CN105740991B - Climate change prediction method and system based on improved BP neural network fitting of multiple climate modes - Google Patents

Climate change prediction method and system based on improved BP neural network fitting of multiple climate modes Download PDF

Info

Publication number
CN105740991B
CN105740991B CN201610109283.2A CN201610109283A CN105740991B CN 105740991 B CN105740991 B CN 105740991B CN 201610109283 A CN201610109283 A CN 201610109283A CN 105740991 B CN105740991 B CN 105740991B
Authority
CN
China
Prior art keywords
climate
prediction
rainfall
data
neural network
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201610109283.2A
Other languages
Chinese (zh)
Other versions
CN105740991A (en
Inventor
钟平安
吴业楠
朱非林
徐斌
李天成
付吉斯
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.)
Hohai University HHU
Original Assignee
Hohai University HHU
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 Hohai University HHU filed Critical Hohai University HHU
Priority to CN201610109283.2A priority Critical patent/CN105740991B/en
Publication of CN105740991A publication Critical patent/CN105740991A/en
Application granted granted Critical
Publication of CN105740991B publication Critical patent/CN105740991B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/10—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
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
    • Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping
    • 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 discloses a climate change prediction method and a climate change prediction system based on fitting of a plurality of climate modes by an improved BP neural network, wherein the method comprises the following steps: collecting rainfall data representing climate change characteristics; carrying out downscaling processing on the climate mode prediction grid data to obtain long series data with the same resolution as the actually measured data; carrying out mean value correction on the weather mode prediction series after the size reduction; establishing an improved BP neural network model and determining a network model structure; training and checking the established BP neural network model; and predicting the future climate space-time change by using the model output. The method improves the input and output data processing of the BP neural network, and fits a plurality of climate modes to predict the future climate change, thereby increasing the reliability of the prediction result, and making up the defects that the deviation is larger in the conventional single mode during the evaluation, the nonuniformity of the basin space is not considered during the multi-mode set evaluation, and the like.

Description

Climate change prediction method and system based on improved BP neural network fitting of multiple climate modes
Technical Field
The invention belongs to the field of climatology and hydrology water resource application, and particularly relates to a method for predicting future climate change by using a climate mode.
Background
Future climate prediction against the background of climate change is one of the major scientific problems of global climate change. The climate change trend in the future is objectively evaluated, the water resource adaptability management is enhanced, and the trend is towards interest and avoiding harm, which is a major strategic need of the country for coping with climate change. Global climate patterns are considered as the primary tool for understanding and attributing past climate changes and making predictions of the future. Therefore, it is very important to select a mode with high performance, whether the mode can reasonably simulate past climate change or not will directly influence the mode to evaluate the current climate change 'reproduction ability' or not, and the reliability of prediction of future climate change can be checked to a certain extent.
At present, methods for predicting the future by using climate modes are mainly divided into two types: one mode is used for forecasting, simulation data and actual measurement data of different climate modes are used for carrying out control tests and evaluating the simulation effect of the modes on the average state, the historical climate is known, and the climate mode with good simulation effect is screened out to forecast the future climate; and the second mode is multi-mode set prediction, wherein the climate mode is weighted by using a subjective weighting method or an objective weighting method according to single-mode performance to obtain the multi-mode set, and the future climate change is predicted. However, when the single mode is used for evaluating the climate change trend of a specific drainage basin, the deviation is often large; when the multi-mode set is used, the nonuniformity of a basin space is not considered when one basin uses the same group of weight values, and the space-time distribution condition of climate change cannot be correctly reflected.
In addition, when the BP neural network is used for forecasting, the application of the BP neural network in climate forecasting is limited due to the fact that forecasting sequence values cannot be extended. How to predict future climate change by using the BP neural network and obtain a result with high reliability is also an urgent technical problem to be solved.
Disclosure of Invention
The purpose of the invention is as follows: an object is to provide a climate change prediction method based on fitting of an improved BP neural network to multiple climate modes, so as to solve the above problems in the prior art. It is a further object to provide a system for carrying out the above method.
The technical scheme is as follows: a climate change prediction method based on improved BP neural network fitting of multiple climate modes comprises the following steps:
step 1, collecting rainfall data representing climate change characteristics;
step 2, carrying out downscaling processing on the prediction grid data of various weather modes to obtain long series data with the same resolution as the actually measured data;
step 3, mean value correction is carried out on the weather mode prediction series after the size reduction;
step 4, establishing an improved BP neural network model and determining a network model structure;
step 5, training and checking the established BP neural network model;
and 6, predicting the future climate space-time change by using the model output.
The step 1 is further as follows:
and collecting actual rainfall data series of historical periods, rainfall grid data of historical periods simulated by a plurality of climate modes and rainfall grid data of predicted future periods in the research area.
In a further embodiment, the step 2 is further:
when the scale of the climate mode prediction grid data is reduced, carrying out year-by-year interpolation on the climate mode grid data by adopting an inverse distance weight interpolation method in ArcGIS; and connecting the space amount in the climate mode space distribution with the rainfall station position in the research area by using a spatial join tool in an Analysis Tools module in ArcGIS to obtain climate mode rainfall prediction long series data with the same resolution as the actually measured data.
The step 3 is further as follows:
correcting multiple weather mode prediction sequences of k rainfall stations in the research area one by one, and calculating the average actual rainfall of the jth rainfall station in the historical period for years
Figure GDA0002192518170000022
j is 1, 2 … k, and the mean value of the meteorological element series in the jth rainfall station mode simulation historical period is calculated
Figure GDA0002192518170000023
j is 1, 2 … k, using the difference delta between the two average values to make year-by-year correction for future forecast climate data series,
Figure GDA0002192518170000021
in the formula, YPre-jtPredicted precipitation magnitude for the future t year of the climate pattern; pjtThe corrected value of the rainfall in the t year is obtained.
The step 4 is further as follows:
establishing the BP neural network model, which comprises an input layer, a hidden layer and an output layer; the input layer is a plurality of corrected climate mode prediction series, and the number of nodes is the number of the climate modes; the number of hidden nodes is determined by a Lippmann empirical formula; the output layer is a multi-mode set rainfall prediction series, and the number of nodes is 1;
and improving the input and output data processing of the BP neural network aiming at the data extension condition that the prediction stage possibly exceeds the extreme value of the training sample, and normalizing the input data to (α):
Figure GDA0002192518170000031
wherein the content of the first and second substances,
Figure GDA0002192518170000032
wherein XkFor the input sequence of samples, Xmin、XmaxRespectively a maximum and a minimum, X, in the input sequence of samplesmax、XPremaxMaximum values of training period and prediction period in sample input sequence, akThe input data is normalized;
for desired output TkThe above changes are also made to normalize the data to (α):
Figure GDA0002192518170000033
wherein, Tmin、TmaxMaximum and minimum values in the desired output sequence for the sample, respectively, bkOutputting a normalized value for the expectation;
when the model training is finished and the parameters are fixed, the model is used for prediction, the output value of the model is still between (0, 1), and the model is required to be converted into an actual value;
Figure GDA0002192518170000034
ckand outputting a value for the model actually.
In a further embodiment, the step 5 further comprises:
training and inspecting the BP neural network; selecting an S-shaped function as an excitation function; setting an initial learning rate and momentum items; randomly generating a weight value and a threshold value in the network; and setting a network training end condition.
In a further embodiment, the step 6 is further that:
network modeling and prediction are carried out on each rainfall station to obtain a multi-mode set rainfall prediction series output by each rainfall station neural network in the research area, rainfall station mode prediction values of different time windows are counted, and spatial interpolation is carried out by utilizing an interpolation module in ArcGIS to obtain future climate space-time distribution in the research area.
A climate change prediction system based on improved BP neural network fitting of multiple climate modes mainly comprises the following modules:
a first module for collecting rainfall data characterizing climate change;
the second module is used for carrying out downscaling processing on the prediction grid data of various weather modes to obtain long series data with the same resolution as the actually measured data;
the third module is used for carrying out mean value correction on the multiple weather mode prediction series after the scale reduction;
the fourth module is used for establishing an improved BP neural network model and determining a network model structure;
a fifth module for training and verifying the established BP neural network model;
a sixth module for predicting future spatiotemporal changes in climate using the model output.
The first module is further used for collecting actual measurement rainfall data series of historical periods, rainfall grid data of historical periods simulated by multiple weather modes and rainfall grid data of predicted future periods in the research area;
the second module is further used for carrying out annual interpolation on the climate mode grid data by adopting an inverse distance weight interpolation method in ArcGIS when carrying out downscaling on the climate mode prediction grid data; connecting the spatial amount in the climate mode spatial distribution with the rainfall station position in the research area by using a spatial join tool in an Analysis Tools module in ArcGIS to obtain climate mode forecast rainfall long series data with the same resolution as the actually measured data;
the third module is further used for correcting the climate mode prediction sequences of k rainfall stations in the research area one by one and calculating the average actual rainfall of the jth rainfall station in the historical period for years
Figure GDA0002192518170000041
j is 1, 2 … k, and the mean value of the meteorological element series in the jth rainfall station mode simulation historical period is calculated
Figure GDA0002192518170000042
j is 1, 2 … k, using the difference delta between the two average values to make year-by-year correction for future forecast climate data series,
Figure GDA0002192518170000043
in the formula, YPre-jtPredicted precipitation magnitude for the future t year of the climate pattern; pjtThe corrected value of the rainfall in the t year is obtained.
The fourth module is further used for establishing the BP neural network model and comprises an input layer, a hidden layer and an output layer; the input layer is a plurality of corrected climate mode prediction series, and the number of nodes is the number of the climate modes; the number of hidden nodes is determined by a Lippmann empirical formula; the output layer is a multi-mode set rainfall prediction series, and the number of nodes is 1;
and improving the input and output data processing of the BP neural network aiming at the data extension condition that the prediction stage possibly exceeds the extreme value of the training sample, and normalizing the input data to (α):
Figure GDA0002192518170000044
wherein the content of the first and second substances,
Figure GDA0002192518170000045
wherein XkFor the input sequence of samples, Xmin、XmaxAre respectively as followsMaximum and minimum, X, in the input sequenceTraining max、XPremaxMaximum values of training period and prediction period in sample input sequence, akThe input data is normalized;
for desired output TkThe above changes are also made to normalize the data to (α):
Figure GDA0002192518170000051
wherein, Tmin、TmaxMaximum and minimum values in the desired output sequence for the sample, respectively, bkOutputting a normalized value for the expectation;
when the model training is finished and the parameters are fixed, the model is used for prediction, the output value of the model is still between (0, 1), and the model is required to be converted into an actual value;
Figure GDA0002192518170000052
ckoutputting a value for the model actually;
the fifth module is further for training and testing the BP neural network; selecting an S-shaped function as an excitation function; setting an initial learning rate and momentum items; randomly generating a weight value and a threshold value in the network; and setting a network training end condition, wherein the sixth module is further used for performing network modeling and prediction on each rainfall station to obtain a multi-mode set rainfall prediction series output by each rainfall station neural network in the research area, counting rainfall station mode prediction values of different time windows, and performing spatial interpolation by using an interpolation module in the ArcGIS to obtain future climate space-time distribution in the research area.
Has the advantages that:
firstly, the method carries out control experiment on the simulation data and the actual measurement data of the climate mode and carries out mean value correction, thereby making up the defect that the prediction of a single climate mode is often large in deviation.
Secondly, the good nonlinear mapping capability and the good recognition and prediction capability of the BP neural network are utilized to carry out multi-mode fitting on each rainfall station in the research area, and the defect that the nonuniformity of a basin space is not considered when the same group of weight values are used in the previous research of a basin is overcome.
Finally, the method improves the defect that the data extension capacity exceeding the extreme value of the training sample is insufficient in the BP neural network prediction stage, improves the input and output data processing of the BP neural network, and increases the reliability of predicting the future climate space-time change by a multi-mode set.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Fig. 2 is a rainfall correction graph after mean value correction.
Fig. 3 is a series of charts of improved BP neural network training and testing for multi-mode aggregate rainfall.
FIG. 4 is a spatial distribution diagram of a multi-modal ensemble predicted future climate change.
Detailed Description
In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention. It will be apparent, however, to one skilled in the art, that the present invention may be practiced without one or more of these specific details. In other instances, well-known features have not been described in order to avoid obscuring the invention.
The applicant considers that the climate mode is used for predicting the climate change, the control test is carried out by using the simulation data and the actual measurement data of the climate mode, and the climate mode or the climate mode set with good evaluation effect is selected for future climate prediction. Research finds that in the existing climate mode prediction future climate change scheme, future trend prediction is performed by using a single mode or weighted multi-mode set method, and the problems of large deviation, no consideration of spatial change and the like occur. The applicant believes that it is important to adopt high-reliability mode data in the process of predicting future climate change.
In order to solve the problems in the prior art, the invention provides a climate change prediction method based on fitting of a plurality of climate modes by an improved BP neural network.
The technical solution of the present invention is further specifically described below by way of an embodiment, and with reference to the accompanying drawings and the implementation process of ArcGIS software.
As shown in fig. 1, a climate change prediction method based on fitting of an improved BP neural network to a plurality of climate modes includes the following steps:
step 1, collecting actual rainfall data series of historical periods in a research area, historical rainfall grid data of weather mode prediction and rainfall grid data of future periods, and arranging the data into data of uniform time length.
Step 2, carrying out downscaling processing on the climate mode prediction grid data to obtain long series data with the same resolution as the actually measured data:
establishing an excel file according to the weather mode prediction grid Data, five-column values of serial numbers, years, longitudes, latitudes and rainfall, importing the excel file into ArcGIS, selecting a Tools of Tools → Add XY Data to position a weather mode grid position, generating a weather mode grid Data layer, and generating an actually measured rainfall Data layer according to the actually measured rainfall station Data in the same way. Selecting a Spatial Analysis Tools → Interpolation → an ID W tool to interpolate the climate mode prediction data to obtain Spatial distribution, and then using the Analysis Tools → Overlay → Spatial Join tool to spatially connect the climate mode Spatial distribution with the position of the actual rainfall station to obtain mode prediction rainfall series data of the actual rainfall station.
And 3, performing mean value correction on the reduced-scale climate mode prediction series:
FIG. 2 is a graph showing the result of mean value correction of model prediction data using historical measured data of a rainfall station.
Correcting climate mode prediction sequences of k rainfall stations in the research area one by one, and calculating the average actual rainfall of the jth rainfall station in the historical period for years
Figure GDA0002192518170000071
j is 1, 2 … k, and the average value of the j-th rainfall station mode prediction historical period meteorological element series is calculated
Figure GDA0002192518170000072
j is 1, 2 … k, using the difference delta between the two average values to make year-by-year correction for future forecast climate data series,
Figure GDA0002192518170000073
Pjt＝Ypre-jt+Δ
In the formula, YPre-jtPredicted precipitation magnitude for the future t year of the climate pattern; pjtThe corrected value of the rainfall in the t year is obtained.
Step 4, establishing an improved BP neural network model and determining a network model structure:
establishing a BP neural network model, which comprises an input layer, a hidden layer and an output layer, wherein the input layer is a plurality of corrected climate mode prediction series, the number of nodes is the number of climate modes, the number of hidden layer nodes is determined by a Lippmann empirical formula, the output layer is a multi-mode set rainfall prediction series, the number of nodes is 1, aiming at the situation that data extension exceeding the extreme value of a training sample possibly occurs in the prediction stage, the input and output data processing of the BP neural network is improved, and the input data is normalized to (α):
Figure GDA0002192518170000074
wherein the content of the first and second substances,
Figure GDA0002192518170000075
Xmin、Xmaxrespectively, a maximum and a minimum, X, in the input sequence of samples (including the training and prediction periods)max、XPremaxMaximum values of training period and prediction period in sample input sequence, akThe input data is normalized to a value.
For desired output TkThe above changes are also made to normalize the data to (α):
Figure GDA0002192518170000081
wherein T ismin、TmaxMaximum and minimum values in the desired output sequence for the sample, respectively, bkNormalized values are output for expectations.
When the model training is finished and the parameters are fixed, the model is used for prediction, the output value of the model is still between (0, 1), and the model is required to be converted into an actual value.
Figure GDA0002192518170000082
ckAnd outputting a value for the model actually.
Step 5, training and checking the established BP neural network model:
FIG. 3 is a comparison graph of a multimode aggregate rainfall series and an actual measurement rainfall series simulated by using an improved BP neural network, wherein a historical period is divided into a model training period and a test period, the training period is matched by using actual measurement data and multimode data, and the result of the test period is predicted and compared with the actual measurement data to verify the model accuracy.
Training and inspecting the BP neural network; selecting an S-shaped function (Sigmoid function) as an excitation function; setting an initial learning rate and momentum items; randomly generating a weight value and a threshold value in the network; and setting a network training end condition (error meets the requirement or the training times).
And 6, predicting the future climate space-time change by using the model output:
fig. 4 shows the spatial and temporal distribution of rainfall in the study area obtained by spatial interpolation using the multi-mode set prediction data of each rainfall station. The darker the color in the graph indicates the greater the rainfall in that area.
Network modeling and prediction are carried out on all rainfall stations in a research area one by one to obtain a multi-mode collective rainfall prediction series output by each rainfall station network in the research area, rainfall station mode prediction values in different time windows (years or periods) are counted, and Spatial Interpolation is carried out by utilizing a Spatial Analvsis Tools → Interpolation → ID W Interpolation module in ArcGIS to obtain future climate space-time distribution in the research area.
In summary, aiming at the defects of the prior art, the invention provides a climate change prediction method based on fitting of a plurality of climate modes by an improved BP neural network, which comprises the steps of carrying out scale reduction and mean value correction on climate mode prediction data, establishing an improved BP neural network model, obtaining a multi-climate mode set by taking the plurality of climate mode data as input and carrying out model training and verification, and predicting future climate space-time change. The improved BP neural network input and output data processing method disclosed by the invention can be used for predicting future climate change by fitting various climate modes, so that the reliability of a prediction result is improved, and the defects that the deviation is large in the conventional single mode during evaluation, the nonuniformity of a watershed space is not considered during multi-mode collective evaluation and the like are overcome.
It should be noted that, although part of the steps of the present invention are implemented by using the ArcGIS software in the above-mentioned embodiments, the gist of the present invention is to provide a method for predicting future climate change by using climate mode, the ArcGIS is one of the technical means for implementing the method, and the implementation process of the software can be regarded as an implementation process of an embodiment of the technical idea of the present invention. Of course, similar software may be adopted or developed, and will not be described herein.
In addition, the sequence of the related steps in the present invention is not limited, i.e., it can be adjusted by those skilled in the art, and the sequence in the present invention is written by way of example and not by way of limitation.
Although the preferred embodiments of the present invention have been described in detail, the present invention is not limited to the details of the embodiments, and various equivalent modifications can be made within the technical spirit of the present invention, and the scope of the present invention is also within the scope of the present invention.

Claims (5)

1. A climate change prediction method based on fitting of an improved BP neural network to multiple climate modes is characterized by comprising the following steps:
step 1, collecting rainfall data representing climate change characteristics, comprising:
collecting actual rainfall data series of historical periods in a research area, rainfall grid data of historical periods simulated by various weather modes and rainfall grid data of predicted future periods;
step 2, carrying out downscaling processing on the prediction grid data of various weather modes to obtain long series data with the same resolution as the actually measured data;
step 3, correcting the reduced-scale climate mode prediction series, and the method comprises the following steps:
correcting multiple weather mode prediction sequences of k rainfall stations in the research area one by one, and calculating the average actual rainfall of the jth rainfall station in the historical period for years
Figure FDA0002192518160000011
Calculating the mean value of the meteorological element series in the jth rainfall station mode simulation historical period
Figure FDA0002192518160000012
The difference delta between the two average values is used to carry out year-by-year correction on the climate data series predicted in the future,
Figure FDA0002192518160000013
in the formula, YPre-jtPredicted precipitation magnitude for the future t year of the climate pattern; pjtThe corrected t-year rainfall value;
step 4, establishing an improved BP neural network model and determining a network model structure;
establishing the BP neural network model, which comprises an input layer, a hidden layer and an output layer; the input layer is a plurality of corrected climate mode prediction series, and the number of nodes is the number of the climate modes; the number of hidden nodes is determined by a Lippmann empirical formula; the output layer is a multi-mode set rainfall prediction series, and the number of nodes is 1;
aiming at the data extension condition that the prediction stage is possible to exceed the extreme value of the training sample, the improved BP neural network model improves the input and output data processing of the BP neural network, and input data is normalized to (α):
Figure FDA0002192518160000014
wherein the content of the first and second substances,
Figure FDA0002192518160000015
wherein XkFor the input sequence of samples, Xmin、XmaxRespectively a maximum and a minimum, X, in the input sequence of samplesTraining max、XPremaxMaximum values of training period and prediction period in sample input sequence, akThe input data is normalized;
for desired output TkThe above changes are also made to normalize the data to (α):
Figure FDA0002192518160000021
wherein, Tmin、TmaxMaximum and minimum values in the desired output sequence for the sample, respectively, bkOutputting a normalized value for the expectation;
when the model training is finished and the parameters are fixed, the model is used for prediction, the output value of the model is still between (0, 1), and the model is required to be converted into an actual value;
Figure FDA0002192518160000022
ckoutputting a value for the model actually;
step 5, training and checking the established BP neural network model;
and 6, predicting the future climate space-time change by using the model output.
2. The method of claim 1, wherein the step 2 is further performed by:
when the scale of the climate mode prediction grid data is reduced, carrying out year-by-year interpolation on the climate mode grid data by adopting an inverse distance weight interpolation method in ArcGIS; and connecting the space amount in the climate mode space distribution with the rainfall station position in the research area by using a spatial join tool in an Analysis Tools module in ArcGIS to obtain climate mode rainfall prediction long series data with the same resolution as the actually measured data.
3. The method of claim 1, wherein the method comprises the steps of: the step 5 further comprises the following steps:
training and inspecting the BP neural network; selecting an S-shaped function as an excitation function; setting an initial learning rate and momentum items; randomly generating a weight value and a threshold value in the network; and setting a network training end condition.
4. The method of claim 1, wherein the method comprises the steps of: the step 6 is further as follows:
network modeling and prediction are carried out on each rainfall station to obtain a multi-mode set rainfall prediction series output by each rainfall station neural network in the research area, rainfall station mode prediction values of different time windows are counted, and spatial interpolation is carried out by utilizing an interpolation module in ArcGIS to obtain future climate space-time distribution in the research area.
5. A climate change prediction system for fitting multiple climate modes based on an improved BP neural network is characterized by comprising the following modules:
a first module for collecting rainfall data characterizing climate change;
the second module is used for carrying out downscaling processing on the prediction grid data of various weather modes to obtain long series data with the same resolution as the actually measured data;
the third module is used for carrying out mean value correction on the multiple weather mode prediction series after the scale reduction;
the fourth module is used for establishing an improved BP neural network model and determining a network model structure;
a fifth module for training and verifying the established BP neural network model;
a sixth module for predicting future climate spatiotemporal changes using the model output;
the first module is further used for collecting actual measurement rainfall data series of historical periods, rainfall grid data of historical periods simulated by multiple weather modes and rainfall grid data of predicted future periods in the research area;
the second module is further used for carrying out annual interpolation on the climate mode grid data by adopting an inverse distance weight interpolation method in ArcGIS when carrying out downscaling on the climate mode prediction grid data; connecting the spatial amount in the climate mode spatial distribution with the rainfall station position in the research area by using a spatial join tool in an Analysis Tools module in ArcGIS to obtain climate mode forecast rainfall long series data with the same resolution as the actually measured data;
the third module is further used for correcting the climate mode prediction sequences of k rainfall stations in the research area one by one and calculating the average actual rainfall of the jth rainfall station in the historical period for years
Figure FDA0002192518160000031
Calculating the mean value of the meteorological element series in the jth rainfall station mode simulation historical period
Figure FDA0002192518160000032
The difference delta between the two average values is used to carry out year-by-year correction on the climate data series predicted in the future,
Figure FDA0002192518160000033
in the formula, YPre-jtPredicting future for climate patternsPrecipitation magnitude in the t year; pjtThe corrected t-year rainfall value;
the fourth module is further used for establishing the BP neural network model and comprises an input layer, a hidden layer and an output layer; the input layer is a plurality of corrected climate mode prediction series, and the number of nodes is the number of the climate modes; the number of hidden nodes is determined by a Lippmann empirical formula; the output layer is a multi-mode set rainfall prediction series, and the number of nodes is 1;
and improving the input and output data processing of the BP neural network aiming at the data extension condition that the prediction stage possibly exceeds the extreme value of the training sample, and normalizing the input data to (α):
Figure FDA0002192518160000034
wherein the content of the first and second substances,
Figure FDA0002192518160000035
wherein XkFor the input sequence of samples, Xmin、XmaxRespectively a maximum and a minimum, X, in the input sequence of samplesTraining max、XPremaxMaximum values of training period and prediction period in sample input sequence, akThe input data is normalized;
for desired output TkThe above changes are also made to normalize the data to (α):
Figure FDA0002192518160000041
wherein, Tmin、TmaxMaximum and minimum values in the desired output sequence for the sample, respectively, bkOutputting a normalized value for the expectation;
when the model training is finished and the parameters are fixed, the model is used for prediction, the output value of the model is still between (0, 1), and the model is required to be converted into an actual value;
Figure FDA0002192518160000042
ckoutputting a value for the model actually;
the fifth module is further for training and testing the BP neural network; selecting an S-shaped function as an excitation function; setting an initial learning rate and momentum items; randomly generating a weight value and a threshold value in the network; setting a network training end condition;
the sixth module is further used for carrying out network modeling and prediction on each rainfall station to obtain a multi-mode set rainfall prediction series output by the neural network of each rainfall station in the research area, counting rainfall station mode prediction values of different time windows, and carrying out spatial interpolation by using an interpolation module in ArcGIS to obtain future climate space-time distribution in the research area.
CN201610109283.2A 2016-02-26 2016-02-26 Climate change prediction method and system based on improved BP neural network fitting of multiple climate modes Active CN105740991B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610109283.2A CN105740991B (en) 2016-02-26 2016-02-26 Climate change prediction method and system based on improved BP neural network fitting of multiple climate modes

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610109283.2A CN105740991B (en) 2016-02-26 2016-02-26 Climate change prediction method and system based on improved BP neural network fitting of multiple climate modes

Publications (2)

Publication Number Publication Date
CN105740991A CN105740991A (en) 2016-07-06
CN105740991B true CN105740991B (en) 2020-04-03

Family

ID=56248681

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610109283.2A Active CN105740991B (en) 2016-02-26 2016-02-26 Climate change prediction method and system based on improved BP neural network fitting of multiple climate modes

Country Status (1)

Country Link
CN (1) CN105740991B (en)

Families Citing this family (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106227706B (en) * 2016-07-25 2018-07-24 河海大学 A kind of more climatic model output data integrated calibrations and uncertain appraisal procedure
CN106874511B (en) * 2017-03-06 2020-05-05 云南电网有限责任公司电力科学研究院 Database based on insulator metal accessory corrosion electric charge quantity prediction system
CN107403237A (en) * 2017-07-03 2017-11-28 上海海洋大学 West and central Pacific ocean stripped tuna fishing ground forecasting procedure based on Different climate condition
CN107423811B (en) * 2017-07-04 2018-12-14 河海大学 The streamflow change attribution recognition methods combined based on BP artificial neural network and Scene Simulation
CN109146194A (en) * 2018-09-05 2019-01-04 重庆工商大学 A kind of Runoff forestry method encoding interconnection vector machine certainly based on change mode convolution
CN109726867B (en) * 2018-12-27 2020-07-28 北京恒泰实达科技股份有限公司 High-resolution electric power weather forecasting method based on multi-mode set
US10871594B2 (en) * 2019-04-30 2020-12-22 ClimateAI, Inc. Methods and systems for climate forecasting using artificial neural networks
CN110442937B (en) * 2019-07-24 2023-01-24 武汉大学 Drainage basin hydrological simulation method integrating satellite remote sensing and machine learning technology
CN110646867A (en) * 2019-08-28 2020-01-03 北京无线电计量测试研究所 Urban drainage monitoring and early warning method and system
CN110555554A (en) * 2019-08-28 2019-12-10 向波 intelligent climate prediction technology based on objective quantification
CN111126684A (en) * 2019-12-13 2020-05-08 北京心中有数科技有限公司 Climate prediction method, climate prediction apparatus, computer-readable storage medium, and server
CN111145304B (en) * 2019-12-27 2023-10-10 新奥数能科技有限公司 Data processing method and device, intelligent terminal and storage medium
CN111488974B (en) * 2020-04-14 2023-05-02 四川北控清洁能源工程有限公司 Ocean wind energy downscaling method based on deep learning neural network
CN111784046B (en) * 2020-06-30 2023-12-22 中国人民解放军国防科技大学 Method for predicting future change trend of storm axis activity
US11687620B2 (en) 2020-12-17 2023-06-27 International Business Machines Corporation Artificial intelligence generated synthetic image data for use with machine language models
CN113011683A (en) * 2021-04-26 2021-06-22 中国科学院地理科学与资源研究所 Crop yield estimation method and system based on corrected crop model
CN113592132B (en) * 2021-04-29 2022-04-05 山东省气象科学研究所(山东省海洋气象科学研究所、山东省气象局培训中心) Rainfall objective forecasting method based on numerical weather forecast and artificial intelligence
CN113361766B (en) * 2021-06-03 2024-04-09 南京信息工程大学 Multi-mode precipitation prediction method integrated with machine learning
CN113850420A (en) * 2021-09-13 2021-12-28 江苏海洋大学 Weather data prediction method for optimizing BP neural network based on improved genetic algorithm

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2464677A (en) * 2008-10-20 2010-04-28 Univ Nottingham Trent A method of analysing data by using an artificial neural network to identify relationships between the data and one or more conditions.
CN101852871A (en) * 2010-05-25 2010-10-06 南京信息工程大学 Short-term climate forecasting method based on empirical mode decomposition and numerical value set forecasting
CN102122370A (en) * 2011-03-07 2011-07-13 北京师范大学 Method for predicting river basin climatic change and analyzing tendency
CN102495937A (en) * 2011-10-18 2012-06-13 南京信息工程大学 Prediction method based on time sequence
CN103838979A (en) * 2014-03-26 2014-06-04 武汉大学 Statistics downscaling method based on SVM algorithm
US9471884B2 (en) * 2014-05-30 2016-10-18 International Business Machines Corporation Multi-model blending
CN104820754A (en) * 2015-05-13 2015-08-05 南京信息工程大学 Space statistical downscaling rainfall estimation method based on geographical difference analysis method

Also Published As

Publication number Publication date
CN105740991A (en) 2016-07-06

Similar Documents

Publication Publication Date Title
CN105740991B (en) Climate change prediction method and system based on improved BP neural network fitting of multiple climate modes
CN111310968B (en) LSTM neural network circulating hydrologic forecasting method based on mutual information
CN111104981B (en) Hydrological prediction precision evaluation method and system based on machine learning
Mahsin et al. Modeling rainfall in Dhaka division of Bangladesh using time series analysis
TW201734837A (en) Multi-sampling model training method and device
CN111665575B (en) Medium-and-long-term rainfall grading coupling forecasting method and system based on statistical power
US10770898B2 (en) Methods and systems for energy use normalization and forecasting
CN112560173B (en) Vehicle weather resistance temperature prediction method and device based on deep learning
CN109598052B (en) Intelligent ammeter life cycle prediction method and device based on correlation coefficient analysis
CN105974495A (en) Method for pre-judging future average cloud amount of target area by using classification fitting method
CN111639803A (en) Prediction method applied to future vegetation index of area under climate change scene
CN108763648A (en) Method and apparatus based on nuclear magnetic resonance T2 distributed acquisition capillary pressure curves
CN115759445A (en) Machine learning and cloud model-based classified flood random forecasting method
Devak et al. Downscaling of precipitation in Mahanadi basin, India
Muthusamy et al. Comparison of different configurations of quantile regression in estimating predictive hydrological uncertainty
CN116541681A (en) Composite disaster space variability identification method based on collaborative kriging interpolation
Oesting et al. Sampling from a Max‐Stable Process Conditional on a Homogeneous Functional with an Application for Downscaling Climate Data
CN113487069B (en) Regional flood disaster risk assessment method based on GRACE daily degradation scale and novel DWSDI index
Lee et al. Forecasting korean stock price index (kospi) using backpropagationn neural network model bayesian chiao’s model and sarima model
Condeixa et al. Wind speed time series analysis using TBATS decomposition and moving blocks bootstrap
Lee Climate change inspector with intentionally biased bootstrapping (CCIIBB ver. 1.0)–methodology development
Liu et al. A method for deterministic statistical downscaling of daily precipitation at a monsoonal site in Eastern China
Herath et al. Downscaling approach to develop future sub-daily IDF relations for Canberra Airport Region, Australia
Chen et al. Uncertainty analysis of hydrologic forecasts based on copulas
CN112749510B (en) Precipitation downscaling method combining support vector machine and interactive factor selection

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant