CN117251705A - Daily natural gas load prediction method - Google Patents
Daily natural gas load prediction method Download PDFInfo
- Publication number
- CN117251705A CN117251705A CN202311283186.1A CN202311283186A CN117251705A CN 117251705 A CN117251705 A CN 117251705A CN 202311283186 A CN202311283186 A CN 202311283186A CN 117251705 A CN117251705 A CN 117251705A
- Authority
- CN
- China
- Prior art keywords
- natural gas
- data
- gas load
- load data
- prediction
- 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.)
- Pending
Links
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 title claims abstract description 192
- 239000003345 natural gas Substances 0.000 title claims abstract description 96
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 27
- 238000010606 normalization Methods 0.000 claims abstract description 16
- 238000012545 processing Methods 0.000 claims abstract description 15
- 230000002159 abnormal effect Effects 0.000 claims abstract description 13
- 238000007781 pre-processing Methods 0.000 claims abstract description 11
- 238000007689 inspection Methods 0.000 claims abstract description 4
- 238000005457 optimization Methods 0.000 claims description 10
- 238000012549 training Methods 0.000 claims description 9
- 238000004458 analytical method Methods 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 3
- 230000007774 longterm Effects 0.000 abstract description 3
- 238000013136 deep learning model Methods 0.000 abstract 1
- 238000004364 calculation method Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 239000011159 matrix material Substances 0.000 description 5
- 238000005065 mining Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000012300 Sequence Analysis Methods 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000007789 gas Substances 0.000 description 1
- 230000002779 inactivation Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- General Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Marketing (AREA)
- Computing Systems (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Water Supply & Treatment (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a daily natural gas load prediction method, which comprises the following steps: designing a natural gas historical load data and characteristic selection preprocessing module, wherein the processing of abnormal values, missing values and repeated values of the load and characteristic data is completed by the module, the stability and randomness inspection is carried out, and the data normalization is completed; designing a natural gas load data decomposition and feature selection module, wherein the module decomposes natural gas original data into a plurality of different subsequences, optimizes the decomposition number and then performs feature selection on each subsequence; and designing a natural gas load data prediction module, wherein the module predicts different subsequences, and carries out signal reconstruction on the prediction result to finally obtain a natural gas load data prediction result of the second day. According to the method, the natural gas data with high complexity and nonlinear instability are decomposed and deeply mined, long-term dependence of the sequence is built through a deep learning model, and the prediction accuracy of the natural gas load is improved.
Description
Technical Field
The invention belongs to the field of time sequence analysis and energy, and particularly relates to a daily natural gas load prediction method.
Background
Along with the influence of global climate change on human living environment, more and more countries begin to pay attention to the development of low-carbon green energy, and natural gas serves as a green clean energy and plays a key supporting role in a clean energy system. The increase in natural gas usage requires gas companies to accurately predict the natural gas consumption for different time periods in time. The model for predicting natural gas consumption is mainly divided into the following three types: traditional models, artificial intelligence models, hybrid models. The traditional model is difficult to process natural gas consumption with nonlinear characteristics, and can only make long-term prediction on natural gas consumption, but cannot make short-term prediction. The artificial intelligence model improves the processing power of nonlinear data, but the generalization power and the interpretability of the artificial intelligence model are still to be improved. The single algorithm is difficult to avoid defects when aiming at specific problems, so that the mixed model optimizes the algorithm by combining different algorithms, and the problems of the single prediction algorithm are overcome. The currently commonly used sequence decomposition methods mainly comprise Wavelet Transform (WT), empirical Mode Decomposition (EMD) and VMD, and compared with other models, the VMD can effectively avoid the predicted delay phenomenon. The transform model was proposed by google in 2017, which shows a strong modeling capability for time series data in the NLP field, and more students use it in time series data prediction.
Disclosure of Invention
In order to solve the technical problems, the invention provides a daily natural gas load prediction method, which covers the design of a natural gas historical load data and characteristic selection preprocessing module, a natural gas load data decomposition and characteristic selection module and a natural gas load data prediction module, can realize the preprocessing and deep characteristic mining of data under the condition of higher complexity of load data, and respectively predicts each subsequence after decomposition to obtain a predicted value, thereby realizing the accurate prediction of daily natural gas load data.
In order to achieve the above purpose, the invention adopts the following technical scheme:
a daily natural gas load prediction method comprising the steps of:
step (1), designing a natural gas historical load data and characteristic data preprocessing module, completing the processing of abnormal values, missing values and repeated values of the load and characteristic data, performing stability and randomness inspection, and completing data normalization, wherein the method comprises the following steps:
the method comprises the steps of (1.1) processing abnormal values, missing values and repeated values of data, namely firstly removing the detected repeated values, calculating a standard deviation sigma of natural gas load data, then detecting the abnormal values according to a 3 sigma principle, deleting the abnormal values, finally processing the missing values by adopting an interpolation filling method, and taking the average value of the front value and the rear value for filling;
step (1.2) checking the stability and randomness of the natural gas load data, adopting an ADF checking method to perform stability checking on the data, judging a non-stable sequence if the result is more than 0, constructing Q statistics to perform randomness checking on the natural gas load data, and judging a non-pure random sequence if the checking result is less than 0.05, so as to prove that the natural gas load data has analysis and prediction significance;
step (1.3) carrying out normalization processing on the natural gas load data and the characteristic data, adopting the maximum normalization, mapping the data with different dimensions between [0-1], and accelerating the training speed of the follow-up neural network;
step (2), designing a natural gas decomposition and feature selection module, decomposing the preprocessed natural gas load data into a plurality of different subsequences, optimizing the decomposition number, and then performing feature selection on each subsequence, wherein the method comprises the following steps:
decomposing the natural gas load data by adopting variation modal decomposition, analyzing and predicting the natural gas load data after decomposing the natural gas load data into a plurality of different subsequences, and reducing the complexity of the natural gas load data;
step (2.2) optimizing the number of subsequences by designing a feature matching degree maximization optimization algorithm, circulating natural gas original data in a range from 2 to the feature number, respectively decomposing and calculating the feature matching degree, comparing the feature matching degree after the circulation is finished, and taking the maximum value as an optimization result;
step (2.3) respectively combining the decomposed subsequences with characteristic data to calculate pearson correlation coefficients, wherein the calculated result is a correlation characteristic greater than 0.3;
step (3), designing a natural gas load data prediction module, firstly predicting different subsequences, and carrying out signal reconstruction on the prediction result to finally obtain a natural gas load data prediction result on the second day, wherein the method comprises the following steps:
step (3.1) constructing an improved converter model to predict sub-sequences, replacing an original Decoder module of the converter model with a full connection layer to obtain a converter prediction model, respectively training the converter prediction model according to the decomposed sub-sequences, and predicting each sub-sequence;
and (3.2) reconstructing the prediction results of each subsequence to obtain a final natural gas load prediction result on the second day, and realizing accurate prediction of the natural gas load data on each day.
Further, the method is suitable for natural gas load data of different cities, regions and enterprises.
Compared with the prior art, the invention has the beneficial effects that:
(1) The current optimization scheme aiming at the decomposition quantity is mostly based on a heuristic search algorithm, has low optimization efficiency and is not suitable for a prediction algorithm. The feature matching degree maximization optimization algorithm designed by the invention can make the feature quantity of each decomposed subsequence as small as possible and close to one, and is beneficial to improving the training speed and accuracy of the subsequent deep learning algorithm.
(2) When the city or region is located in a subtropical zone or even a tropical zone, the relevance between the natural gas load data and the air temperature is not great, the influence factors are difficult to evaluate, the load data has nonlinear and unstable characteristics, and the data analysis and the feature mining are difficult to effectively perform. In the existing researches, most algorithms aim at data with large temperature correlation in temperate regions, and the prediction accuracy is high. But there is a significant drop in accuracy when the algorithm is applied to subtropical or tropical regions. According to the method, the original data is decomposed to reduce the data complexity, an improved transducer model is designed to improve the long-term dependence modeling capability of the subsequence, and finally, higher prediction precision can be realized for different regions.
In summary, the method comprises the steps of preprocessing the natural gas load historical data, decomposing the original sequence by using a modal decomposition method, designing a decomposition quantity optimization algorithm, realizing analysis and feature mining of high-complexity data, designing an improved transducer model, predicting the subsequence, synthesizing signals, and realizing accurate prediction of the natural gas load data in the next day.
Drawings
FIG. 1 is a block diagram of a daily natural gas load prediction method of the present invention;
FIG. 2 is a flowchart of a VMD algorithm;
FIG. 3 is a flow chart of a feature matching degree maximization optimization algorithm;
fig. 4 is a block diagram of a transducer model.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
As shown in fig. 1, the invention relates to a daily natural gas load prediction method, which comprises the steps of designing a natural gas historical load data and characteristic selection preprocessing module 1, designing a natural gas load data decomposition and characteristic selection module 2 and designing a natural gas load data prediction module 3, and can realize preprocessing of data and deep feature mining under the condition of higher complexity of load data, and respectively predict each subsequence after decomposition to obtain a predicted value so as to realize accurate prediction of daily natural gas load data.
As shown in fig. 1, a daily natural gas load prediction method of the present invention includes the steps of:
step (1) designs a natural gas historical load data and characteristic data preprocessing module 1, which completes the processing of abnormal values, missing values and repeated values of load and characteristic data, performs stability and randomness inspection, completes data normalization, and specifically realizes the following steps:
(1) aiming at the problem that the quality of the natural gas load data collected by the sensor cannot be directly analyzed, abnormal values, missing values and repeated values of the data are processed. Firstly, removing the detected repeated values, calculating the standard deviation sigma of the natural gas load data, then detecting abnormal values according to the 3 sigma principle, deleting the abnormal values, finally adopting an interpolation filling method to process the missing values, and taking the average value of the front value and the rear value for filling;
(2) the method comprises the steps of checking the stability and randomness of natural gas load data, adopting an ADF checking method to perform stability checking on the data, judging a non-stable sequence if a result is larger than 0, constructing Q statistics to perform randomness checking on the natural gas load data according to chi-square distribution with degree of freedom s, judging a non-pure random sequence if a checking result is smaller than 0.05, and proving that the data has analysis and prediction significance;
(3) carrying out normalization processing on the natural gas load data and the characteristic data, adopting the maximum normalization, mapping the data with different dimensions between [0-1], and accelerating the training speed of the follow-up neural network;
wherein x is load data, min and max are maximum and minimum values in the load data, respectively, x * Is the load data after normalization processing.
Step (2) designs a natural gas decomposition and feature selection module 2, which decomposes natural gas raw data into a plurality of different subsequences, optimizes the decomposition number, and then performs feature selection on each subsequence, wherein the method is specifically implemented as follows:
(1) aiming at the characteristics of high complexity and nonlinear instability of the natural gas load data, the natural gas load data is decomposed by adopting a Variation Mode Decomposition (VMD), and is analyzed and predicted after being decomposed into a plurality of different subsequences, so that the complexity of the data can be effectively reduced, and the VMD algorithm flow is shown in figure 2. The VMD core idea is to construct and solve the variation problem, firstly construct the variation problem, assume that the natural gas load data is decomposed into a plurality of components, the component sequence is a modal component with limited bandwidth of the center frequency, meanwhile, the sum of the estimated bandwidths of all modes is minimum, the constraint condition is that the sum of all modes is equal to the original signal, introduce a Lagrange multiplier to convert the constraint variation problem into the non-constraint variation problem, continuously update the modal component and the center frequency until the convergence condition is met, and decompose the natural gas load data into a plurality of modes;
(2) the number of sequences of VMD decomposition is manually selected, so that the optimization algorithm for maximizing the feature matching degree is designed to optimize the number of sequences of natural gas decomposition, and the algorithm flow is shown in figure 3. Firstly, selecting characteristics affecting natural gas load, initializing the number of decomposition sequences to be 2, decomposing load data, calculating pearson correlation coefficients between subsequences and the characteristics, carrying out normalization processing, enlarging the gap between the maximum value and the minimum value, then calculating the characteristic matching degree according to the proportion of the maximum correlation coefficient in the subsequences to the sum of all correlation coefficients, increasing the number of decomposition from 2 to the number of characteristics, and sequentially carrying out the characteristic matching degree calculation, wherein the number of decomposition corresponding to the obtained maximum characteristic matching degree is the number of optimized decomposition sequences;
wherein u is i Is a subsequence obtained by decomposing load data, Y j Is characteristic data ρ ij Is the pearson correlation coefficient between the subsequence and the feature, P k Is the feature matching value of each sequence, m is the number of features, n is the amount of data each feature contains, i is the index of the sub-sequence, q is the index of the data in the sub-sequence, j is the index of the feature,the (q) th data representing the (i) th group of subsequences>Mean value of the subsequence of group i, +.>The q-th data representing the j-th group of features, < >>Mean value representing the j-th group of features, softmax is normalized exponential function, z ij The normalized pearson correlation coefficient of the ith group of subsequences and the jth group of features is shown, max () is a maximum function, and k is the number of subsequences for decomposing the natural gas load data.
(3) And respectively combining the decomposed subsequences with characteristic data to calculate the pearson correlation coefficient, wherein the calculated result is more than 0.3 and is the correlation characteristic.
Step (3) designing a natural gas load data prediction module 3, firstly predicting different subsequences, and carrying out signal reconstruction on the prediction result to finally obtain a natural gas load data prediction result of the second day, wherein the method is specifically realized as follows:
(1) constructing an improved converter model to predict sub-sequences, replacing an original Decoder module of the converter model by a full connection layer to obtain a converter prediction model, respectively training the converter prediction model according to the decomposed sub-sequences, and predicting each sub-sequence;
the structure of the transducer model is shown in fig. 4, firstly, the position coding layer divides the natural gas load data into different input sequences and adds position information, then the encoder performs multi-head attention calculation in the encoder to generate a tensor with a dimension (window size, batch size and vector dimension), then the tensor is subjected to residual connection, layer normalization and feedforward connection, and meanwhile, random inactivation treatment is performed on the tensor, and finally the decoder consists of a layer of full-connection layer to generate a final prediction result.
Compared with the traditional cyclic neural network and convolutional neural network, the transducer model provides a brand-new position coding mechanism for capturing time sequence information between input data, and the principle is that sine functions and cosine functions with different frequencies are added into an input sequence after normalization as position codes. The position coding formula is as follows:
where pos is the length of the index sequence, i is the index of the dimension, d model Is the word vector dimension and PE represents the position of the sequence.
The self-attention mechanism can be seen as establishing the interaction between different forms of input vectors in a linear projection space. The core process is that attention weight is obtained through calculation of a query matrix Q and a key value matrix K, and then the attention weight is acted on a value matrix V to obtain weight and output. For the input Q, K and V, the calculation formula of the output vector is as follows:
wherein T represents transpose of matrix, d k Representing the word embedding dimension, attention () represents the Attention calculation value.
The multi-head attention mechanism is to process multiple groups of self-attention processing on the original input sequence, and splice the self-attention results of each group together. The calculation formula is as follows:
MultiHead(Q,K,V)＝Concat(head 1 ,head 2 ,...,head h )W O
wherein head i A value representing the attention of the i-th group,respectively representing linear transformation matrixes corresponding to the ith group of attention Q, K, V, multiHead () representing a multi-head attention mechanism output value, concat () representing that a plurality of groups of attention are spliced, W O Representing the linear transformation matrix during the stitching process.
The encoder mainly consists of multi-head attention, residual connection and layer normalization and feedforward connection. Multiple head attention may avoid introducing future information. The residual connection is mainly used for solving the problem of multi-layer network training. The layer normalization makes the input mean variance of each layer of neurons consistent, and accelerates convergence. The feed forward connection is a two-layer fully connected layer. The activation function of the first layer is Relu and is not used for the second time. The original decoder structure is replaced by a linear layer, and the predicted value of the natural gas consumption is directly output. Doing so can reduce training parameters by about half of the scale, avoiding overfitting and reducing the cumulative error of predictions to some extent.
(2) Reconstructing the prediction results of each subsequence to obtain a final natural gas load prediction result on the second day, and realizing accurate prediction of the natural gas load data on each day.
In summary, the method comprises the steps of designing a natural gas historical load data and characteristic selection preprocessing module, designing a natural gas load data decomposition and characteristic selection module and designing a natural gas load data prediction module, preprocessing data and deep characteristic mining can be realized under the condition of higher complexity of the load data, and predicting each decomposed subsequence to obtain a predicted value, so that accurate prediction of daily natural gas load data is realized.
What is not described in detail in the present specification belongs to the prior art known to those skilled in the art.
The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.
Claims (2)
1. A method of daily natural gas load prediction comprising the steps of:
step (1), designing a natural gas historical load data and characteristic data preprocessing module, completing the processing of abnormal values, missing values and repeated values of the load and characteristic data, performing stability and randomness inspection, and completing data normalization, wherein the method comprises the following steps:
the method comprises the steps of (1.1) processing abnormal values, missing values and repeated values of data, namely firstly removing the detected repeated values, calculating a standard deviation sigma of natural gas load data, then detecting the abnormal values according to a 3 sigma principle, deleting the abnormal values, finally processing the missing values by adopting an interpolation filling method, and taking the average value of the front value and the rear value for filling;
step (1.2) checking the stability and randomness of the natural gas load data, adopting an ADF checking method to perform stability checking on the data, judging a non-stable sequence if the result is more than 0, constructing Q statistics to perform randomness checking on the natural gas load data, and judging a non-pure random sequence if the checking result is less than 0.05, so as to prove that the natural gas load data has analysis and prediction significance;
step (1.3) carrying out normalization processing on the natural gas load data and the characteristic data, adopting the maximum normalization, mapping the data with different dimensions between [0-1], and accelerating the training speed of the follow-up neural network;
step (2), designing a natural gas decomposition and feature selection module, decomposing the preprocessed natural gas load data into a plurality of different subsequences, optimizing the decomposition number, and then performing feature selection on each subsequence, wherein the method comprises the following steps:
decomposing the natural gas load data by adopting variation modal decomposition, analyzing and predicting the natural gas load data after decomposing the natural gas load data into a plurality of different subsequences, and reducing the complexity of the natural gas load data;
step (2.2) optimizing the number of subsequences by designing a feature matching degree maximization optimization algorithm, circulating natural gas original data in a range from 2 to the feature number, respectively decomposing and calculating the feature matching degree, comparing the feature matching degree after the circulation is finished, and taking the maximum value as an optimization result;
step (2.3) respectively combining the decomposed subsequences with characteristic data to calculate pearson correlation coefficients, wherein the calculated result is a correlation characteristic greater than 0.3;
step (3), designing a natural gas load data prediction module, firstly predicting different subsequences, and carrying out signal reconstruction on the prediction result to finally obtain a natural gas load data prediction result on the second day, wherein the method comprises the following steps:
step (3.1), constructing an improved converter model to predict sub-sequences, replacing an original Decoder module of the converter model with a full connection layer to form a converter prediction model, respectively training the converter prediction model according to the decomposed sub-sequences, and predicting each sub-sequence;
and (3.2) reconstructing the prediction results of each subsequence to obtain a final natural gas load prediction result on the second day, and realizing accurate prediction of the natural gas load data on each day.
2. A daily natural gas load prediction method as claimed in claim 1, wherein the method is applicable to natural gas load data of different cities, regions and enterprises.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311283186.1A CN117251705A (en) | 2023-10-07 | 2023-10-07 | Daily natural gas load prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311283186.1A CN117251705A (en) | 2023-10-07 | 2023-10-07 | Daily natural gas load prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117251705A true CN117251705A (en) | 2023-12-19 |
Family
ID=89129052
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311283186.1A Pending CN117251705A (en) | 2023-10-07 | 2023-10-07 | Daily natural gas load prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117251705A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117808175A (en) * | 2024-03-01 | 2024-04-02 | 南京信息工程大学 | Short-term multi-energy load prediction method based on DTformer |
-
2023
- 2023-10-07 CN CN202311283186.1A patent/CN117251705A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117808175A (en) * | 2024-03-01 | 2024-04-02 | 南京信息工程大学 | Short-term multi-energy load prediction method based on DTformer |
| CN117808175B (en) * | 2024-03-01 | 2024-05-17 | 南京信息工程大学 | DTformer-based short-term multi-energy load prediction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112990556A (en) | User power consumption prediction method based on Prophet-LSTM model | |
| CN111242377B (en) | Short-term wind speed prediction method integrating deep learning and data denoising | |
| CN107292446B (en) | Hybrid wind speed prediction method based on component relevance wavelet decomposition | |
| CN113177633A (en) | Deep decoupling time sequence prediction method | |
| CN114611792A (en) | Atmospheric ozone concentration prediction method based on mixed CNN-Transformer model | |
| CN117251705A (en) | Daily natural gas load prediction method | |
| CN111832825A (en) | Wind power prediction method and system integrating long-term and short-term memory network and extreme learning machine | |
| CN112884236B (en) | Short-term load prediction method and system based on VDM decomposition and LSTM improvement | |
| CN112668611B (en) | Kmeans and CEEMD-PE-LSTM-based short-term photovoltaic power generation power prediction method | |
| CN112434891A (en) | Method for predicting solar irradiance time sequence based on WCNN-ALSTM | |
| CN113420868A (en) | Traveling salesman problem solving method and system based on deep reinforcement learning | |
| CN115204035A (en) | Generator set operation parameter prediction method and device based on multi-scale time sequence data fusion model and storage medium | |
| CN111141879B (en) | Deep learning air quality monitoring method, device and equipment | |
| CN115034430A (en) | Carbon emission prediction method, device, terminal and storage medium | |
| CN116702831A (en) | Hybrid short-term wind power prediction method considering massive loss of data | |
| CN114676645B (en) | Non-stationary time sequence prediction method and system | |
| Xinxin et al. | Short-term wind speed forecasting based on a hybrid model of ICEEMDAN, MFE, LSTM and informer | |
| Li et al. | Short-term wind power prediction based on modal reconstruction and CNN-BiLSTM | |
| CN113076545A (en) | Deep learning-based kernel fuzzy test sequence generation method | |
| Murata et al. | Automatic CNN compression system for autonomous driving | |
| CN115809725A (en) | Multi-factor short-term electric quantity prediction method and device | |
| CN115730716A (en) | Method for predicting medium-term and long-term power consumption of communication base station based on improved Transformer model | |
| CN115081551A (en) | RVM line loss model building method and system based on K-Means clustering and optimization | |
| CN114386666A (en) | Wind power plant short-term wind speed prediction method based on space-time correlation | |
| CN114139783A (en) | Wind power short-term power prediction method and device based on nonlinear weighted combination |
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 |