CN114595639A - Atmospheric pollutant diffusion numerical simulation method based on deep learning - Google Patents

Atmospheric pollutant diffusion numerical simulation method based on deep learning Download PDF

Info

Publication number
CN114595639A
CN114595639A CN202210316067.0A CN202210316067A CN114595639A CN 114595639 A CN114595639 A CN 114595639A CN 202210316067 A CN202210316067 A CN 202210316067A CN 114595639 A CN114595639 A CN 114595639A
Authority
CN
China
Prior art keywords
neural network
deep neural
space
training
sets
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
CN202210316067.0A
Other languages
Chinese (zh)
Inventor
李鸣野
张耀华
李小明
张畦霖
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.)
Taiyuan Zecheng Information Technology Co ltd
Original Assignee
Taiyuan Zecheng Information Technology Co ltd
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 Taiyuan Zecheng Information Technology Co ltd filed Critical Taiyuan Zecheng Information Technology Co ltd
Priority to CN202210316067.0A priority Critical patent/CN114595639A/en
Publication of CN114595639A publication Critical patent/CN114595639A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling
    • 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
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Abstract

The invention provides a numerical simulation method for atmospheric pollutant diffusion based on deep learning, which comprises the following steps: carrying out data cleaning, and selecting N sets of data sets from monitoring data of all space points collected for multiple times in a specific area, wherein N is an integer and is more than or equal to 10; randomly distributing N sets of data sets into a group of training sets and a group of verification sets; constructing a deep neural network; training a deep neural network by using a training set and a validation set for cross validation; and predicting concentration values of all unknown space points in a specific area by using the trained deep neural network, and inserting the predicted concentration values of all the unknown space points into a space grid of the specific area to construct a refined smooth concentration field. The atmospheric pollutant diffusion numerical simulation method based on deep learning improves the smoothness and the fitting degree of spatial interpolation data.

Description

Atmospheric pollutant diffusion numerical simulation method based on deep learning
Technical Field
The invention relates to the technical field of atmospheric environment, in particular to a spatial interpolation method which is constructed based on a deep neural network and used for numerical simulation of pollutant diffusion in the atmospheric environment.
Background
Atmospheric environmental pollution is a complex phenomenon, and the concentration of atmospheric pollutants at different times and places is influenced by a plurality of factors, the concentration of the atmospheric pollutants is different, and the air quality is greatly different. At present, an environment quality monitoring station monitors atmospheric environment conditions by monitoring limited discrete space point data and processing the monitoring data, and the processed result often cannot evaluate the air quality of a monitoring area, so that a numerical simulation method is utilized to establish a space grid, simulated space points are added, space grid interpolation is carried out, and space distribution is carried out on atmospheric environment pollution monitoring data. The current spatial grid interpolation can only consider spatial autocorrelation factors, attribute similarity of continuous variation of atmospheric pollutants in space and time is difficult to finely consider, the attribute similarity is limited by various conditions, and smoothness and fitting degree of spatial grid interpolation data need to be further improved.
Disclosure of Invention
In order to improve smoothness and fitting degree of spatial interpolation data in numerical simulation of pollutant diffusion in atmospheric environment in the prior art, the invention provides a numerical simulation method of atmospheric pollutant diffusion based on deep learning.
The technical scheme of the invention is realized as follows:
a numerical simulation method for atmospheric pollutant diffusion based on deep learning comprises the following steps: carrying out data cleaning, and selecting N sets of data sets from monitoring data of all space points collected for multiple times in a specific area, wherein N is an integer and is more than or equal to 10; randomly distributing the N sets of data sets into a group of training sets and a group of verification sets; constructing a deep neural network; training a deep neural network by using a training set and a validation set for cross validation; predicting concentration values of all unknown space points in a specific area by using the trained deep neural network, inserting the predicted concentration values of all unknown space points into a space grid of the specific area, and constructing a refined smooth concentration field;
the specific method for constructing the deep neural network is as follows: defining eight attribute input values of an input layer of any space point, wherein the eight attribute input values are the wind speed, the wind direction, the temperature, the precipitation, the terrain and the distance from a most similar building of the space point, the included angle between the position of the most similar building and the space point and the wind direction of the space point, the distance or the time interval between the space point and another space point which is most adjacent to the space point, carrying out normalization pretreatment on the eight attribute input values, establishing a first group of fully-connected layers with more than two layers, carrying out tensor expansion on the eight attribute input values after the normalization pretreatment, carrying out nonlinear activation treatment on the output results of the first group of fully-connected layers on a first layer activation layer, establishing a second group of fully-connected layers with more than two layers, inputting the output results of the first layer activation layer into the second group of fully-connected layers to carry out tensor reduction and reduce to 1 parameter value, and carrying out activation treatment on the output results of the second group of fully-connected layers on a second layer activation layer, the second layer of activation layer outputs the weight concentration value, and the output layer outputs the predicted concentration value corresponding to the space point.
Preferably, the specific way of training the deep neural network by using the training set and the validation set to cross-validate is as follows: inputting the eight attribute values of each space point of the training set into a deep neural network for training, and obtaining a concentration weight value corresponding to each space point after training; calculating a predicted concentration value of the space point according to the concentration weight value of the space point and an actual concentration value of another space point close to the space point; comparing the actual concentration value and the predicted concentration value of each space point in the training set, and optimally training the deep neural network through a back propagation mechanism of the deep neural network; and inputting the eight attribute values of one space point on the verification set into the optimized deep neural network for cross verification every time of optimization, namely comparing the difference value between the actual concentration value and the predicted concentration value of the space on the verification set, and finishing the training of the deep neural network when the difference value is minimum.
Preferably, when the deep neural network is constructed, the first group of fully-connected layers are two fully-connected layers, and the two fully-connected layers firstly expand the eight attribute values to 32 parameter values and then expand the 32 parameter values to 128 parameter values; the second group of full connection layers is set as four layers of full connection layers, and 128 parameter values processed by the first layer of activation layer are firstly reduced to 64 parameter values, then 64 parameter values are reduced to 16 parameter values, then 16 parameter values are reduced to 4 parameter values, and finally 4 parameter values are reduced to 1 parameter value.
Preferably, the first layer active layer performs nonlinear activation processing by using an activated function Rectified Linear Unit, and the second layer active layer performs activation processing of a (1, 0) interval by using an activated function SoftMax.
Preferably, the specific way of predicting the concentration values of all unknown spatial points in a specific area by using the trained deep neural network is as follows: a concentration diffusion field with a spatial grid structure is arranged in a specific area, collected spatial points and unknown spatial points are spatial points on a spatial grid, whether all the unknown spatial points are predicted by a deep neural network is observed, the unknown spatial points which are not predicted are predicted, and prediction of all the unknown spatial points is completed in a recursion manner; when the predicted concentration value of the unknown spatial point is predicted, the actual concentration value of another spatial point closest to the unknown spatial point can be selected to acquire the actual concentration value of the spatial point, and the predicted concentration value of the unknown spatial point which has already been predicted can also be selected.
Preferably, when the deep neural network is optimally trained through a back propagation mechanism of the deep neural network, a cross entropy loss function is selected for optimal training.
Preferably, the data in the training set occupies ninety percent of the N sets of data sets, and the data in the validation set occupies ten percent of the N sets of data sets.
The invention has the beneficial effects that: according to the atmospheric pollutant diffusion numerical simulation method based on deep learning, the predicted concentration value of the unknown space point in the specific passing area is output by constructing and training a deep neural network, then the predicted concentration value of the unknown space point is interpolated into the space grid of the specific area, and the smoothness and the fitting degree of the interpolation data of the space grid are improved. The method can carry out flexible self-defined interpolation aiming at the atmospheric pollution diffusion field of the multi-scale environment, and carry out interpolation value compensation aiming at irregular space grids or space grids with data loss. Diffusion concentration points with time continuity in the concentration diffusion field realize space-time smooth interpolation compensation by utilizing the time precedence relationship of adjacent space points; and diffusing concentration points with position continuity in the concentration diffusion field, and realizing the spatial smooth interpolation compensation by using the position relation of adjacent space points. The prediction accuracy can be continuously updated and updated iteratively according to the expansion of the training set by utilizing the deep neural network, so that a self-iterative optimization interpolation method is formed; the deep neural network has stronger generalization and can carry out interpolation prediction on interpolation fields of different types according to the diffusion characteristics of different pollutants.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a frame structure diagram of a numerical simulation method for atmospheric pollutant diffusion based on deep learning according to the present invention;
fig. 2 is a schematic structural diagram of the deep neural network constructed as shown in fig. 1.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example (b): as shown in fig. 1, a numerical simulation method for atmospheric pollutant diffusion based on deep learning includes: carrying out data cleaning, and selecting N sets of data sets from monitoring data of all space points collected for multiple times in a specific area, wherein N is an integer and is more than or equal to 10; randomly distributing the N sets of data sets into a group of training sets and a group of verification sets; constructing a deep neural network; training a deep neural network by using a training set and a validation set for cross validation; and predicting concentration values of all unknown space points in a specific area by using the trained deep neural network, and inserting the predicted concentration values of all the unknown space points into a space grid of the specific area to construct a refined smooth concentration field.
As shown in fig. 2, for the diffusion concentration points with position continuity in the concentration diffusion field, the specific way of constructing the deep neural network is as follows: defining eight attribute input values as the wind speed, wind direction, temperature, precipitation, terrain of any space point, the included angle between the most similar building and the space point and the wind direction of the space point, the distance between the space point and another space point which is most adjacent to the space point, namely the position relationship of adjacent space points, normalizing the eight attribute input values, establishing a first group of fully-connected layers with more than two layers, carrying out tensor expansion on the eight attribute input values after the normalization pretreatment, carrying out nonlinear activation on the output results of the first group of fully-connected layers on a first layer of active layers, establishing two second groups of fully-connected layers, inputting the output results of the first group of active layers to the second group of fully-connected layers for tensor reduction and reduction to 1 parameter value, carrying out activation on the output results of the second group of fully-connected layers on a second layer of active layers, the second layer of activation layer outputs the weight concentration value, and the output layer outputs the predicted concentration value corresponding to the space point.
For the diffusion concentration points with time continuity in the concentration diffusion field, the specific method for constructing the deep neural network is as follows: defining eight attribute input values as the wind speed, wind direction, temperature, precipitation, terrain of any space point, the included angle between the position of the most similar building and the space point and the wind direction of the space point, the time interval between the space point and the other space point which is most adjacent to the space point, namely the time sequence relation of the adjacent space points, carrying out normalization pretreatment on the eight attribute input values, establishing a first group of fully-connected layers with more than two layers, carrying out tensor expansion on the eight attribute input values after the normalization pretreatment, carrying out nonlinear activation treatment on the output results of the first group of fully-connected layers on a first layer activation layer, establishing two second groups of fully-connected layers, inputting the output results of the first layer activation layer to the second group of fully-connected layers for tensor reduction and reduction to 1 parameter value, carrying out activation treatment on the output results of the second group of fully-connected layers on the second layer activation layer, the second layer of activation layer outputs the weight concentration value, and the output layer outputs the predicted concentration value corresponding to the space point.
The complement of the spatio-temporal smooth interpolation for the same spatial locus point A is ensured to be continuous on the simulation time axis, e.g. for a diffusion simulation process with a temporal resolution of 1 hour interval, T0The time is the beginning of simulation, and the concentration of the corresponding point A is B0；T1The time is diffusion simulation lasting 1 hour, and the concentration of the corresponding point A is B1；T2The time is diffusion simulation lasting 2 hours, which corresponds to the concentration B of point A2. The time point of the space-time interpolation value compensation carried out by the method of the invention is positioned at T1And T2Time T in betweena，TaCan be set to any value between 0 and 1 hour according to requirements, and the concentration result B can be obtained through training and prediction by the methodaAnd by analogy, the concentration interpolation compensation of the point A at all unknown moments in the simulation process can be completed.
When the deep neural network is constructed, the first group of fully connected layers are set to be two fully connected layers, the eight attribute values are expanded to 32 parameter values by the two fully connected layers, and then the 32 parameter values are expanded to 128 parameter values by the two fully connected layers; the second group of full connection layers firstly reduces 128 parameter values processed by the first layer of activation layer through an activation function Rectified Linear Unit (ReLU) to 64 parameter values, then reduces 64 parameter values to 16 parameter values, then reduces 16 parameter values to 4 parameter values, finally reduces 4 parameter values to 1 parameter value, and inputs 1 parameter value to the second layer of activation layer to perform activation processing of a (1, 0) interval through an activation function SoftMax.
The specific mode of training the deep neural network by using the training set and the verification set for cross validation is as follows: inputting the eight attribute values of each space point of the training set into a deep neural network for training, and obtaining a concentration weight value corresponding to each space point after training; calculating a predicted concentration value of the space point according to the concentration weight value of the space point and an actual concentration value of another space point close to the space point; comparing the actual concentration value and the predicted concentration value of each space point in the training set, and optimally training the deep neural network by selecting a cross entropy loss function through a back propagation mechanism of the deep neural network; and inputting the eight attribute values of one space point on the verification set into the optimized deep neural network for cross verification every time of optimization, namely comparing the difference value between the actual concentration value and the predicted concentration value of the space on the verification set, and finishing the training of the deep neural network when the difference value is minimum.
The specific mode for predicting the concentration values of all unknown space points in a specific area by using the trained deep neural network is as follows: a concentration diffusion field with a spatial grid structure is arranged in a specific area, collected spatial points and unknown spatial points are spatial points on a spatial grid, whether all the unknown spatial points are predicted by a deep neural network is observed, the unknown spatial points which are not predicted are predicted, and prediction of all the unknown spatial points is completed in a recursion manner; when the predicted concentration value of the unknown space point is predicted, the actual concentration value of another nearest space point can be selected to acquire the actual concentration value of the space point, or the predicted concentration value of the unknown space point which has already been predicted can be selected, and the acquired space point is the space point in the training set and the verification set.
The data in the training set occupies ninety percent of the N sets of data, and the data in the validation set occupies ten percent of the N sets of data.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (7)

1. A numerical simulation method of atmospheric pollutant diffusion based on deep learning is characterized by comprising the following steps: carrying out data cleaning, and selecting N sets of data sets from monitoring data of all space points collected for multiple times in a specific area, wherein N is an integer and is more than or equal to 10; randomly distributing the N sets of data sets into a group of training sets and a group of verification sets; constructing a deep neural network; training a deep neural network by using a training set and a validation set for cross validation; predicting concentration values of all unknown space points in a specific area by using the trained deep neural network, inserting the predicted concentration values of all unknown space points into a space grid of the specific area, and constructing a refined smooth concentration field;
the specific method for constructing the deep neural network is as follows: defining eight attribute input values as the wind speed, wind direction, temperature, precipitation, terrain and distance of the nearest building to the space point, the included angle between the position of the nearest building and the space point and the wind direction of the space point, the distance or time interval between the space point and another space point nearest to the space point, normalizing the eight attribute input values, establishing a first group of fully-connected layers with more than two layers, carrying out tensor expansion on the eight attribute input values after the normalization, carrying out nonlinear activation processing on the output results of the first group of fully-connected layers on a first layer activation layer, establishing a second group of fully-connected layers with more than two layers, inputting the output results of the first layer activation layer to the second group of fully-connected layers for tensor reduction and reduction to 1 parameter value, and carrying out activation processing on the output results of the second group of fully-connected layers on a second layer activation layer, the second layer of activation layer outputs the weight concentration value, and the output layer outputs the predicted concentration value corresponding to the space point.
2. The atmospheric pollutant diffusion numerical simulation method based on deep learning of claim 1, wherein the deep neural network is trained by cross validation of a training set and a validation set in a specific manner that: inputting the eight attribute values of each space point of the training set into a deep neural network for training, and obtaining a concentration weight value corresponding to each space point after training; calculating a predicted concentration value of the space point according to the concentration weight value of the space point and an actual concentration value of another space point close to the space point; comparing the actual concentration value and the predicted concentration value of each space point in the training set, and optimally training the deep neural network through a back propagation mechanism of the deep neural network; and inputting the eight attribute values of one space point on the verification set into the optimized deep neural network for cross verification every time of optimization, namely comparing the difference value between the actual concentration value and the predicted concentration value of the space on the verification set, and finishing the training of the deep neural network when the difference value is minimum.
3. The numerical simulation method for atmospheric pollutant diffusion based on deep learning of claim 1, wherein when the deep neural network is constructed, the first group of fully-connected layers is two fully-connected layers, and the two fully-connected layers firstly expand eight attribute values to 32 parameter values and then expand the 32 parameter values to 128 parameter values; the second group of full connection layers is set as four layers of full connection layers, and 128 parameter values processed by the first layer of activation layer are firstly reduced to 64 parameter values, then 64 parameter values are reduced to 16 parameter values, then 16 parameter values are reduced to 4 parameter values, and finally 4 parameter values are reduced to 1 parameter value.
4. The numerical simulation method for atmospheric pollutant diffusion based on deep learning of claim 1, wherein the first layer of active layer is activated nonlinearly by using activated function Rectified Linear Unit, and the second layer of active layer is activated for interval (1, 0) by using activated function SoftMax.
5. The numerical simulation method for atmospheric pollutant diffusion based on deep learning of claim 2 is characterized in that the specific way of predicting the concentration values of all unknown spatial points in a specific area by using the trained deep neural network is as follows: a concentration diffusion field with a spatial grid structure is arranged in a specific area, collected spatial points and unknown spatial points are spatial points on a spatial grid, whether all the unknown spatial points are predicted by a deep neural network is observed, the unknown spatial points which are not predicted are predicted, and prediction of all the unknown spatial points is completed in a recursion manner; when the predicted concentration value of the unknown spatial point is predicted, the actual concentration value of another spatial point closest to the unknown spatial point can be selected to acquire the actual concentration value of the spatial point, and the predicted concentration value of the unknown spatial point which has already been predicted can also be selected.
6. The atmospheric pollutant diffusion numerical simulation method based on deep learning of claim 2 is characterized in that when the deep neural network is optimally trained through a back propagation mechanism of the deep neural network, a cross entropy loss function is selected for optimal training.
7. The deep learning-based numerical simulation method of atmospheric pollutant diffusion according to claim 2, characterized in that the data in the training set occupy ninety percent of the N sets of data sets, and the data in the validation set occupy ten percent of the N sets of data sets.
CN202210316067.0A 2022-03-29 2022-03-29 Atmospheric pollutant diffusion numerical simulation method based on deep learning Pending CN114595639A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210316067.0A CN114595639A (en) 2022-03-29 2022-03-29 Atmospheric pollutant diffusion numerical simulation method based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210316067.0A CN114595639A (en) 2022-03-29 2022-03-29 Atmospheric pollutant diffusion numerical simulation method based on deep learning

Publications (1)

Publication Number Publication Date
CN114595639A true CN114595639A (en) 2022-06-07

Family

ID=81809845

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210316067.0A Pending CN114595639A (en) 2022-03-29 2022-03-29 Atmospheric pollutant diffusion numerical simulation method based on deep learning

Country Status (1)

Country Link
CN (1) CN114595639A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116448966A (en) * 2023-06-15 2023-07-18 吉林省中农阳光数据有限公司 Air quality assessment method based on combination of intelligent Internet of things and deep learning

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116448966A (en) * 2023-06-15 2023-07-18 吉林省中农阳光数据有限公司 Air quality assessment method based on combination of intelligent Internet of things and deep learning
CN116448966B (en) * 2023-06-15 2023-09-12 吉林省中农阳光数据有限公司 Air quality assessment method based on combination of intelligent Internet of things and deep learning

Similar Documents

Publication Publication Date Title
Liang et al. A novel wind speed prediction strategy based on Bi-LSTM, MOOFADA and transfer learning for centralized control centers
CN109214575B (en) Ultrashort-term wind power prediction method based on small-wavelength short-term memory network
CN110059878B (en) Photovoltaic power generation power prediction model based on CNN LSTM and construction method thereof
JP5888640B2 (en) Photovoltaic power generation prediction apparatus, solar power generation prediction method, and solar power generation prediction program
CN110084610B (en) Network transaction fraud detection system based on twin neural network
CN111079931A (en) State space probabilistic multi-time-series prediction method based on graph neural network
CN113361801A (en) Hybrid rolling short-term load prediction method based on GAN network and semi-supervised regression
CN109784562B (en) Smart power grid power load prediction method based on big data space-time clustering
CN114595639A (en) Atmospheric pollutant diffusion numerical simulation method based on deep learning
CN111105005B (en) Wind power prediction method
CN114694379B (en) Traffic flow prediction method and system based on self-adaptive dynamic graph convolution
CN115051358A (en) Wind power generation short-term load power prediction method and wind power generation system
Xing et al. Comparative study of reformed neural network based short‐term wind power forecasting models
CN114039659A (en) High-order dispersion prediction calculation method for optical fiber communication
CN113269421A (en) Atmospheric pollution traceability calculation method based on automatic monitoring station data and genetic algorithm
Chen et al. Spatial-temporal cellular traffic prediction: A novel method based on causality and graph attention network
CN111797979A (en) Vibration transmission system based on LSTM model
Shang et al. Predicting Hourly $\mathbf {PM} _ {2.5} $ Concentrations Based on Random Forest and Ensemble Neural Network
CN116722529A (en) Short-term photovoltaic power prediction method and system
CN115640845A (en) Method for generating few-category samples of neural network of graph based on generation of confrontation network
CN115619028A (en) Clustering algorithm fusion-based power load accurate prediction method
CN115796327A (en) Wind power interval prediction method based on VMD (vertical vector decomposition) and IWOA-F-GRU (empirical mode decomposition) -based models
CN112752290B (en) Method and equipment for predicting data traffic of wireless base station
CN113449920A (en) Wind power prediction method, system and computer readable medium
CN113191526A (en) Short-term wind speed interval multi-objective optimization prediction method and system based on random sensitivity

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