CN117273200A - Runoff interval forecasting method based on convolution optimization algorithm and Pyraformer neural network - Google Patents
Runoff interval forecasting method based on convolution optimization algorithm and Pyraformer neural network Download PDFInfo
- Publication number
- CN117273200A CN117273200A CN202311112445.4A CN202311112445A CN117273200A CN 117273200 A CN117273200 A CN 117273200A CN 202311112445 A CN202311112445 A CN 202311112445A CN 117273200 A CN117273200 A CN 117273200A
- Authority
- CN
- China
- Prior art keywords
- interval
- runoff
- prediction
- index
- value
- 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.)
- Granted
Links
- 238000005457 optimization Methods 0.000 title claims abstract description 43
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 17
- 238000013277 forecasting method Methods 0.000 title claims abstract description 17
- 108090000672 Annexin A5 Proteins 0.000 claims abstract description 36
- 238000000034 method Methods 0.000 claims abstract description 34
- 238000012937 correction Methods 0.000 claims abstract description 25
- 238000003062 neural network model Methods 0.000 claims abstract description 11
- 239000002131 composite material Substances 0.000 claims abstract description 6
- 239000011159 matrix material Substances 0.000 claims description 31
- 230000006870 function Effects 0.000 claims description 21
- 238000005259 measurement Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 11
- 108091081062 Repeated sequence (DNA) Proteins 0.000 claims description 9
- 230000000737 periodic effect Effects 0.000 claims description 9
- 238000011156 evaluation Methods 0.000 claims description 7
- 230000007246 mechanism Effects 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 6
- 230000000694 effects Effects 0.000 claims description 5
- 241000039077 Copula Species 0.000 claims description 3
- 238000013459 approach Methods 0.000 claims description 3
- 230000001419 dependent effect Effects 0.000 claims description 3
- 238000012938 design process Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 claims description 3
- 230000003094 perturbing effect Effects 0.000 claims description 3
- 230000008859 change Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000013467 fragmentation Methods 0.000 description 1
- 238000006062 fragmentation reaction Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000001556 precipitation Methods 0.000 description 1
- 238000002922 simulated annealing Methods 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
- 239000005436 troposphere Substances 0.000 description 1
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Tourism & Hospitality (AREA)
- Operations Research (AREA)
- Educational Administration (AREA)
- Marketing (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- General Business, Economics & Management (AREA)
- Algebra (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Pure & Applied Mathematics (AREA)
- Databases & Information Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a runoff interval forecasting method based on a convolution optimization algorithm and a Pyraformer neural network, which comprises the steps of collecting data, and optimizing a Pyraformer neural network model by using the convolution optimization algorithm as a prediction model of the runoff interval; the coverage rate index PICP, the average width index PINAW and the symmetry index PIS are utilized to define a composite objective function CSWC of a convolution optimization algorithm as an adaptability function of the model, and the parameters of the prediction model are optimized in a parameter feasible domain range; predicting the upper boundary and the lower boundary of the runoff interval value by using a prediction model; meanwhile, the accuracy of the upper boundary and the lower boundary is evaluated and corrected by using a coverage rate index PICP, an average width index PINAW and a symmetry index PIS; obtaining a prediction result of error data by using the upper and lower boundary errors; and sending the predicted result into a designed error correction strategy to obtain a final interval predicted result. The method solves the problem that the runoff interval forecast by the local optimization algorithm is inaccurate in the prior art.
Description
Technical Field
The invention relates to the technical field of runoff interval forecasting, in particular to a runoff interval forecasting method based on a convolution optimization algorithm and a Pyraformer neural network.
Background
Under the common influence of global climate change and human activities, extreme weather events frequently occur, the air moisture content of the global near-surface and troposphere rises, and strong precipitation in the range of human activity areas is also exacerbated, which increases the flood risk year by year. Meanwhile, the characteristics of the river basin are changed, and the phenomenon of runoff fragmentation is increasingly serious, so that hydrologic elements such as river runoff are obviously changed.
The runoff prediction model is a mathematical model for predicting river runoff, is a simplified representation of a real hydrologic system, and predicts the change condition of the river flow path flow in a future period according to the hydrologic characteristics, meteorological conditions, topography and other information of a river basin, and is usually composed of an upper boundary and a lower boundary of runoff prediction. In the radial flow interval prediction, besides a complete radial flow interval prediction model, model parameters also determine the key of prediction accuracy, and the parameter problem of the radial flow interval prediction model is mainly a manual trial and error method, so that researchers are required to have higher experience knowledge, but the efficiency is low, and the result is difficult to reach the optimum.
Along with the development of modern computer technology and the coming of big data age, more abundant data resources are easy to obtain, and the local optimization algorithm is used for carrying out parameter optimization on a runoff interval forecast model, such as a Rosenblock method, a Powell method and a Simplex method, the local optimization algorithm provides convenience for the parameter rate of the runoff interval forecast model to a certain extent, but because of the nonlinearity of the runoff interval forecast model, the local optimization algorithm is easy to sink into the local optimization, and the global optimal calibration result is difficult to capture, so the global optimization algorithm is proposed, such as a particle swarm algorithm, a differential evolution algorithm, a simulated annealing algorithm and the like, has higher robustness, and the interference of local extremum can be effectively avoided. However, in the actual prediction of the runoff interval, there are often requirements on multiple indexes such as flow, flow process, etc., and these indexes are mutually restricted, so that the optimization cannot be achieved at the same time. There is therefore a need for a more accurate method of forecasting radial intervals to provide more comprehensive forecast information.
Disclosure of Invention
The invention aims to: the invention aims to provide a runoff interval forecasting method based on a convolution optimization algorithm and a Pyraformer neural network, which is used for accurately and rapidly forecasting upper and lower boundary intervals of runoff.
The technical scheme is as follows: in order to achieve the above purpose, the runoff interval forecasting method based on a convolution optimization algorithm and a Pyraformer neural network comprises the following steps:
step 1: collecting data, including a geographic potential dependent sequence, a proximity sequence, a periodic repeated sequence and a historical meteorological sequence of a runoff region;
step 2: defining a coverage index PICP, an average width index PINAW and a symmetry index PIS;
step 3: optimizing a Pyraformer neural network model by adopting a convolution optimization algorithm to serve as a prediction model of a runoff interval;
step 4: operating the radial flow interval prediction model obtained in the step 3 by using the data in the step 1 to obtain a parameter feasible region of the model; defining a composite objective function CSWC of a convolution optimization algorithm by using the index in the step 2 as an adaptability function of the model, and optimizing predicted model parameters of a runoff interval in a parameter feasible domain range;
step 5: inputting the data set in the step 1 into the runoff interval prediction model after optimizing parameters in the step 4, extracting time-space dimension characteristics of the data, and predicting upper and lower boundaries of the runoff interval value by combining with the fully-connected prediction output layer; performing accuracy evaluation and correction on the upper and lower boundaries of the predicted runoff interval by using the indexes in the step 2;
step 6: calculating the upper and lower boundary errors of the predicted runoff interval to obtain error data, and sending the error data into a prediction model of the runoff interval after optimizing parameters to obtain a prediction result of the error data;
step 7: and designing an error correction strategy, and sending the prediction result of the error data into the error correction strategy to obtain a final interval prediction result.
The data described in step 1 is the entire basin containing |v|=n sub-basins, and the set of sub-basin control sites is defined as v= { V 1 ,...,v n ,...,v N N represents the nth sub-watershed site; the dependency matrix between the sub-watershed control sites is l= { L geo ,L rel }, wherein L geo Is a geographic adjacency matrix L rel An adjacency matrix for potential spatial dependencies;
for any two different sub-watershed i and j, its geographic adjacency matrix and latent dependency matrix distribution are expressed as:
wherein CEPMI (v) i ,v j ) Copula entropy bias mutual information between the sub-watershed i and j is represented; l (L) rel =1, meaning that the partial mutual information of sub-watershed i and j is greater than a given threshold δ;
according to the distribution of the geographical adjacency matrix and the potential dependency matrix of any two different sub-watersheds i and j, a subscript T is intercepted along a time axis g The time sequence is a geographic potential dependency sequence
Analyzing the time dependence of the basin runoff process on periodicity and adjacent moments, and respectively intercepting subscripts T along a time axis h 、T p And T m The time sequence segment is the input of the model at the adjacent time, the periodical repeated time and the historical meteorological sequence component, t 0 The time sequence sets are respectively expressed as
Is provided withRepresenting the runoff value at time t in the sub-basin n, then +.>The method comprises the steps of collecting runoff values of all sub-watershed at t time, namely a target value;
construction of sample set (x t ,Y t ) T=1, 2,., N, N represents the number of sub-watershed, i.e. the number of sample points, where x t =[L,K,P,T] T For the input data set, data information of a geographic potential dependency sequence L, a nearby sequence K, a periodic repeated sequence P and a historical meteorological sequence T is contained.
The indexes in the step 2 are respectively as follows:
coverage index PICP:
when the target value is covered by the forecast interval Shi Buer function c i The value of (2) is 1, otherwise 0;
average width index PINAW:
symmetry index PIS:
wherein N represents the size of the sample set, R is the value range of the target vector, i is the sample point, y i Representing the measured value of the i-th sample point; LB (LB) i And UB i Respectively represent y i Lower and upper limits of (2).
The optimization method in the step 3 is as follows: introducing a convolution optimization algorithm MCOA to optimize parameters of the Pyraformer neural network model including learning rate, hidden layer number and number of detail points so as to improve the convergence rate of the model, wherein the specific optimization process is as follows: inputting the number of the initialized population, the dimension of the individual position, the maximum iteration times, the convolution kernel parameters and the fitness function; performing a location update strategy; updating the global optimal solution and the position thereof; and outputting the parameters of the optimal solution assigned to the Pyraformer neural network model.
The method for obtaining the parameter feasible region of the runoff interval prediction model in the step 4 is as follows: initializing parameters of a runoff interval prediction model, inputting data, obtaining an optimal parameter set theta of the runoff interval prediction model, slightly perturbing the set theta, obtaining a parameter feasible region F, wherein the feasible region F is expressed as:
F=[θ-ω≤θ≤θ+ω]。
the formula of the composite objective function CSWC in the step 4 is as follows:
wherein a, b and c are weight values of indexes, and the smaller CSWC is, the higher the coverage rate of the forecast interval is, the smaller the width is, and the higher the symmetry is;
in the CSWC function, PIARW is used as a main measurement index, a is set to be a larger value to ensure that the coverage rate of a forecast interval is better, and then a more ideal forecast interval is obtained, and the sum form among PICP, PIARW and index PIS in the CSWC function is used to avoid the extreme condition that PIARW is 0 in the product form, namely the upper boundary and the lower boundary of the forecast interval are overlapped, and the forecast width is 0.
And 5, predicting the upper and lower boundaries of the runoff interval value, wherein the specific process is as follows:
constructing a large matrix data set by using a geographic potential dependency sequence, a proximity sequence, a periodic repeated sequence and a historical meteorological element sequence, and sending the large matrix data set into a prediction model of a runoff interval, and constructing a multi-resolution C element tree by using a coarse scale construction module CSCM, wherein the coarse scale nodes summarize information of corresponding fine scale C nodes; the pyramid attention module PAM is introduced to further capture time-space dependencies of different ranges, and the attention mechanism in the pyramid diagram is utilized to transmit messages; the output of the fully connected network is directly used as the prediction of the upper and lower boundaries of the interval prediction;
wherein the pyramid attention module PAM comprises a scale time connection and a scale space connection, the scale time connection forming a C-ary tree, wherein each parent node has C child nodes; using the finest scale of the pyramid graph to correlate with the hourly observations of the original time series, the nodes on the coarse scale are considered as time-of-day, week, and month features of the time series; connecting adjacent nodes through a scale space, and capturing a remote dependency relationship in a coarse scale;
in the attention mechanism, X and Y represent input and output of single attention, respectively, and X is first linearly transformed into three different matrices, i.e. query q=xw Q Key k=xw K Value v=xw V WhereinW Q ,W K ,W V Respectively a weight matrix of inquiry, a weight matrix of key, a Q of ith row in Q i Is any row in K, i.e. a key, the output of the attention mechanism is:
wherein,transpose of line I in K, D K Representing the dimension of the key; i is the sample point, y i The measured value of the i-th sample point is shown.
And carrying out single-step prediction on the historical sequence, after the sequence is subjected to PAM coding of a pyramid attention module, assembling the features of the last node in the pyramid graph on all scales, inputting the spliced features into a fully-connected layer for prediction, outputting two predicted values, wherein a larger value is used as the upper boundary of a prediction interval at the moment, and a smaller value is used as the lower boundary of the prediction interval at the moment.
The evaluation method in the step 5 comprises the following steps:
when the value of the coverage rate index PICP is close to 1, the more actual measurement values are covered by the prediction interval, the better the effect of the prediction model of the runoff interval is; the average width index PINAW is as narrow as possible in an effective forecasting interval under the condition of higher coverage rate, so that the forecasting interval contains effective information;
when the forecast interval covers the actual measurement value, the symmetry index PIS takes the value between 0 and 0.5; when the actual measurement is in the center of the prediction interval, namely the optimal symmetry condition, the symmetry index PIS takes a value of 0; when the symmetry index PIS is close to 0, the symmetry of the prediction interval about the measured value is better; when the measured value is smaller than the lower boundary of the predicted interval and larger than the upper boundary of the predicted interval, the value of the symmetry index PIS is larger than 0.5, and the degree of asymmetry of the predicted interval to the measured value is high;
the correction method comprises the following steps: calculating the evaluation results of a coverage index PICP, an average width index PINAW and a symmetry index PIS according to the upper boundary and the lower boundary of the prediction interval, and giving a confidence probability of 100 (1-alpha)%, and correcting the prediction of the prediction interval which cannot meet the index by the following formula:
wherein,forecasting the upper and lower boundaries of the runoff interval value, wherein alpha is the significance level, and the forecasting interval is expressed as +.>The target vector is located in the forecast interval +.>The probability of (2) is expressed as
The error data in step 6 is expressed as:
in the formula, UB_Err i 、And UB con Respectively representing an upper boundary error, a forecast upper boundary and a set upper boundary; LB_Err i 、/>And LB con Respectively representing a lower boundary error, a forecast lower boundary and a set lower boundary; wherein forecast boundary->Andis the value of the predicted result of the runoff interval prediction model and corrected by the result of the three indexes;
the method for obtaining the prediction result of the error data comprises the following steps: the UB_Err is used for i And LB_Err i The error data is input into a prediction model of the runoff interval to obtain prediction results of the error data, wherein the prediction results are respectively upper boundary prediction results UB_Err pre And lower boundary prediction result LB_Err pre 。
The design process of the error correction strategy described in the step 7 is as follows:
according to the prediction result of the runoff interval prediction model after optimizing the parametersAnd->Calculating a coverage rate index PICP value;
if the coverage index PICP value is higher than the set confidence level, the adjustment strategy is as follows: the reduction of the average width index PINAW and the symmetry index PIS is more important than the increase of the coverage index PICP to improve the quality of the forecast interval;
when the upper boundary error UB_Err pre In the positive direction, the upper boundary prediction result UB is represented pre Greater than the set upper boundary UB con Then the upper boundary prediction result UB pre Subtracting the upper boundary error UB_Err pre To reduce the width of the forecast interval;
when the upper boundary error UB_Err pre When negative, the upper boundary prediction result UB is represented pre Less than the set upper boundary UB con Upper boundary error UB _ Err pre Switching to zero to avoid increasing the forecast interval width; LB (LB) pre Is to adjust policy and UB pre Conversely;
if the coverage index PICP value is lower than the set confidence level, it is more important to increase PICP than to decrease PINAW and PIS, so the strategy is adjusted as follows: upper boundary error ub_err pre And lower boundary error LB_Err pre Remain unchanged to ensure that the boundary of the forecast interval approaches the set interval edgeA boundary; when the coverage rate index PICP is increased, the quality of the forecast interval is improved;
the final interval prediction result obtained through the difference correction strategy is expressed as:
wherein UB is cal And LB cal Representing the correction upper and correction lower boundaries, respectively.
The beneficial effects are that: the invention has the following advantages: 1. the prediction model of the runoff interval constructed by the method can better process time sequence data and capture the long-term dependence relationship between the data, and outputs the predicted upper and lower boundaries by combining with the fully-connected prediction layer to form a complete prediction interval, so that more comprehensive prediction information is provided; 2. the accuracy of the upper boundary and the lower boundary of the predicted runoff interval is evaluated and corrected by using a coverage rate index PICP, an average width index PINAW and a symmetry index PIS, so that a more accurate predicted runoff interval is obtained; 3. and an error correction strategy is provided for adjusting the upper boundary and the lower boundary of the forecast interval, so that possible deviation of the forecast interval is corrected in time, and a more reliable forecast result is provided.
Drawings
FIG. 1 is a schematic flow chart of the present invention;
fig. 2 is a flowchart of a Pyraformer neural network model.
Detailed Description
The technical scheme of the present invention will be described in detail with reference to the following examples and the accompanying drawings.
As shown in fig. 1, a runoff interval forecasting method based on a convolution optimization algorithm and a Pyraformer neural network comprises the following steps:
step 1: collecting data, including a geographic potential dependent sequence, a proximity sequence, a periodic repeated sequence and a historical meteorological sequence of a runoff region;
step 2: defining a coverage index PICP, an average width index PINAW and a symmetry index PIS;
step 3: optimizing a Pyraformer neural network model by adopting a convolution optimization algorithm to serve as a prediction model of a runoff interval;
step 4: using the data in the step 1 to run a prediction model of the runoff interval to obtain a parameter feasible region of the model; defining a composite objective function CSWC of a convolution optimization algorithm by using a coverage rate index PICP, an average width index PINAW and a symmetry index PIS as an adaptability function of the model, and optimizing prediction model parameters of a runoff interval in a parameter feasible domain range;
step 5: the data set in the step 1 is input into a prediction model of the runoff interval after the optimization parameters to extract the time dimension and the space dimension characteristics of the data, and the upper boundary and the lower boundary of the value of the runoff interval are predicted by combining with a fully-connected prediction output layer; evaluating and correcting the accuracy of the upper and lower boundaries of the predicted runoff interval by using a coverage rate index PICP, an average width index PINAW and a symmetry index PIS;
step 6: calculating the upper and lower boundary errors of the predicted runoff interval to obtain error data, and sending the error data into a prediction model of the runoff interval after optimizing parameters to obtain a prediction result of the error data;
step 7: and designing an error correction strategy, and sending the prediction result of the error data into the error correction strategy to obtain a final interval prediction result.
The data described in step 1 is the entire basin containing |v|=n sub-basins, and the set of sub-basin control sites is defined as v= { V 1 ,...,v n ,...,v N N represents the nth sub-watershed site; the dependency matrix between the sub-watershed control sites is l= { L geo ,L rel }, wherein L geo Is a geographic adjacency matrix L rel An adjacency matrix for potential spatial dependencies;
for any two different sub-watershed i and j, its geographic adjacency matrix and latent dependency matrix distribution are expressed as:
wherein CEPMI (v) i ,v j ) Copula entropy bias mutual information between the sub-watershed i and j is represented; l (L) rel =1, meaning that the partial mutual information of sub-watershed i and j is greater than a given threshold δ;
according to the distribution of the geographical adjacency matrix and the potential dependency matrix of any two different sub-watersheds i and j, a subscript T is intercepted along a time axis g The time sequence is a geographic potential dependency sequence
Analyzing the time dependence of the basin runoff process on periodicity and adjacent moments, and respectively intercepting subscripts T along a time axis h 、T p And T m The time sequence segment is the input of the model at the adjacent time, the periodical repeated time and the historical meteorological sequence component, t 0 The time sequence sets are respectively expressed as
Is provided withRepresenting the runoff value at time t in the sub-basin n, then +.>The method comprises the steps of collecting runoff values of all sub-watershed at t time, namely a target value;
construction of sample set (x t ,Y t ) T=1, 2,., N, N represents the number of sub-watershed, i.e. the number of sample points, where x t =[L,K,P,T] T For the input data set, data information of a geographic potential dependency sequence L, a nearby sequence K, a periodic repeated sequence P and a historical meteorological sequence T is contained.
The indexes in the step 2 are respectively as follows:
coverage index PICP:
when the target value is covered by the forecast interval Shi Buer function c i The value of (2) is 1, otherwise 0;
average width index PINAW:
symmetry index PIS:
wherein N represents the size of the sample set, R is the value range of the target vector, i is the sample point, y i Representing the measured value of the i-th sample point; LB (LB) i And UB i Respectively represent y i Lower and upper limits of (2).
The optimization method in the step 3 is as follows: introducing a convolution optimization algorithm MCOA to optimize parameters of the Pyraformer neural network model including learning rate, hidden layer number and number of detail points so as to improve the convergence rate of the model, wherein the specific optimization process is as follows: inputting the number of the initialized population, the dimension of the individual position, the maximum iteration times, the convolution kernel parameters and the fitness function; performing a location update strategy; updating the global optimal solution and the position thereof; and outputting the parameters of the optimal solution assigned to the Pyraformer neural network model, as shown in fig. 2.
The method for obtaining the parameter feasible region of the runoff interval prediction model in the step 4 is as follows: initializing parameters of a runoff interval prediction model, inputting data, obtaining an optimal parameter set theta of the runoff interval prediction model, slightly perturbing the set theta, obtaining a parameter feasible region F, wherein the feasible region F is expressed as:
F=[θ-ω≤θ≤θ+ω]。
the formula of the composite objective function CSWC in the step 4 is as follows:
wherein a, b and c are weight values of indexes, and the smaller CSWC is, the higher the coverage rate of the forecast interval is, the smaller the width is, and the higher the symmetry is;
in the CSWC function, PIARW is used as a main measurement index, a is set to be a larger value to ensure that the coverage rate of a forecast interval is better, and then a more ideal forecast interval is obtained, and the sum form among PICP, PIARW and index PIS in the CSWC function is used to avoid the extreme condition that PIARW is 0 in the product form, namely the upper boundary and the lower boundary of the forecast interval are overlapped, and the forecast width is 0.
And 5, predicting the upper and lower boundaries of the runoff interval value, wherein the specific process is as follows:
constructing a large matrix data set by using a geographic potential dependency sequence, a proximity sequence, a periodic repeated sequence and a historical meteorological element sequence, and sending the large matrix data set into a prediction model of a runoff interval, and constructing a multi-resolution C element tree by using a coarse scale construction module CSCM, wherein the coarse scale nodes summarize information of corresponding fine scale C nodes; the pyramid attention module PAM is introduced to further capture time-space dependencies of different ranges, and the attention mechanism in the pyramid diagram is utilized to transmit messages; the output of the fully connected network is directly used as the prediction of the upper and lower boundaries of the interval prediction;
wherein the pyramid attention module PAM comprises a scale time connection and a scale space connection, the scale time connection forming a C-ary tree, wherein each parent node has C child nodes; using the finest scale of the pyramid graph to correlate with the hourly observations of the original time series, the nodes on the coarse scale are considered as time-of-day, week, and month features of the time series; connecting adjacent nodes through a scale space, and capturing a remote dependency relationship in a coarse scale;
in the attention mechanism, X and Y represent input and output of single attention respectively, and X is first linearly transformed into three different matrices, i.e. query q=xW Q Key k=xw K Value v=xw V WhereinW Q ,W K ,W V Respectively a weight matrix of inquiry, a weight matrix of key, a Q of ith row in Q i Is any row in K, i.e. a key, the output of the attention mechanism is:
wherein,transpose of line I in K, D K Representing the dimension of the key; i is the sample point, y i The measured value of the i-th sample point is shown.
And carrying out single-step prediction on the historical sequence, after the sequence is subjected to PAM coding of a pyramid attention module, assembling the features of the last node in the pyramid graph on all scales, inputting the spliced features into a fully-connected layer for prediction, outputting two predicted values, wherein a larger value is used as the upper boundary of a prediction interval at the moment, and a smaller value is used as the lower boundary of the prediction interval at the moment.
The evaluation method in the step 5 comprises the following steps:
when the value of the coverage rate index PICP is close to 1, the more actual measurement values are covered by the prediction interval, the better the effect of the prediction model of the runoff interval is; the average width index PINAW is as narrow as possible in an effective forecasting interval under the condition of higher coverage rate, so that the forecasting interval contains effective information;
when the forecast interval covers the actual measurement value, the symmetry index PIS takes the value between 0 and 0.5; when the actual measurement is in the center of the prediction interval, namely the optimal symmetry condition, the symmetry index PIS takes a value of 0; when the symmetry index PIS is close to 0, the symmetry of the prediction interval about the measured value is better; when the measured value is smaller than the lower boundary of the predicted interval and larger than the upper boundary of the predicted interval, the value of the symmetry index PIS is larger than 0.5, and the degree of asymmetry of the predicted interval to the measured value is high;
the correction method comprises the following steps: calculating the evaluation results of a coverage index PICP, an average width index PINAW and a symmetry index PIS according to the upper boundary and the lower boundary of the prediction interval, and giving a confidence probability of 100 (1-alpha)%, and correcting the prediction of the prediction interval which cannot meet the index by the following formula:
wherein,forecasting the upper and lower boundaries of the runoff interval value, wherein alpha is the significance level, and the forecasting interval is expressed as +.>The target vector is located in the forecast interval +.>The probability of (2) is expressed as
The error data in step 6 is expressed as:
in the formula, UB_Err i 、And UB con Respectively representing an upper boundary error, a forecast upper boundary and a set upper boundary; LB_Err i 、/>And LB con Respectively representing a lower boundary error, a forecast lower boundary and a set lower boundary; wherein forecast boundary->Andis the value of the predicted result of the runoff interval prediction model and corrected by the result of the three indexes;
the method for obtaining the prediction result of the error data comprises the following steps: the UB_Err is used for i And LB_Err i The error data is input into a prediction model of the runoff interval to obtain prediction results of the error data, wherein the prediction results are respectively upper boundary prediction results UB_Err pre And lower boundary prediction result LB_Err pre 。
The design process of the error correction strategy described in the step 7 is as follows:
according to the prediction result of the runoff interval prediction model after optimizing the parametersAnd->Calculating a coverage rate index PICP value;
if the coverage index PICP value is higher than the set confidence level, the adjustment strategy is as follows: the reduction of the average width index PINAW and the symmetry index PIS is more important than the increase of the coverage index PICP to improve the quality of the forecast interval;
when the upper boundary error UB_Err pre In the positive direction, the upper boundary prediction result UB is represented pre Greater than the set upper boundary UB con Then the upper boundary prediction result UB pre Subtracting the upper boundary error UB_Err pre To reduce the width of the forecast interval;
when the upper boundary error UB_Err pre When negative, the upper boundary prediction result UB is represented pre Less than the set upper boundary UB con Upper boundary error UB _ Err pre Switching to zero to avoid increasing the forecast interval width; LB (LB) pre Is to adjust policy and UB pre Conversely;
if the coverage index PICP value is lower than the set confidence level, it is more important to increase PICP than to decrease PINAW and PIS, so the strategy is adjusted as follows: upper boundary error ub_err pre And lower boundary error LB_Err pre Keeping unchanged to ensure that the forecast interval boundary approaches the set interval boundary; when the coverage rate index PICP is increased, the quality of the forecast interval is improved;
the final interval prediction result obtained through the difference correction strategy is expressed as:
wherein UB is cal And LB cal Representing the correction upper and correction lower boundaries, respectively.
Claims (10)
1. The runoff interval forecasting method based on the convolution optimization algorithm and the Pyraformer neural network is characterized by comprising the following steps of:
step 1: collecting data, including a geographic potential dependent sequence, a proximity sequence, a periodic repeated sequence and a historical meteorological sequence of a runoff region;
step 2: defining a coverage index PICP, an average width index PINAW and a symmetry index PIS;
step 3: optimizing a Pyraformer neural network model by adopting a convolution optimization algorithm to serve as a prediction model of a runoff interval;
step 4: operating the radial flow interval prediction model obtained in the step 3 by using the data in the step 1 to obtain a parameter feasible region of the model; defining a composite objective function CSWC of a convolution optimization algorithm by using the index in the step 2 as an adaptability function of the model, and optimizing predicted model parameters of a runoff interval in a parameter feasible domain range;
step 5: inputting the data set in the step 1 into the runoff interval prediction model after optimizing parameters in the step 4, extracting time-space dimension characteristics of the data, and predicting upper and lower boundaries of the runoff interval value by combining with the fully-connected prediction output layer; performing accuracy evaluation and correction on the upper and lower boundaries of the predicted runoff interval by using the indexes in the step 2;
step 6: calculating the upper and lower boundary errors of the predicted runoff interval to obtain error data, and sending the error data into a prediction model of the runoff interval after optimizing parameters to obtain a prediction result of the error data;
step 7: and designing an error correction strategy, and sending the prediction result of the error data into the error correction strategy to obtain a final interval prediction result.
2. The runoff interval forecasting method based on the convolution optimization algorithm and the Pyraformer neural network according to claim 1, wherein the data in the step 1 is an entire drainage basin including |v|=n sub-drainage basins, and the set of sub-drainage basin control sites is defined as v= { V 1 ,...,v n ,...,v N N represents the nth sub-watershed site; the dependency matrix between the sub-watershed control sites is l= { L geo ,L rel }, wherein L geo Is a geographic adjacency matrix L rel An adjacency matrix for potential spatial dependencies;
for any two different sub-watershed i and j, its geographic adjacency matrix and latent dependency matrix distribution are expressed as:
wherein CEPMI (v) i ,v j ) Copula entropy bias mutual information between the sub-watershed i and j is represented; l (L) rel =1, meaning that the partial mutual information of sub-watershed i and j is greater than a given threshold δ;
according to the distribution of the geographical adjacency matrix and the potential dependency matrix of any two different sub-watersheds i and j, a subscript T is intercepted along a time axis g The time sequence is a geographic potential dependency sequence
Analyzing the time dependence of the basin runoff process on periodicity and near moments,cut off subscripts T along time axis h 、T p And T m The time sequence segment is the input of the model at the adjacent time, the periodical repeated time and the historical meteorological sequence component, t 0 The time sequence sets are respectively expressed as
Is provided withRepresenting the runoff value at time t in the sub-basin n, then +.>The method comprises the steps of collecting runoff values of all sub-watershed at t time, namely a target value;
construction of sample set (x t ,Y t ) T=1, 2,., N, N represents the number of sub-watershed, i.e. the number of sample points, where x t =[L,K,P,T] T For the input data set, data information of a geographic potential dependency sequence L, a nearby sequence K, a periodic repeated sequence P and a historical meteorological sequence T is contained.
3. The runoff interval forecasting method based on the convolution optimization algorithm and the Pyraformer neural network according to claim 1, wherein the indexes in the step 2 are respectively as follows:
coverage index PICP:
when the target value is forecastedInter-overlay Shi Buer function c i The value of (2) is 1, otherwise 0;
average width index PINAW:
symmetry index PIS:
wherein N represents the size of the sample set, R is the value range of the target vector, i is the sample point, y i Representing the measured value of the i-th sample point; LB (LB) i And UB i Respectively represent y i Lower and upper limits of (2).
4. The runoff interval forecasting method based on the convolution optimization algorithm and the Pyraformer neural network according to claim 1, wherein the optimization method in the step 3 is as follows: introducing a convolution optimization algorithm MCOA to optimize parameters of the Pyraformer neural network model including learning rate, hidden layer number and number of detail points so as to improve the convergence rate of the model, wherein the specific optimization process is as follows: inputting the number of the initialized population, the dimension of the individual position, the maximum iteration times, the convolution kernel parameters and the fitness function; performing a location update strategy; updating the global optimal solution and the position thereof; and outputting the parameters of the optimal solution assigned to the Pyraformer neural network model.
5. The runoff interval forecasting method based on the convolution optimization algorithm and the Pyraformer neural network according to claim 1, wherein the method for obtaining the parameter feasible region of the runoff interval forecasting model in step 4 is as follows: initializing parameters of a runoff interval prediction model, inputting data, obtaining an optimal parameter set theta of the runoff interval prediction model, slightly perturbing the set theta, obtaining a parameter feasible region F, wherein the feasible region F is expressed as:
F=[θ-ω≤θ≤θ+ω]。
6. the runoff interval forecasting method based on the convolution optimization algorithm and the Pyraformer neural network according to claim 1, wherein the complex objective function CSWC formula in the step 4 is as follows:
wherein a, b and c are weight values of indexes, and the smaller CSWC is, the higher the coverage rate of the forecast interval is, the smaller the width is, and the higher the symmetry is;
in the CSWC function, PIARW is used as a main measurement index, a is set to be a larger value to ensure that the coverage rate of a forecast interval is better, and then a more ideal forecast interval is obtained, and the sum form among PICP, PIARW and index PIS in the CSWC function is used to avoid the extreme condition that PIARW is 0 in the product form, namely the upper boundary and the lower boundary of the forecast interval are overlapped, and the forecast width is 0.
7. The runoff interval forecasting method based on the convolution optimization algorithm and the Pyraformer neural network according to claim 1, wherein the forecasting of the upper and lower boundaries of the runoff interval value in the step 5 comprises the following specific processes:
constructing a large matrix data set by using a geographic potential dependency sequence, a proximity sequence, a periodic repeated sequence and a historical meteorological element sequence, and sending the large matrix data set into a prediction model of a runoff interval, and constructing a multi-resolution C element tree by using a coarse scale construction module CSCM, wherein the coarse scale nodes summarize information of corresponding fine scale C nodes; the pyramid attention module PAM is introduced to further capture time-space dependencies of different ranges, and the attention mechanism in the pyramid diagram is utilized to transmit messages; the output of the fully-connected network is directly used as the prediction of the upper and lower boundaries of the interval prediction, the larger value is used as the upper boundary of the time prediction interval, and the smaller value is used as the lower boundary of the time prediction interval.
8. The runoff interval forecasting method based on the convolution optimization algorithm and the Pyraformer neural network according to claim 1, wherein the evaluating method in the step 5 is as follows:
when the value of the coverage rate index PICP is close to 1, the more actual measurement values are covered by the prediction interval, the better the effect of the prediction model of the runoff interval is; the average width index PINAW is as narrow as possible in an effective forecasting interval under the condition of higher coverage rate, so that the forecasting interval contains effective information;
when the forecast interval covers the actual measurement value, the symmetry index PIS takes the value between 0 and 0.5; when the actual measurement is in the center of the prediction interval, namely the optimal symmetry condition, the symmetry index PIS takes a value of 0; when the symmetry index PIS is close to 0, the symmetry of the prediction interval about the measured value is better; when the measured value is smaller than the lower boundary of the predicted interval and larger than the upper boundary of the predicted interval, the value of the symmetry index PIS is larger than 0.5, and the degree of asymmetry of the predicted interval to the measured value is high;
the correction method comprises the following steps: calculating the evaluation results of a coverage index PICP, an average width index PINAW and a symmetry index PIS according to the upper boundary and the lower boundary of the prediction interval, and giving a confidence probability of 100 (1-alpha)%, and correcting the prediction of the prediction interval which cannot meet the index by the following formula:
wherein,forecasting the upper and lower boundaries of the runoff interval value, wherein alpha is the significance level, and the forecasting interval is expressed as +.>The target vector is located in the forecast interval +.>The probability of (2) is expressed as
9. The runoff interval forecasting method based on the convolution optimization algorithm and the Pyraformer neural network according to claim 1, wherein the error data in step 6 is expressed as:
in the formula, UB_Err i 、And UB con Respectively representing an upper boundary error, a forecast upper boundary and a set upper boundary; LB_Err i 、/>And LB con Respectively representing a lower boundary error, a forecast lower boundary and a set lower boundary; wherein forecast boundary->Andis the value of the predicted result of the runoff interval prediction model and corrected by the result of the three indexes;
the method for obtaining the prediction result of the error data comprises the following steps: the UB_Err is used for i And LB_Err i The error data is input into a prediction model of the runoff interval to obtain prediction results of the error data, wherein the prediction results are respectively upper boundary prediction results UB_Err pre And lower boundary prediction result LB_Err pre 。
10. The runoff interval forecasting method based on the convolution optimization algorithm and the Pyraformer neural network according to claim 1, wherein the error correction strategy design process in the step 7 is as follows:
according to the prediction result of the runoff interval prediction model after optimizing the parametersAnd->Calculating a coverage rate index PICP value;
if the coverage index PICP value is higher than the set confidence level, the adjustment strategy is as follows:
when the upper boundary error UB_Err pre In the positive direction, the upper boundary prediction result UB is represented pre Greater than the set upper boundary UB con Then the upper boundary prediction result UB pre Subtracting the upper boundary error UB_Err pre To reduce the width of the forecast interval;
when the upper boundary error UB_Err pre When negative, the upper boundary prediction result UB is represented pre Less than the set upper boundary UB con Upper boundary error UB _ Err pre Switching to zero to avoid increasing the forecast interval width; LB (LB) pre Is to adjust policy and UB pre Conversely;
if the coverage index PICP value is lower than the set confidence level, the adjustment strategy is as follows: upper boundary error ub_err pre And lower boundary error LB_Err pre Keeping unchanged to ensure that the forecast interval boundary approaches the set interval boundary; when the coverage rate index PICP is increased, the quality of the forecast interval is improved;
the final interval prediction result obtained through the difference correction strategy is expressed as:
wherein UB is cal And LB cal Representing the correction upper and correction lower boundaries, respectively.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311112445.4A CN117273200B (en) | 2023-08-31 | 2023-08-31 | Runoff interval forecasting method based on convolution optimization algorithm and Pyraformer neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311112445.4A CN117273200B (en) | 2023-08-31 | 2023-08-31 | Runoff interval forecasting method based on convolution optimization algorithm and Pyraformer neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117273200A true CN117273200A (en) | 2023-12-22 |
CN117273200B CN117273200B (en) | 2024-09-06 |
Family
ID=89209532
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311112445.4A Active CN117273200B (en) | 2023-08-31 | 2023-08-31 | Runoff interval forecasting method based on convolution optimization algorithm and Pyraformer neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117273200B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117973645A (en) * | 2024-04-02 | 2024-05-03 | 华东交通大学 | Photovoltaic power prediction method |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106991278A (en) * | 2017-03-21 | 2017-07-28 | 武汉大学 | It is a kind of to gather precipitation forecast and the coupling process of real-time flood probability forecast |
CN110188922A (en) * | 2019-05-05 | 2019-08-30 | 中国长江电力股份有限公司 | A kind of long-term Runoff Forecast method in the RBF neural based on runoff mechanism |
CN110458722A (en) * | 2019-07-25 | 2019-11-15 | 淮阴工学院 | Flood interval prediction method based on multiple target random vector function connection network |
CN112711896A (en) * | 2021-01-05 | 2021-04-27 | 浙江大学 | Complex reservoir group optimal scheduling method considering multi-source forecast error uncertainty |
CN113570159A (en) * | 2021-08-26 | 2021-10-29 | 西安理工大学 | Runoff prediction method, system and computer storage medium |
CN114282431A (en) * | 2021-12-09 | 2022-04-05 | 淮阴工学院 | Runoff interval prediction method and system based on improved SCA and QRGRU |
WO2022135265A1 (en) * | 2021-01-14 | 2022-06-30 | 中国长江三峡集团有限公司 | Failure warning and analysis method for reservoir dispatching rules under effects of climate change |
CN115495991A (en) * | 2022-09-29 | 2022-12-20 | 河海大学 | Rainfall interval prediction method based on time convolution network |
-
2023
- 2023-08-31 CN CN202311112445.4A patent/CN117273200B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106991278A (en) * | 2017-03-21 | 2017-07-28 | 武汉大学 | It is a kind of to gather precipitation forecast and the coupling process of real-time flood probability forecast |
CN110188922A (en) * | 2019-05-05 | 2019-08-30 | 中国长江电力股份有限公司 | A kind of long-term Runoff Forecast method in the RBF neural based on runoff mechanism |
CN110458722A (en) * | 2019-07-25 | 2019-11-15 | 淮阴工学院 | Flood interval prediction method based on multiple target random vector function connection network |
CN112711896A (en) * | 2021-01-05 | 2021-04-27 | 浙江大学 | Complex reservoir group optimal scheduling method considering multi-source forecast error uncertainty |
WO2022135265A1 (en) * | 2021-01-14 | 2022-06-30 | 中国长江三峡集团有限公司 | Failure warning and analysis method for reservoir dispatching rules under effects of climate change |
CN113570159A (en) * | 2021-08-26 | 2021-10-29 | 西安理工大学 | Runoff prediction method, system and computer storage medium |
CN114282431A (en) * | 2021-12-09 | 2022-04-05 | 淮阴工学院 | Runoff interval prediction method and system based on improved SCA and QRGRU |
CN115495991A (en) * | 2022-09-29 | 2022-12-20 | 河海大学 | Rainfall interval prediction method based on time convolution network |
Non-Patent Citations (4)
Title |
---|
"Determination of Input for Artificial Neural Networks for Flood Forecasting Using the Copula Entropy Method", 《JOURNAL OF HYDROLOGIC ENGINEERING》, vol. 19, no. 11, 1 November 2014 (2014-11-01), pages 25 - 29 * |
卢韦伟;周建中;陈璐;叶磊;: "考虑预报因子选择的神经网络降雨径流模型", 水电能源科学, no. 06, 25 June 2013 (2013-06-25), pages 21 - 25 * |
张海荣: "耦合天气预报的流域短期水文预报方法研究", 《中国博士学位论文全文数据库 基础科学辑 (月刊) 2018年第10期》, 15 August 2018 (2018-08-15), pages 51 - 89 * |
李逸: "基于长短期记忆神经网络及其变体的骆马湖水位预报模型研究", 《中国优秀硕士学位论文全文数据库 基础科学辑 (月刊) 2023 年 第01期》, 15 January 2023 (2023-01-15), pages 74 - 81 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117973645A (en) * | 2024-04-02 | 2024-05-03 | 华东交通大学 | Photovoltaic power prediction method |
Also Published As
Publication number | Publication date |
---|---|
CN117273200B (en) | 2024-09-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111027175B (en) | Method for evaluating social and economic influences of flood based on coupling model integrated simulation | |
CN113128793A (en) | Photovoltaic power combination prediction method and system based on multi-source data fusion | |
CN112288164B (en) | Wind power combined prediction method considering spatial correlation and correcting numerical weather forecast | |
CN111898660B (en) | Hydrological simulation method for average fusion of multi-source data based on Bayesian mode | |
Shin et al. | Seasonal forecasting of daily mean air temperatures using a coupled global climate model and machine learning algorithm for field-scale agricultural management | |
CN106951979A (en) | The crop maturity phase Forecasting Methodology that remote sensing, crop modeling are merged with weather forecast | |
CN110442937A (en) | A kind of valley water security management method merging satellite remote sensing and machine learning techniques | |
CN117273200B (en) | Runoff interval forecasting method based on convolution optimization algorithm and Pyraformer neural network | |
CN114861840B (en) | Multi-source precipitation data fusion method | |
WO2023245399A1 (en) | Rice production potential simulation method based on land system and climate change coupling | |
CN112862274A (en) | WRF-LES and Deepar combined wind power prediction method | |
CN114372631A (en) | Data-lacking area runoff prediction method based on small sample learning and LSTM | |
CN116484189A (en) | ERA5 precipitation product downscaling method based on deep learning | |
CN115544889A (en) | Numerical mode precipitation deviation correction method based on deep learning | |
CN114219345B (en) | Secondary air quality prediction optimization method based on data mining | |
CN117909888B (en) | Intelligent artificial intelligence climate prediction method | |
CN114819264A (en) | Photovoltaic power station irradiance ultra-short term prediction method based on space-time dependence and storage medium | |
CN116826737A (en) | Photovoltaic power prediction method, device, storage medium and equipment | |
Li et al. | An Integrated Artificial Neural Network-based Precipitation Revision Model. | |
CN118095034B (en) | Hydrologic forecasting method and system based on hydrologic model and LSTM dual coupling under influence of water taking activity | |
CN117172037B (en) | Distributed hydrologic forecasting method, device, computer equipment and medium | |
CN114330641A (en) | Method for establishing short-term wind speed correction model based on deep learning of complex terrain | |
Raja et al. | Drought prediction and validation for desert region using machine learning methods | |
CN117076738A (en) | Medium-long term weather prediction method, system and medium based on variable grid model | |
CN115496128A (en) | Urban waterlogging risk forecasting method based on raininess-raininess characteristic parameter combined distribution |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |