CN116542701A - Carbon price prediction method and system based on CNN-LSTM combination model - Google Patents

Carbon price prediction method and system based on CNN-LSTM combination model Download PDF

Info

Publication number
CN116542701A
CN116542701A CN202310356370.8A CN202310356370A CN116542701A CN 116542701 A CN116542701 A CN 116542701A CN 202310356370 A CN202310356370 A CN 202310356370A CN 116542701 A CN116542701 A CN 116542701A
Authority
CN
China
Prior art keywords
carbon
lstm
data
carbon price
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
Application number
CN202310356370.8A
Other languages
Chinese (zh)
Inventor
郭宇辰
加鹤萍
杨争林
刘敦楠
许晓敏
郑亚先
冯树海
曾丹
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Electric Power Research Institute Co Ltd CEPRI
North China Electric Power University
Original Assignee
China Electric Power Research Institute Co Ltd CEPRI
North China Electric Power University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by China Electric Power Research Institute Co Ltd CEPRI, North China Electric Power University filed Critical China Electric Power Research Institute Co Ltd CEPRI
Priority to CN202310356370.8A priority Critical patent/CN116542701A/en
Publication of CN116542701A publication Critical patent/CN116542701A/en
Pending legal-status Critical Current

Links

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
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • G06Q30/0202—Market predictions or forecasting for commercial activities
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • 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
    • 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
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • G06Q30/0206—Price or cost determination based on market factors
    • 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
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/80—Management or planning
    • Y02P90/84—Greenhouse gas [GHG] management systems

Abstract

The invention relates to a carbon price prediction method and a system based on a CNN-LSTM combination model, comprising the following steps: step 1, acquiring historical carbon price time series data, and carrying out normalization processing on the historical carbon price time series data: step 2, inputting the normalized carbon number sequence data into a CNN module, and carrying out convolution calculation on the carbon number sequence data to extract potential feature vectors; step 3, inputting the normalized carbon price data and the extracted feature vector into an LSTM module for training and completing prediction; and 4, carrying out inverse normalization on the result output by the LSTM module to obtain a final prediction result. The method and the device can simultaneously improve the accuracy and the processing speed of the prediction model.

Description

Carbon price prediction method and system based on CNN-LSTM combination model
Technical Field
The invention belongs to the technical field of carbon price prediction, relates to a carbon price prediction method and a carbon price prediction system, and particularly relates to a carbon price prediction method and a carbon price prediction system based on a CNN-LSTM (computer numerical control-based virtual machine modeling) combined model.
Background
In the trade market of carbon emission rights in various countries, carbon prices are in a fluctuation state for a long time, so that decisions and investments related to the carbon market have certain risks. The financial price sequence often has an average regression trend, namely the short-term price fluctuates up and down around a fixed value, so that accurate carbon price prediction is realized, and the market management mechanism is facilitated to realize effective regulation and control of the carbon price and efficient performance of a control and discharge enterprise, so that the method has important significance in predicting the carbon price.
The carbon price historical data can be regarded as a time sequence in mathematics, the time sequence is a set of random variables obtained by observing a certain phenomenon or process according to a given sampling rate in time sequence, the time sequence data can record the historical data and show the change process of things, the time sequence data generally comprises a plurality of potential development rules, and the analysis and prediction of proper influence factors are selected from the time sequence data, so that the analysis and prediction of the things and the summary rules are greatly facilitated for human beings.
Currently, three main types of methods exist to realize the prediction of carbon prices. Firstly, a carbon price prediction method based on a metering model, such as an autoregressive model, has good fitting prediction capability on a data sequence meeting a hypothesis condition, however, because the model is simplified and fixed, and the hypothesis condition is that a time sequence has linear and stable characteristics, the method is contrary to the characteristics of a highly complex carbon price sequence, and therefore, accurate carbon price prediction cannot be realized by carbon market price prediction.
Secondly, in order to solve the nonlinear relation of the price sequence, artificial intelligent models such as an artificial neural network and the like are proposed, hidden relations in the carbon price sequence can be effectively mined, and errors are minimized, however, the problems of local overfitting, complex calculation amount and information loss are easy to occur in the model training process, and the performance of a single model cannot meet the requirement of carbon price prediction precision.
Finally, in order to grasp various characteristics and influencing factors in the carbon price time sequence more comprehensively and clearly, a combined prediction model processing non-stationarity, non-linearity and multi-scale carbon price prediction method is proposed, and a relatively accurate and steady prediction result is obtained.
The convolutional neural network CNN model is a popular method in the deep learning field, can effectively extract data features, and has wider application scenes. The alternating use of the convolution layer and the pooling layer not only can effectively extract potential features of input data, but also can reduce errors caused by manually extracting the features. The calculation process of the pooling layer is simple, each submatrix of the input data set is compressed, and common pooling standards mainly have two kinds: maximum pooling and average pooling, namely taking the maximum value or average value of the corresponding area as the element value after pooling.
The long-term memory neural network LSTM model is an improved model based on a cyclic neural network (RNN), and can solve the problems of gradient explosion, gradient disappearance and difficulty in long-term storage of historical data in a long-time sequence training process.
However, there is a limitation in the conventional single model, for example, the CNN model may ignore the correlation with the whole in the pooling process, so that the training result converges to the local minimum value instead of the global minimum value. And the LSTM model of the long-term memory neural network needs to process huge calculation amount in processing long-term sequences.
Therefore, the invention provides a carbon price prediction method and a system based on a CNN-LSTM combined model, and the accuracy and the processing speed of the prediction model can be improved simultaneously by improving or combining different models.
No published patent documents identical or similar to the present invention are found upon searching.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, and provides a carbon price prediction method and a system based on a CNN-LSTM combined model, which can simultaneously improve the accuracy and the processing speed of the prediction model.
The invention solves the practical problems by adopting the following technical scheme:
a carbon price prediction method based on a CNN-LSTM combination model comprises the following steps:
step 1, acquiring historical carbon price time series data, and carrying out normalization processing on the historical carbon price time series data:
step 2, inputting the normalized carbon number sequence data into a CNN module, and carrying out convolution calculation on the carbon number sequence data to extract potential feature vectors;
step 3, inputting the normalized carbon price data and the extracted feature vector into an LSTM module for training and completing prediction;
and 4, carrying out inverse normalization on the result output by the LSTM module to obtain a final prediction result.
The specific method of the step 1 is as follows:
the linear conversion normalization processing adopts a maximum and minimum normalization method to map the original carbon number sequence data to intervals [0,1], so that the processing speed of the model on the carbon number sequence is accelerated, and the calculation process is as follows:
wherein x is historical carbon price time series data, min represents the minimum value of the historical carbon price time series data, and max represents the maximum value of the historical carbon price time series data.
Moreover, the specific steps of the step 2 include:
(1) Setting the number of convolution kernels to be 3;
(2) Setting the step length of a pooling layer to be 1;
(3) Setting the number of the combined layers of the convolution layers to be 2-4, collecting key data from a time sequence through the convolution layers, and extracting data characteristics;
(4) And utilizing a pooling layer to prevent overfitting, setting a full connection layer to integrate the convolved features, outputting the classification condition and probability of the vector features, and connecting the classification condition and probability to an LSTM layer.
Moreover, the specific steps of the step 3 include:
(1) Setting the layer number of the LSTM model;
(2) Setting an optimizer trained by an LSTM module as an Adam optimizer;
(3) And outputting the processing result of the model through the full connection layer.
A carbon price prediction system based on a CNN-LSTM combination model, comprising:
the data processing module is used for acquiring historical carbon price time sequence data and carrying out normalization processing on the historical carbon price time sequence data:
the feature vector extraction module is used for inputting the normalized carbon price sequence data into the CNN module, and carrying out convolution calculation on the carbon price sequence data to extract potential features;
and the model training module is used for inputting the normalized carbon price data and the extracted feature vectors into the LSTM module for training and completing prediction.
And the prediction result output module is used for carrying out inverse normalization on the result output by the LSTM module to obtain a final prediction result.
Moreover, the data processing module is further configured to:
the maximum and minimum normalization method is adopted to carry out linear conversion normalization processing, original carbon price sequence data are mapped to intervals [0,1], the processing speed of the model on the carbon price sequence is accelerated, and the calculation process is as follows:
wherein x is historical carbon price time series data, min represents the minimum value of the historical carbon price time series data, and max represents the maximum value of the historical carbon price time series data.
The calculation method in the step 4 is as follows:
x”＝y' (max-min) +min
wherein y' is a normalized price prediction value, and x "is an inverse normalized actual price prediction value.
The invention has the advantages and beneficial effects that:
1. the invention provides a carbon price prediction method and a system based on a CNN-LSTM combination model, which are further improved on the basis of a traditional artificial neural network, so that the problem of gradient explosion or information deletion possibly occurring in the traditional method when a time sequence is processed is effectively solved, and the processing capacity of information is enhanced while the complexity of an algorithm is reduced.
2. The invention provides a carbon price prediction method and a system based on a CNN-LSTM combination model, which attempt to maximize and comprehensively utilize the information of a carbon price time sequence by aggregating and optimizing the information processed by the two models, so that the utilization rate of carbon price data is improved to reduce the prediction risk of the model in the unknown future, and the reliability of the prediction performance of the model is improved.
3. According to the carbon price prediction method and system based on the CNN-LSTM combined model, the extraction capacity of the data features is enhanced through the combined model, the change of the carbon price sequence can be analyzed in the time dimension, the relation between the data features is considered, and therefore the accuracy of the prediction model is improved.
4. The invention provides a carbon price prediction method and a system based on a CNN-LSTM combination model, which realize parallel learning through local weight sharing of a neural network, reduce complexity of feature extraction and classification, thereby simplifying calculation process and improving model training speed.
5. The invention provides a carbon price prediction method and a system based on a CNN-LSTM combination model, which can better learn the dependency relationship in data by selectively reserving or deleting the historical time sequence by using a gating mechanism, thereby ensuring the accuracy of prediction.
6. According to the LSTM carbon price prediction method and system combined with the convolutional neural network, information features are deeply mined through a modularized algorithm, errors caused by manually extracting the features are greatly reduced, a prediction model is optimized, and prediction accuracy is further improved.
Drawings
FIG. 1 is a European energy exchange carbon price historical data set of the present invention;
fig. 2 is a schematic diagram of the internal unit structure of the convolutional neural network CNN model of the present invention;
FIG. 3 is a schematic diagram of the internal unit structure of the LSTM model of the long-term and short-term memory neural network according to the present invention;
FIG. 4 is a schematic diagram of a CNN-LSTM combined model according to the present invention;
FIG. 5 is a graph comparing the multi-model predictions of the present invention.
Detailed Description
Embodiments of the invention are described in further detail below with reference to the attached drawing figures:
a carbon price prediction method based on a CNN-LSTM combination model comprises the following steps:
step 1, acquiring historical carbon price time series data, and carrying out normalization processing on the historical carbon price time series data:
the specific method of the step 1 is as follows:
the linear conversion normalization processing adopts a maximum and minimum normalization method to map the original carbon number sequence data to intervals [0,1], so that the processing speed of the model on the carbon number sequence is accelerated, and the calculation process is as follows:
wherein x is historical carbon price time series data, min represents the minimum value of the historical carbon price time series data, and max represents the maximum value of the historical carbon price time series data.
Step 2, inputting the normalized carbon number sequence data into a CNN module, and carrying out convolution calculation on the carbon number sequence data to extract potential feature vectors;
the specific steps of the step 2 include:
(1) Setting the number of convolution kernels to be 3;
the number of convolution kernels of the convolution neural network unit is set to be y, and the convolution kernels are generally set to be 3 because the convolution kernels have to be larger than 1 to have the effect of improving the receptive field, and the convolution kernels with even numbers cannot guarantee that the characteristic value size is unchanged.
(2) Setting the step length of a pooling layer to be 1;
the step size of the pooling layer is n, and since the pooling operation scans an input tensor by using one matrix window, and the value in each window reduces the number of elements by taking the maximum value or taking the average value, the step size determines how many matrix windows will be used, if the step size is 1, each matrix window will be used, if the step size is 2, every 1 matrix window in each dimension will be used, and in general, in order to make full use of data information, the step size is set to 1.
The convolution calculation process is as follows:
wherein N is the number of input matrixes and X k Represents the kth input matrix, W k The kth sub-convolution kernel matrix, representing the convolution kernel, s (i, j), represents the values of the corresponding position elements of the corresponding output matrix of the convolution kernel W.
(3) Setting the number of the combined layers of the convolution layers to be 2-4, collecting key data from a time sequence through the convolution layers, and extracting data characteristics;
the combination of the convolution layer and the pooling layer in the model is set to x times in the hidden layer and is used for removing interference and noise information in sample data, and a full-connection layer is arranged at the back of the combination, so that influence of characteristic positions on classification is reduced. The more the number of combined layers, the heavier the calculation task of the model in training, and the number of combined layers is generally set to be 2 to 4.
(4) And utilizing a pooling layer to prevent overfitting, setting a full connection layer to integrate the convolved features, outputting the classification condition and probability of the vector features, and connecting the classification condition and probability to an LSTM layer.
And step 3, inputting the normalized carbon price data and the extracted feature vector into an LSTM module for training and completing prediction.
The specific steps of the step 3 include:
(1) Setting the layer number of the LSTM model;
when the test data is sufficiently large, the non-linear fitting capacity and prediction accuracy of the model can be enhanced by increasing the hidden layer number of the LSTM module, but the calculation amount of the model can be increased, and even the problem of overfitting is caused, so that the LSTM layer number is usually set to be 2-4 layers, and meanwhile, the performance of the model is influenced by the number of neurons of each layer, so that each layer is usually selected to be gradually increased from using 2 or 3 neurons.
(2) Setting an optimizer trained by an LSTM module as an Adam optimizer;
the batch_size is set to 64 and the optimization function is prevented from falling into a local optimum by randomly inactivating neurons. The iteration number of the model is set to be 50 in the initial training, and the optimal value of the model parameter is determined by continuously adjusting the model parameter according to the training set result.
(3) Outputting a processing result of the model, namely predicted future carbon price trend, through the full-connection layer;
step 4, carrying out inverse normalization on the result output by the LSTM module to obtain a final prediction result;
x”＝y' (max-min) +min
wherein y' is a normalized price prediction value, and x "is an inverse normalized actual price prediction value.
In this embodiment, the specific calculation process of the LSTM module is as follows:
1) Forgetting the stage: forget about the door forgets useless history information. This stage mainly controls the input data of the last state node:
f t ＝σ(W f ·[h t-1 ,x t ]+b f )
i t ＝σ(W i ·(x t ,h t-1 )+b i )
wherein f t 、i t Respectively representing a forgetting gate and an input gate, W f 、W i And b f 、b i The weight matrix and the offset of each gate are respectively represented, and sigma represents the Sigmoid function.
2) Updating: selectively recording input information
C t ＝tanh(W c ·(x t ,h t-1 )+b c )
U t ＝i t ·C t +f t ·C t-1
Wherein U is t And C · Representing the cell state and the tanh control gate, respectively, tanh being the activation function.
The training time is prolonged by considering too many hidden models, so that two LSTM layers are set to perform associated prediction processing on data, and a method of randomly discarding neurons is adopted between the LSTM layers to prevent overfitting.
3) Output stage: it is decided what information is to be output as the current state.
O t ＝σ(W o ·(x t ,h t-1 )+b o )
h t ＝O t ·tanh(U t )
Wherein h is t Represents hidden layer output, O t Representing an output gate.
In the embodiment, the difference between the predicted value and the actual true value of various models is calculated by adopting root mean square error and mean square error, so that the effectiveness and the accuracy of the model are proved.
Since Root Mean Square Error (RMSE) is more sensitive to outliers in the results, the difference between the predicted and true values is measured using this calculation method:
wherein y is c Is the actual value of the c-th sample point, y c' Is the predicted value of the c-th sample, and N is the number of sample points.
Wherein the Mean Square Error (MSE) can be used to evaluate the degree of variation of the resulting error:
a carbon price prediction system based on a CNN-LSTM combination model, comprising:
the data processing module is used for acquiring historical carbon price time sequence data and carrying out normalization processing on the historical carbon price time sequence data:
the feature vector extraction module is used for inputting the normalized carbon price sequence data into the CNN module, and carrying out convolution calculation on the carbon price sequence data to extract potential features;
and the model training module is used for inputting the normalized carbon price data and the extracted feature vectors into the LSTM module for training and completing prediction.
And the prediction result output module is used for carrying out inverse normalization on the result output by the LSTM module to obtain a final prediction result.
The data processing module is further configured to:
the maximum and minimum normalization method is adopted to carry out linear conversion normalization processing, original carbon price sequence data are mapped to intervals [0,1], the processing speed of the model on the carbon price sequence is accelerated, and the calculation process is as follows:
wherein x is historical carbon price time series data, min represents the minimum value of the historical carbon price time series data, and max represents the maximum value of the historical carbon price time series data.
The invention is further illustrated by the following examples:
first, the carbon price history data is obtained to create a raw data set, as shown in fig. 1, which is the carbon price data of the european energy exchange in recent years, specifically, the carbon price data during the european energy exchanges 2018.1.8 to 2022.2.28. Then, the time sequence data are divided according to the proportion of 2:1, and the first two thirds of data are used as training sets of adjustment parameters during model learning and training; the latter third of the dataset is used as a test set for evaluating the accuracy and effectiveness of model predictions, does not participate in the training process, and then the data is normalized.
And secondly, constructing a convolutional neural network and a long-short-term memory neural network by using Python, and constructing a CNN model structure and an LSTM model structure by relying on a Tensorflow bottom framework. The CNN model mainly comprises a convolution layer, a pooling layer and a full-connection layer, and is shown in fig. 2, wherein the features are extracted by mainly carrying out convolution calculation on data, and finally, local features are combined into global features; the LSTM model is shown in fig. 3, with a specially designed gate structure to enhance the ability of the neuron to select information.
In this embodiment, the convolution kernel of the CNN module is set to 3, the step size of the pooling layer is 1, the number of combinations of the pooling layer and the convolution layer is 2, and the maximum pooling is used to reduce the output dimension size through feature mapping in the convolution layer. The test result of the training set shows that the model performance is optimal when the number of LSTM layers in the LSTM module is 2 and the number of neurons in each layer is 64.
Then, a combination and superposition of a deep learning link library Keras and a Pandas support model are used for constructing a CNN-LSTM combination prediction model structure shown in fig. 4, an original carbon price sequence is firstly input into a CNN module, the CNN extracts the characteristics of variables in a time sequence through a convolution kernel, variable characteristic results are pooled and fused through a pooling layer to reduce output dimensions, information in advance of the CNN module is input into an LSTM module, time sequence characteristic information is screened again through a forgetting stage, prediction is carried out through the LSTM layer, and finally a prediction result of the model is output.
Finally, the prediction results of running each model are shown in fig. 5, which shows the prediction results of test sets of different models, and the prediction results of the CNN-LSTM model can be obtained more accurately by comparing the prediction results of various models with the true values. The error checking analysis is shown in table 1. Compared with other models, the CNN-LSTM model has the advantages that the fitting effect is optimal, the root mean square error in the test set is 2.88, the prediction capability is strong when the data change sharply, and other models have larger hysteresis.
TABLE 1
Therefore, compared with a plurality of models, the method has good performance, can realize more accurate carbon price prediction results, provides references for carbon price prediction research, and is beneficial to establishing a perfect market mechanism. After accurate prediction of the carbon price is realized by adopting the CNN-LSTM model, as the carbon price represents the income or risk caused by carbon emission of operation activities of producers and investors, planning and deployment can be performed in advance according to the future carbon price prediction trend, so that the national energy structure adjustment and the industrial structure upgrading are accelerated, the ecological environment protection is enhanced, the carbon-peak carbon neutralization target is realized early, the comprehensive national force of China is improved, and the state of China is consolidated.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

Claims (7)

1. A carbon price prediction method based on a CNN-LSTM combination model is characterized by comprising the following steps of: the method comprises the following steps:
step 1, acquiring historical carbon price time series data, and carrying out normalization processing on the historical carbon price time series data:
step 2, inputting the normalized carbon number sequence data into a CNN module, and carrying out convolution calculation on the carbon number sequence data to extract potential feature vectors;
step 3, inputting the normalized carbon price data and the extracted feature vector into an LSTM module for training and completing prediction;
and 4, carrying out inverse normalization on the result output by the LSTM module to obtain a final prediction result.
2. The carbon price prediction method based on the CNN-LSTM combination model of claim 1, which is characterized by comprising the following steps: the specific method of the step 1 is as follows:
the linear conversion normalization processing adopts a maximum and minimum normalization method to map the original carbon number sequence data to intervals [0,1], so that the processing speed of the model on the carbon number sequence is accelerated, and the calculation process is as follows:
wherein x is historical carbon price time series data, min represents the minimum value of the historical carbon price time series data, and max represents the maximum value of the historical carbon price time series data.
3. The carbon price prediction method based on the CNN-LSTM combination model of claim 1, which is characterized by comprising the following steps: the specific steps of the step 2 include:
(1) Setting the number of convolution kernels to be 3;
(2) Setting the step length of a pooling layer to be 1;
(3) Setting the number of the combined layers of the convolution layers to be 2-4, collecting key data from a time sequence through the convolution layers, and extracting data characteristics;
(4) And utilizing a pooling layer to prevent overfitting, setting a full connection layer to integrate the convolved features, outputting the classification condition and probability of the vector features, and connecting the classification condition and probability to an LSTM layer.
4. The carbon price prediction method based on the CNN-LSTM combination model of claim 1, which is characterized by comprising the following steps: the specific steps of the step 3 include:
(1) Setting the layer number of the LSTM model;
(2) Setting an optimizer trained by an LSTM module as an Adam optimizer;
(3) And outputting the processing result of the model through the full connection layer.
5. A carbon price prediction system based on a CNN-LSTM combination model is characterized in that: comprising the following steps:
the data processing module is used for acquiring historical carbon price time sequence data and carrying out normalization processing on the historical carbon price time sequence data:
the feature vector extraction module is used for inputting the normalized carbon price sequence data into the CNN module, and carrying out convolution calculation on the carbon price sequence data to extract potential features;
and the model training module is used for inputting the normalized carbon price data and the extracted feature vectors into the LSTM module for training and completing prediction.
And the prediction result output module is used for carrying out inverse normalization on the result output by the LSTM module to obtain a final prediction result.
6. The CNN-LSTM combination model-based carbon price prediction system of claim 5, wherein: the data processing module is further configured to:
the maximum and minimum normalization method is adopted to carry out linear conversion normalization processing, original carbon price sequence data are mapped to intervals [0,1], the processing speed of the model on the carbon price sequence is accelerated, and the calculation process is as follows:
wherein x is historical carbon price time series data, min represents the minimum value of the historical carbon price time series data, and max represents the maximum value of the historical carbon price time series data.
7. The CNN-LSTM combination model-based carbon price prediction system of claim 5, wherein: the calculation method of the prediction result output module comprises the following steps:
x”＝y' (max-min) +min
wherein y' is a normalized price prediction value, and x "is an inverse normalized actual price prediction value.
CN202310356370.8A 2023-04-06 2023-04-06 Carbon price prediction method and system based on CNN-LSTM combination model Pending CN116542701A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310356370.8A CN116542701A (en) 2023-04-06 2023-04-06 Carbon price prediction method and system based on CNN-LSTM combination model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310356370.8A CN116542701A (en) 2023-04-06 2023-04-06 Carbon price prediction method and system based on CNN-LSTM combination model

Publications (1)

Publication Number Publication Date
CN116542701A true CN116542701A (en) 2023-08-04

Family

ID=87442497

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310356370.8A Pending CN116542701A (en) 2023-04-06 2023-04-06 Carbon price prediction method and system based on CNN-LSTM combination model

Country Status (1)

Country Link
CN (1) CN116542701A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117251295A (en) * 2023-11-15 2023-12-19 成方金融科技有限公司 Training method, device, equipment and medium of resource prediction model

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117251295A (en) * 2023-11-15 2023-12-19 成方金融科技有限公司 Training method, device, equipment and medium of resource prediction model
CN117251295B (en) * 2023-11-15 2024-02-02 成方金融科技有限公司 Training method, device, equipment and medium of resource prediction model

Similar Documents

Publication Publication Date Title
Zhang et al. At-lstm: An attention-based lstm model for financial time series prediction
CN110909926A (en) TCN-LSTM-based solar photovoltaic power generation prediction method
CN110084610B (en) Network transaction fraud detection system based on twin neural network
CN111967343B (en) Detection method based on fusion of simple neural network and extreme gradient lifting model
CN111563706A (en) Multivariable logistics freight volume prediction method based on LSTM network
CN112990556A (en) User power consumption prediction method based on Prophet-LSTM model
CN111814956B (en) Multi-task learning air quality prediction method based on multi-dimensional secondary feature extraction
CN111626785A (en) CNN-LSTM network fund price prediction method based on attention combination
CN110264270A (en) A kind of behavior prediction method, apparatus, equipment and storage medium
CN110956309A (en) Flow activity prediction method based on CRF and LSTM
CN112288137A (en) LSTM short-term load prediction method and device considering electricity price and Attention mechanism
CN112363896A (en) Log anomaly detection system
CN111222689A (en) LSTM load prediction method, medium, and electronic device based on multi-scale temporal features
CN111985719A (en) Power load prediction method based on improved long-term and short-term memory network
CN113095484A (en) Stock price prediction method based on LSTM neural network
CN116542701A (en) Carbon price prediction method and system based on CNN-LSTM combination model
Urgun et al. Composite system reliability analysis using deep learning enhanced by transfer learning
CN115308558A (en) Method and device for predicting service life of CMOS (complementary Metal oxide semiconductor) device, electronic equipment and medium
CN116703464A (en) Electric automobile charging demand modeling method and device, electronic equipment and storage medium
CN113449919B (en) Power consumption prediction method and system based on feature and trend perception
CN117557304B (en) Electric quantity and electricity price level fusion prediction method based on modal decomposition and neural network
Urgun et al. Composite power system reliability evaluation using importance sampling and convolutional neural networks
CN114118401A (en) Neural network-based power distribution network flow prediction method, system, device and storage medium
CN113033898A (en) Electrical load prediction method and system based on K-means clustering and BI-LSTM neural network
Bidyuk et al. An Approach to Identifying and Filling Data Gaps in Machine Learning Procedures

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