting the second natural gas consumption prediction temporary data into the second prediction module.
Specifically, in this embodiment, a time-series addition model y (k) is constructed based on the change trend, seasonal change, and cyclic period fluctuation of the natural gas consumption historical data and the future influence factor data:
y(k)＝g(k)+s(k)+h(k)+ε k (14)
wherein: g (k) is a trend term which represents the variation trend of the time series on the non-period; s (k) is a seasonal term, which may be year, quarter, month, week, or day, etc.; h (k) is a cycle period item which can represent holidays, special dates and the like; epsilon k Is an error term.
When data fitting is carried out, the trend item reflects the change of time series trend, in particular the trend that the time series continuously rises or falls towards a certain direction or stays on a certain level. The seasonal item reflects seasonal variation of a time series, the seasonal variation refers to a rule that the usage amount of natural gas varies with the seasonal variation is objective, the seasonal item refers to a time period which is elapsed after the seasonal variation of one period is completed, and the seasonal item is a generalized season and can refer to year, quarter, month, week, day and the like. For example, in a year, since heating is required in the winter, the consumption of natural gas will peak at a high level in the winter, which is a seasonal periodic fluctuation. The cyclic period item reflects non-seasonal periodic fluctuation, and the period of the cyclic period item is a specific time period; for example, on holidays such as spring festival, national day, etc. or some other special date, there will be a peak in natural gas consumption, the change of which is regularly cyclable. The error term is the change of the consumption of the natural gas caused by some random factors, and the factors have unpredictable and irregular effects, such as relevant policies of national platform departure and the like. In the embodiment, the trend term, the seasonal term, the cycle period term and the error term are all fitted by using a fourier series.
And the second prediction module is used for inputting the corresponding covariates selected by the data analysis and screening module and the preliminary prediction data obtained by the first prediction module into a pre-constructed second prediction module, and calculating the covariates and the preliminary prediction data to obtain final prediction data of the natural gas consumption. The second prediction module predicts the multi-dimensional influence factor prediction temporary data, the multi-dimensional influence factor historical data, the first natural gas consumption prediction temporary data and the second natural gas consumption prediction temporary data which are calculated by the first calculation sub-module based on a pre-constructed second neural network model to obtain final natural gas consumption prediction data.
Specifically, in this embodiment, the second prediction model adopts an MLP multilayer sensing machine, a trained MLP multilayer sensing machine network is constructed based on the multi-dimensional influence factor prediction temporary data, the multi-dimensional influence factor historical data, the first natural gas consumption prediction temporary data, and the second natural gas consumption prediction temporary data, iterative training is performed, and model parameters are optimized, and then the multi-dimensional influence factor prediction temporary data, the multi-dimensional influence factor historical data, the first natural gas consumption prediction temporary data, and the second natural gas consumption prediction temporary data are used as input predictions of the MLP multilayer sensing machine network to obtain final natural gas consumption.
The data processing system for natural gas consumption prediction of the embodiment selects the influence factors influencing the natural gas consumption based on the grey system model, the cyclic neural network model and the addition model, preliminarily predicts the data of the selected influence factors, and finally predicts the natural gas consumption based on the preliminarily predicted data by the MLP multilayer perceptron.
Example two
Fig. 4 is a flow chart of an embodiment of the data processing method for predicting natural gas consumption according to the present invention. The method for predicting natural gas consumption according to the embodiment is implemented based on the data processing system for predicting natural gas consumption according to the first embodiment, and specifically includes the following steps:
s1: and (6) acquiring data.
And acquiring historical data of the natural gas consumption, and data of historical influence factors and data of future influence factors of the natural gas consumption. Specifically, the step is based on the historical data and the influence factors of the natural gas consumption acquired by the data acquisition module in the natural gas consumption prediction system.
S2: and (4) preprocessing data.
And preprocessing the data of the historical influence factors and the data of the future influence factors, and converting the data of each influence factor into dimensionless data. Specifically, in this step, normalization processing is performed on each influence factor before analysis and screening is performed on each influence factor based on a data preprocessing module in the natural gas consumption prediction system, and data of each influence factor is converted into dimensionless data.
S3: and (4) analyzing and screening data.
And selecting a corresponding covariate based on the relationship between the data of the history influence factors preprocessed in the step S2 and the history data of the natural gas consumption by the data analysis screening module in the natural gas consumption prediction system. Specifically, the correlation, the causal relationship, the hysteresis order and the correlation between the data of the history influence factors after the normalization processing and the history data of the natural gas consumption are analyzed to obtain a corresponding first covariate, a corresponding second covariate and a corresponding third covariate.
Referring to fig. 5, the specific steps of analyzing the relationship between the historical influence factor and the historical data of natural gas consumption in step S3 are as follows:
s301: and (5) analyzing the relevance.
The correlation degree analysis is carried out on the data of the historical influence factors and the historical data of the natural gas consumption based on a first analysis submodule in the natural gas consumption prediction system. In this embodiment, a grey correlation analysis method is used to perform correlation analysis on coarse-grained historical influence factors. Specifically, first, a correlation coefficient ζ of each historical influence factor corresponding to each time series is calculated i (k) '; then, the correlation coefficient zeta corresponding to each time series based on the historical influence factors i (k) ' calculating the correlation coefficient ζ of each historical influence factor i (k) (ii) a Finally according to the correlation coefficient zeta i (k) And sorting the relevance degrees from big to small, and selecting a plurality of items of data of historical influence factors with the top relevance degree as first covariates to be input into the first prediction module according to the precision of the predicted data.
S302: causal relationship and hysteresis order analysis.
The causal relationship and the hysteresis order analysis are carried out on the data of the historical influence factors and the historical data of the natural gas consumption by a second analysis submodule in the natural gas consumption prediction system. In this example, the granular historical impact factors were subjected to causal and hysteresis order analysis using the granger causal test. Specifically, a linear regression model is established, and for the historical data of the natural gas consumption in each time series dimension, the gram cause-effect relationship between the historical data and other influence factors in the corresponding time series dimension is analyzed one by one; and then, checking whether the influence factors can obviously influence the historical data of the natural gas consumption one by one, and selecting the data of the influence factors with the hysteresis order meeting the preset conditions as the input parameters of the second prediction module directly.
S303: and (5) carrying out correlation analysis.
The third analysis submodule in the natural gas consumption prediction system performs correlation analysis on the data of the historical influence factors and the historical data of the natural gas consumption. In the embodiment, a correlation coefficient between each historical influence factor and the historical data of the natural gas consumption is calculated by adopting a Pearson correlation analysis method. Specifically, the Pearson correlation coefficient of each historical influence factor is calculated, and the historical influence factors with the correlation coefficients in an expected range are selected as third covariates to be input into the first prediction module.
S4: and (4) primarily predicting the related data of the natural gas consumption.
In this step, based on the first prediction module in the natural gas consumption prediction system, the historical data of the natural gas consumption obtained in step S1, the data of the future influence factors preprocessed in step S2, and the corresponding covariates selected in step S3 are input into the first prediction module for preliminary prediction, so as to obtain preliminary prediction data.
Referring to fig. 6, the specific steps of the step S4 in which the first prediction module predicts the preliminary prediction data are as follows:
s401: and calculating the multi-dimensional influence factor prediction temporary data.
Inputting a first covariate into a pre-constructed grey system model for prediction based on a first calculation submodule in the natural gas consumption prediction system, calculating a first accumulated predicted value, restoring an actual predicted value according to the first accumulated predicted value, and inputting the obtained multi-dimensional influence factor prediction temporary data into a second prediction module. In this embodiment, the gray system model employs a GM (1,1) gray system model.
S402: and calculating multi-dimensional influence factor historical data.
In the step, a second covariate serving as multi-dimensional influence factor historical data is directly input into a second prediction module on the basis of a second calculation submodule in the natural gas consumption prediction system. In this embodiment, the multi-dimensional influence factor historical data is data of each influence factor based on each time series dimension; for example, when the input second covariate is data corresponding to fine-grained indexes such as natural gas yield and temperature, the multi-dimensional influence factor historical data is historical data of the indexes such as natural gas yield and temperature after the normalization processing.
S403: and calculating the first natural gas consumption prediction temporary data.
The third calculation sub-module in the natural gas consumption prediction system inputs the historical data of the natural gas consumption and the third covariate into a first neural network model which is constructed in advance for fitting, and the temporary data of the first natural gas consumption prediction are obtained and input into a second prediction module. In this embodiment, the GRU recurrent neural network is used to fit the historical data of the usage amount of natural gas to each of the third covariates. The GRU circulation neural network model comprises two hidden layers with neuron number of 64 and a full connection layer, and the GRU circulation neural network model takes historical data of natural gas usage and data of a plurality of influence factors which are selected by correlation analysis and strongly related to the natural gas usage as input of the GRU circulation neural network model, iterative training is carried out, and first temporary data of natural gas consumption prediction are obtained.
S404: and calculating second natural gas consumption prediction temporary data.
The historical data of the natural gas consumption and the data of the future influence factors are input into a pre-constructed addition model to be fitted based on a fourth calculation submodule in the natural gas consumption prediction system, and second temporary natural gas consumption prediction data are obtained and input into a second prediction module. In the embodiment, a time series addition model is constructed based on the change trend, seasonal change and cycle period fluctuation of the natural gas consumption historical data and the future influence factor data, and the trend term, the seasonal term, the cycle period term and the error term are fitted through the time series addition model to obtain second temporary natural gas consumption prediction data.
S5: and predicting the consumption of natural gas.
In this step, the preliminary prediction data obtained in the step S4 is input to a second neural network model pre-constructed in a second prediction module for final prediction based on the second prediction module in the natural gas consumption prediction system, so as to obtain final prediction data of the natural gas consumption.
Specifically, the multi-dimensional influence factor prediction temporary data, the multi-dimensional influence factor historical data, the first natural gas consumption prediction temporary data and the second natural gas consumption prediction temporary data are input into a second neural network model for fitting, and final prediction data of the natural gas consumption are generated. In this embodiment, the second prediction model adopts an MLP multilayer perceptron, an MLP multilayer perceptron network is trained based on the multi-dimensional influence factor prediction temporary data, the multi-dimensional influence factor historical data, the first natural gas consumption prediction temporary data and the second natural gas consumption prediction temporary data, iterative training is performed and model parameters are optimized, and then the multi-dimensional influence factor prediction temporary data, the multi-dimensional influence factor historical data, the first natural gas consumption prediction temporary data and the second natural gas consumption prediction temporary data are used as input of the MLP multilayer perceptron network to predict and obtain final natural gas consumption.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. The above description is only an embodiment of the present invention, and not intended to limit the scope of the present invention, and all equivalent structures made by using the contents of the present specification and the drawings can be directly or indirectly applied to other related technical fields, and are within the scope of the present invention.
Claims (10)
1. A gas consumption forecast data processing system, comprising:
the data acquisition module is used for acquiring historical data of natural gas consumption and data of natural gas consumption influence factors, wherein the influence factors comprise historical influence factors and future influence factors;
the data analysis and screening module is used for analyzing the relation between the historical influence factor data acquired by the data acquisition module and the historical natural gas consumption data and selecting corresponding covariates;
the data analysis and screening module is further used for respectively carrying out association degree analysis of coarse-grained data, causal relationship and hysteresis order analysis of fine-grained data and correlation analysis of fine-grained data on the data of the historical influence factors and the natural gas consumption, and selecting corresponding first covariates, second covariates and third covariates;
the first prediction module is used for carrying out preliminary prediction according to the historical data of the natural gas consumption, the future influence factor data and the corresponding covariates selected by the data analysis and screening module, which are acquired by the data acquisition module, and calculating to obtain corresponding preliminary prediction data;
the first prediction module is further used for obtaining multi-dimensional influence factor prediction temporary data according to the first covariate prediction, obtaining multi-dimensional influence factor historical data according to the second covariate, obtaining first natural gas consumption prediction temporary data according to the third covariate fitting, and obtaining second natural gas consumption prediction temporary data according to the natural gas consumption historical data and the future influence factors by fitting, and the second natural gas consumption prediction temporary data is obtained by fitting through an addition model based on Fourier series when fitting is carried out according to the natural gas consumption historical data and the future influence factors; and
the second prediction module is used for carrying out final prediction according to the preliminary prediction data obtained by the first prediction module and calculating to obtain final prediction data of the natural gas consumption;
the second prediction module is further used for fitting the multi-dimensional influence factor prediction temporary data, the multi-dimensional influence factor historical data, the first natural gas consumption prediction temporary data and the second natural gas consumption prediction temporary data to obtain final prediction data of the natural gas consumption.
2. The natural gas consumption prediction data processing system according to claim 1, further comprising a data preprocessing module, configured to perform normalization processing on the data of the influence factors of the natural gas consumption acquired by the data acquisition module, and convert the data of each influence factor into dimensionless data; the formula of the normalization processing is as follows:
wherein: k is the time series dimension, k is 1,2, …, n is the number of the time series dimension; i is an influence factor, i is 1,2, …, m is the number of the influence factors; x is a radical of a fluorine atom i (k) ' is the data of the ith influencing factor in the k-th time series dimension, max [ x i (k)']Is the minimum value of the ith influence factor, min x i (k)']Is the maximum value of the ith influencing factor.
3. The system of claim 2, wherein the data analysis screening module comprises:
the first analysis submodule is used for analyzing the correlation degree between the preprocessed data of the historical influence factors and the natural gas consumption, calculating the correlation coefficient of each historical influence factor, and selecting data corresponding to a plurality of historical influence factors with the correlation degree ranking at the front as a first covariate to be input into the first prediction module;
the second analysis submodule is used for carrying out causal relationship and hysteresis order analysis on the preprocessed data of the historical influence factors and the natural gas consumption, and selecting data corresponding to the historical influence factors which have causal relationship with the natural gas consumption and have hysteresis order more than or equal to the prediction period number as second covariates to be directly input into the second prediction module; and
and the third analysis submodule is used for carrying out correlation analysis on the preprocessed data of the historical influence factors and the natural gas consumption, calculating correlation coefficients of the historical influence factors, and selecting data corresponding to the historical influence factors with the correlation coefficients in an expected range as third covariates to be input into the first prediction module.
4. The system of claim 3, wherein the first analysis sub-module is configured to perform correlation analysis between the preprocessed data of the historical influence factors and the natural gas consumption to obtain a correlation coefficient of each historical influence factor by the following formula:
wherein: zeta i (k) ' for each historical influence factor, corresponding to a correlation coefficient on each time series;
the third analysis submodule is used for carrying out correlation analysis on the preprocessed data of the historical influence factors and the natural gas consumption through the following formula, so that a correlation coefficient of each historical influence factor is obtained:
wherein: c o v (x) 0 ,x i ) For natural gas consumption historical data x 0 Influence factor x of natural gas consumption i The covariance of (a); σ x 0 For natural gas consumption historical data x 0 Standard deviation of (a), σ x i Influencing factor x for natural gas consumption i Standard deviation of (2).
5. The gas consumption prediction data processing system of claim 3, wherein the first prediction module comprises:
the first calculation submodule is used for inputting the first covariate input by the first analysis submodule into a pre-constructed gray system model for prediction, calculating a first accumulated predicted value, restoring an actual predicted value according to the first accumulated predicted value to obtain multi-dimensional influence factor prediction temporary data, and inputting the multi-dimensional influence factor prediction temporary data into the second prediction module;
the second calculation submodule is used for directly inputting a second covariate selected by the second analysis submodule into the second prediction module, and the second covariate is multi-dimensional influence factor historical data calculated by the second analysis submodule;
the third calculation sub-module is used for inputting the historical data of the natural gas consumption obtained by the data obtaining module and a third covariate selected by the third analysis unit into a first neural network model which is constructed in advance for fitting to obtain first temporary data of the natural gas consumption prediction, and inputting the first temporary data of the natural gas consumption prediction into the second prediction module; and
and the fourth calculation submodule is used for inputting the historical natural gas consumption data acquired by the data acquisition module and the future influence factors into a pre-constructed addition model for fitting to obtain second temporary natural gas consumption prediction data and inputting the second temporary natural gas consumption prediction data into the second prediction module.
6. The natural gas consumption prediction data processing system of claim 5, wherein the gray system model used by the first computation sub-module is:
wherein:
an original array of first covariates; z is a radical of formula (1) (k) A whitened background value; alpha is a coefficient of development; b is the amount of gray effect;
the first neural network model adopted by the third calculation submodule is as follows:
wherein:
is the output of the kth time series; sigma is an activation function; w is a o The weight parameters are obtained by training a GRU (generalized regression Unit) recurrent neural network model; h is k The state of the hidden layer on the kth time sequence;
the fourth calculation submodule adopts an addition model as follows:
y(k)＝g(k)+s(k)+h(k)+ε k
wherein: g (k) is a trend term which represents the variation trend of the time series on the non-period; s (k) is a seasonal term representing a year, quarter, month, week, or day; h (k) is a cycle period item which represents holidays and special dates; epsilon k Is an error term.
7. The natural gas consumption prediction data processing system according to claim 5, wherein the second prediction module is further configured to input the multidimensional influence factor prediction temporary data, the multidimensional influence factor historical data, the first natural gas consumption prediction temporary data, and the second natural gas consumption prediction temporary data into a second neural network model which is constructed in advance for fitting, so as to generate final prediction data of natural gas consumption.
8. A data processing method for natural gas consumption prediction is characterized by comprising the following steps:
s1: acquiring historical data of natural gas consumption, data of historical influence factors of the natural gas consumption and data of future influence factors;
s2: preprocessing the data of the historical influence factors and the data of the future influence factors, and converting the data of each influence factor into dimensionless data;
s3: analyzing the relationship between the data of the preprocessed historical influence factors and the historical data of the natural gas consumption in the step S2, and selecting corresponding covariates; the covariates comprise a first covariate, a second covariate and a third covariate which are obtained by respectively carrying out association analysis on coarse-grained data, causal relationship and hysteresis order analysis on fine-grained data and correlation analysis on the fine-grained data on the data of the historical influence factors and the natural gas consumption;
s4: inputting the historical natural gas consumption data acquired in the step S1, the data of the future influence factors preprocessed in the step S2 and the corresponding covariates selected in the step S3 into a first prediction module for preliminary prediction to obtain preliminary prediction data; the preliminary prediction data comprise multidimensional influence factor prediction temporary data obtained based on first covariate prediction, multidimensional influence factor historical data obtained based on second covariates, first natural gas consumption prediction temporary data obtained based on third covariate fitting and second natural gas consumption prediction temporary data obtained based on natural gas consumption historical data and future influence factors through fitting, and the second natural gas consumption prediction temporary data are obtained through fitting by adopting an addition model based on Fourier series when fitting is carried out according to the natural gas consumption historical data and the future influence factors;
s5: inputting the preliminary prediction data obtained in the step S4 into a second prediction module for final prediction to obtain final prediction data of natural gas consumption; the final prediction data of the natural gas consumption is the final prediction data of the natural gas consumption obtained by fitting based on the multi-dimensional influence factor prediction temporary data, the multi-dimensional influence factor historical data, the first natural gas consumption prediction temporary data and the second natural gas consumption prediction temporary data.
9. The method of claim 8, wherein the step S3 of analyzing the relationship between the historical influence factors and the historical data of natural gas consumption comprises the following specific steps:
s301: analyzing the association degree of the data of the historical influence factors and the historical data of the natural gas consumption, calculating the association coefficient of each historical influence factor, sorting each historical influence factor according to the association degree, and selecting a plurality of historical influence factors with the top association degree as first covariates to be input into a first prediction module;
s302: carrying out causal relationship and hysteresis order analysis on the data of the historical influence factors and the historical data of the natural gas consumption, and selecting the historical influence factors which have causal relationship with the natural gas consumption and have hysteresis order more than or equal to the prediction period number as second covariates to be input into the first prediction module;
s303: and performing correlation analysis on the data of the historical influence factors and the historical data of the natural gas consumption, calculating correlation coefficients of the historical influence factors, and selecting the historical influence factors with the correlation coefficients in an expected range as third covariates to be input into the first prediction module.
10. The method as claimed in claim 8, wherein the step S4 of predicting the preliminary prediction data by the first prediction module comprises the specific steps of:
s401: inputting the first covariate into a pre-constructed gray system model for prediction, calculating a first accumulated predicted value, restoring an actual predicted value according to the first accumulated predicted value to obtain multi-dimensional influence factor prediction temporary data, and inputting the multi-dimensional influence factor prediction temporary data into a second prediction module;
s402: directly inputting the second covariate serving as multi-dimensional influence factor historical data into a second prediction module;
s403: inputting historical data of natural gas consumption and a third covariate into a first neural network model which is constructed in advance for fitting to obtain first temporary natural gas consumption prediction data, and inputting the first temporary natural gas consumption prediction data into a second prediction module;
s404: and inputting the historical data of the natural gas consumption and the data of the future influence factors into a pre-constructed addition model for fitting to obtain second temporary natural gas consumption prediction data, and inputting the second temporary natural gas consumption prediction data into a second prediction module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110590350.8A CN113326983B (en) | 2021-05-28 | 2021-05-28 | Data processing system and method for natural gas consumption prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110590350.8A CN113326983B (en) | 2021-05-28 | 2021-05-28 | Data processing system and method for natural gas consumption prediction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113326983A CN113326983A (en) | 2021-08-31 |
| CN113326983B true CN113326983B (en) | 2022-08-26 |
Family
ID=77421998
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110590350.8A Active CN113326983B (en) | 2021-05-28 | 2021-05-28 | Data processing system and method for natural gas consumption prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113326983B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109919173A (en) * | 2019-01-11 | 2019-06-21 | 国网浙江省电力有限公司宁波供电公司 | A kind of multilist fusion energy behavior analysis method based on subtractive clustering model |
| CN109993364A (en) * | 2019-04-01 | 2019-07-09 | 北京恒华龙信数据科技有限公司 | A kind of prediction technique and device of natural gas gas consumption |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103473605A (en) * | 2013-08-22 | 2013-12-25 | 中国能源建设集团广东省电力设计研究院 | Method and system for predicting energy consumption |
| GB2543281A (en) * | 2015-10-13 | 2017-04-19 | British Gas Trading Ltd | System for energy consumption prediction |
| CN105894113A (en) * | 2016-03-31 | 2016-08-24 | 中国石油天然气股份有限公司规划总院 | Natural gas short-period demand prediction method |
| CN108230049A (en) * | 2018-02-09 | 2018-06-29 | 新智数字科技有限公司 | The Forecasting Methodology and system of order |
| CN109447346B (en) * | 2018-10-26 | 2022-07-05 | 冶金自动化研究设计院 | Converter oxygen consumption prediction method based on gray prediction and neural network combined model |
| CN112686470A (en) * | 2021-01-18 | 2021-04-20 | 国网河北省电力有限公司经济技术研究院 | Power grid saturation load prediction method and device and terminal equipment |
-
2021
- 2021-05-28 CN CN202110590350.8A patent/CN113326983B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109919173A (en) * | 2019-01-11 | 2019-06-21 | 国网浙江省电力有限公司宁波供电公司 | A kind of multilist fusion energy behavior analysis method based on subtractive clustering model |
| CN109993364A (en) * | 2019-04-01 | 2019-07-09 | 北京恒华龙信数据科技有限公司 | A kind of prediction technique and device of natural gas gas consumption |
Non-Patent Citations (1)
| Title |
|---|
| "Evaluation and Forecasting of Gas Consumption by Statistical Analysis";F.B.GORUCU;《Energy Sources》;20100624;第267-276页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113326983A (en) | 2021-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111860979B (en) | Short-term load prediction method based on TCN and IPSO-LSSVM combined model | |
| CN111105332B (en) | Highway intelligent pre-maintenance method and system based on artificial neural network | |
| CN110414788B (en) | Electric energy quality prediction method based on similar days and improved LSTM | |
| CN111260136A (en) | Building short-term load prediction method based on ARIMA-LSTM combined model | |
| CN111563706A (en) | Multivariable logistics freight volume prediction method based on LSTM network | |
| CN110232203B (en) | Knowledge distillation optimization RNN short-term power failure prediction method, storage medium and equipment | |
| Moody | Economic forecasting: Challenges and neural network solutions | |
| CN116757534B (en) | Intelligent refrigerator reliability analysis method based on neural training network | |
| CN110705743A (en) | New energy consumption electric quantity prediction method based on long-term and short-term memory neural network | |
| CN112308288A (en) | Particle swarm optimization LSSVM-based default user probability prediction method | |
| CN111680786B (en) | Time sequence prediction method based on improved weight gating unit | |
| CN114676822B (en) | Multi-attribute fusion air quality forecasting method based on deep learning | |
| CN106980910B (en) | Medium-and-long-term power load measuring and calculating system and method | |
| CN111091241A (en) | BP neural network-based drug sales prediction and decision method and system | |
| CN113408869A (en) | Power distribution network construction target risk assessment method | |
| CN112288137A (en) | LSTM short-term load prediction method and device considering electricity price and Attention mechanism | |
| CN113393057A (en) | Wheat yield integrated prediction method based on deep fusion machine learning model | |
| CN113053113A (en) | PSO-Welsch-Ridge-based anomaly detection method and device | |
| CN115096357A (en) | Indoor environment quality prediction method based on CEEMDAN-PCA-LSTM | |
| CN114154716A (en) | Enterprise energy consumption prediction method and device based on graph neural network | |
| CN113326983B (en) | Data processing system and method for natural gas consumption prediction | |
| CN116703644A (en) | Attention-RNN-based short-term power load prediction method | |
| CN111435471A (en) | Heat supply gas consumption prediction model based on L STM | |
| CN115730717A (en) | Power load prediction method and system based on combination of transfer learning strategy and multiple channels | |
| WO2022222230A1 (en) | Indicator prediction method and apparatus based on machine learning, and device and storage medium |
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 |