CN110648014B - Regional wind power prediction method and system based on space-time quantile regression - Google Patents
Regional wind power prediction method and system based on space-time quantile regression Download PDFInfo
- Publication number
- CN110648014B CN110648014B CN201910802636.0A CN201910802636A CN110648014B CN 110648014 B CN110648014 B CN 110648014B CN 201910802636 A CN201910802636 A CN 201910802636A CN 110648014 B CN110648014 B CN 110648014B
- Authority
- CN
- China
- Prior art keywords
- wind power
- model
- quantile regression
- regional
- data
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000010248 power generation Methods 0.000 claims abstract description 30
- 238000010586 diagram Methods 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000012360 testing method Methods 0.000 claims abstract description 20
- 238000012795 verification Methods 0.000 claims abstract description 14
- 238000013528 artificial neural network Methods 0.000 claims description 32
- 238000011176 pooling Methods 0.000 claims description 25
- 238000013527 convolutional neural network Methods 0.000 claims description 22
- 230000015654 memory Effects 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 7
- 238000011156 evaluation Methods 0.000 claims description 7
- 230000007787 long-term memory Effects 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 239000000284 extract Substances 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 3
- 238000003860 storage Methods 0.000 claims description 3
- 238000012417 linear regression Methods 0.000 claims 1
- 230000006403 short-term memory Effects 0.000 claims 1
- 238000009826 distribution Methods 0.000 description 23
- 230000008569 process Effects 0.000 description 10
- 230000006870 function Effects 0.000 description 8
- 238000000605 extraction Methods 0.000 description 6
- 230000008901 benefit Effects 0.000 description 5
- 230000007774 longterm Effects 0.000 description 5
- 230000000306 recurrent effect Effects 0.000 description 5
- 230000001419 dependent effect Effects 0.000 description 3
- 238000005315 distribution function Methods 0.000 description 3
- 108090000672 Annexin A5 Proteins 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 241000039077 Copula Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
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
- 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
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- 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
- G06N3/049—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
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- 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
- Y02A30/00—Adapting or protecting infrastructure or their operation
-
- 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
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Computational Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Pure & Applied Mathematics (AREA)
- Operations Research (AREA)
- Biomedical Technology (AREA)
- Marketing (AREA)
- Mathematical Optimization (AREA)
- Artificial Intelligence (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Mathematical Analysis (AREA)
- Computing Systems (AREA)
- Public Health (AREA)
- Databases & Information Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Probability & Statistics with Applications (AREA)
- Evolutionary Biology (AREA)
- Algebra (AREA)
- Bioinformatics & Computational Biology (AREA)
- Development Economics (AREA)
Abstract
The invention provides a regional wind power prediction method and system based on spatio-temporal quantile regression, which comprises the steps of collecting operation and numerical weather forecast data in a preset time period of a plurality of wind power plants, converting the collected data into a characteristic diagram, and establishing a training set, a verification set and a test set; establishing a space-time quantile regression model, and training and optimizing the model by utilizing a training set, a verification set and a test set; collecting operation data and environment data of each wind power plant in real time, and predicting regional wind power generation in a certain period of time in the future according to an optimized spatio-temporal quantile regression model; according to the method, the short-term unparameterized probability prediction is carried out on the regional wind power through the space-time quantile regression model, the problem of selection of the interpretation variables in the regional wind power prediction when large input information is carried out is solved, the accuracy and the reliability of the prediction are greatly improved, and a specific solution is provided for the regional wind power generation probability prediction with big data.
Description
Technical Field
The disclosure relates to the technical field of wind power prediction, in particular to a regional wind power prediction method and system based on space-time quantile regression.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
Because wind power generation has the characteristics of intermittency and volatility, high-proportion wind power can form a serious challenge for safe and stable operation of a power system. And the accurate short-term wind power prediction can effectively relieve the adverse effect of wind power. Previous research has focused primarily on point prediction for wind power generation. However, due to the inherent intermittency and volatility of wind, prediction errors inevitably exist. Alternatively, probabilistic predictions may quantify wind uncertainty information, such as confidence intervals, probability distribution functions, quantiles, and the like. The probabilistic wind power prediction method can be divided into parametric prediction and nonparametric prediction according to whether the distribution type of the probability is assumed.
The inventor of the present disclosure finds that wind power is mainly predicted at present by using parameterized probability prediction, and usually random variables are assumed to obey certain functional distribution, such as beta distribution, lognormal distribution and gaussian distribution. However, the actual wind power may not meet the assumed distribution, or even any known distribution, and unreasonable distribution assumptions may directly result in a deviation of the probabilistic prediction result.
Moreover, most of the existing wind power probability prediction methods aim at a single wind power plant, the regional wind power plants are rarely considered, and some researchers provide a copula function-based method for predicting the error distribution of the total wind power of a plurality of wind power plants, however, the method is based on the point prediction result of each single wind power plant, which may cause the accumulation of prediction errors and the distortion of the distribution prediction result; meanwhile, due to typical regional characteristics of wind resources, regions with rich wind resources can be distributed into a plurality of wind power plants, so that strong spatiotemporal and nonlinear correlations exist among the plurality of wind power plants, and due to the large data set and the complex correlations, regional wind power prediction is more complex than that of a single wind power plant, so that great difficulty is brought to the accuracy and reliability of regional wind power prediction.
Disclosure of Invention
In order to overcome the defects of the prior art, the regional wind power prediction method and system based on space-time quantile regression are provided, the problem of selection of interpretation variables in regional wind power prediction when large input information is available is solved, accuracy and reliability of wind power prediction are greatly improved, and a specific solution is provided for regional wind power generation probability prediction with big data characteristics.
In order to achieve the purpose, the following technical scheme is adopted in the disclosure:
in a first aspect, the present disclosure provides a regional wind power prediction method based on spatio-temporal quantile regression;
a regional wind power prediction method based on space-time quantile regression comprises the following steps:
collecting operation and numerical weather forecast data in a preset time period of a plurality of wind power plants, converting the collected data into a characteristic diagram, and establishing a training set, a verification set and a test set;
establishing a space-time quantile regression model, training the model by using a training set, diagnosing the adaptability of the model and optimizing the hyper-parameters of the model by using a verification set, evaluating the reliability and the sharpness of the model by using a test set, and further optimizing the model according to an evaluation result;
and acquiring operation data and numerical weather forecast data of each wind power plant in real time, and predicting regional wind power generation in a certain period of time in the future according to the optimized spatio-temporal quantile regression model.
As some possible implementation manners, the operation and historical numerical weather forecast data in the preset time period of the plurality of wind farms are any data related to wind power generation, including but not limited to power generation, wind direction, wind speed, temperature, humidity and air pressure of each wind farm.
As some possible implementation manners, the acquired data is converted into a feature map, specifically: and calculating the correlation between each wind power plant and regional wind power, selecting the wind power plant with the maximum correlation coefficient as a basis, and arranging the wind power plants in a descending order according to the distance between each wind power plant and the basic wind power plant to generate a characteristic diagram.
By way of further limitation, a series of time-continuous feature maps serve as a set of input feature maps, and the time-continuous feature maps reflect the spatial relative position and time correlation between wind farms.
As possible implementation modes, the space-time quantile regression model comprises a mixed neural network and a quantile regression algorithm, key features are extracted by using the mixed neural network, and a nonlinear regression model is constructed; and (3) according to the quantile regression algorithm, enabling the constructed nonlinear regression model to obtain a corresponding quantile, and estimating a space-time quantile regression model by a gradient descent method to realize probability prediction of regional wind power generation.
As a further limitation, the hybrid neural network is a serial neural network combining a convolutional neural network and a long-term memory neural network, and the convolutional neural network comprises an input layer, a convolutional layer, a pooling layer, a full-link layer and an output layer;
the feature graph enters the convolutional layer through the input layer and is convolved with the convolution kernel, the convolutional layer outputs a convolution result to the pooling layer, and the pooling layer further reduces the dimension and extracts features of the convolution result through sampling by adopting a maximum pooling method;
and after the processing of the convolutional layer and the pooling layer, extracting and converting the input features into high-level information features, classifying the high-level information features by using the connecting layer, and outputting the classified high-level information features to the long-time memory neural network for optimization through the output layer to obtain a final recognition result.
As a further limitation, the convolutional neural network is modeled by two convolutional layers and two pooling layers, the number of convolutional kernels of the two convolutional layers is 16 and 32 in sequence, the format is 4 × 4, kernels in the pooling layers are designed to be 2 × 2, high-level information features are extracted into the long-term memory neural network through convolution and pooling, and a final prediction result is obtained after optimization.
In a second aspect, the present disclosure provides a regional wind power prediction system based on spatio-temporal quantile regression;
a regional wind power prediction system based on spatio-temporal quantile regression comprises:
a data acquisition and pre-processing module configured to: collecting operation and numerical weather forecast data in a preset time period of a plurality of wind power plants, converting the collected data into a characteristic diagram, and establishing a training set, a verification set and a test set;
a model building module configured to: establishing a space-time quantile regression model, training the model by using a training set, diagnosing the adaptability of the model and optimizing the hyper-parameters of the model by using a verification set, evaluating the reliability and the sharpness of the model by using a test set, and further optimizing the model according to an evaluation result;
a prediction module configured to: and acquiring operation data and numerical weather forecast data of each wind power plant in real time, and predicting regional wind power generation in a certain period of time in the future according to the trained and optimized spatio-temporal quantile regression model.
In a third aspect, the present disclosure provides a computer-readable storage medium, on which a computer program is stored, which when executed by a processor, implements the steps in the method for regional wind power prediction based on spatio-temporal quantile regression according to the present disclosure.
In a fourth aspect, the present disclosure provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor implements the steps in the method for regional wind power prediction based on spatio-temporal quantile regression according to the present disclosure when executing the program.
Compared with the prior art, the beneficial effect of this disclosure is:
1. the method comprises the steps of reorganizing input high-dimensional data into a feature map, extracting features by a Hybrid Neural Network (HNN), extracting representative features by using the advantages of the hybrid neural network, constructing a nonlinear regression model, enabling the model to obtain quantiles according to the rules of Quantile Regression (QR) and carrying out probability prediction on regional wind power generation.
2. The method realizes regional wind power prediction, and is different from prediction of a single wind power plant, the prediction of wind power generation in the whole region can be better realized, and the influence of the intermittency and volatility of the wind power generation and the high permeability of the wind power on the safe and stable operation of a power system in the wind power plant in the whole region is reduced, so that an operator can be guided to make a decision effectively from the macroscopic view of power grid operation, and the stability of the power system is improved.
3. According to the method, the SQR which is a new regional wind power prediction unparameterized probability method is utilized, the advantages of QR and HNN are utilized, representative features are extracted, long-term relevant information is recorded, different quantiles are obtained, uncertainty of regional wind power prediction can be accurately and comprehensively described, and accurate and reliable prediction of regional wind power is achieved.
4. According to the method, the time series characteristic diagram is constructed for the high-dimensional input data, the relative position and the time correlation of the space between the wind power plants are reflected, and preparation is made for HNN characteristic extraction, so that the effectiveness and the flexibility of the method are further improved.
5. In the SQR method adopted by the disclosure, HNN is formed by serial connection between CNN and LSTM, namely HNN can be divided into two layers, namely CNN layer and LSTM layer, therefore HNN not only can effectively process high-dimensional data, but also has long-term memory capacity, and SQR combined by QR and HNN can better predict nonparametric probability of regional wind power generation.
Drawings
Fig. 1 is a flowchart of a regional wind power prediction method based on spatio-temporal quantile regression according to embodiment 1 of the present disclosure.
FIG. 2 is a time sequence of characteristic patterns of input data according to embodiment 1 of the present disclosure
Fig. 3 is a schematic structural diagram of CNN according to embodiment 1 of the present disclosure.
Fig. 4 is a schematic diagram of a recursive structure of the RNN according to embodiment 1 of the present disclosure.
Fig. 5 is a schematic diagram of basic units of an LSTM network according to embodiment 1 of the present disclosure.
Fig. 6 is a schematic structural diagram of the SQR according to embodiment 1 of the present disclosure.
Fig. 7 is a schematic diagram of the location distribution of 10 wind power plants in a certain area according to embodiment 1 of the present disclosure.
Fig. 8 is a schematic diagram of a normalized regional wind farm time series according to embodiment 1 of the present disclosure.
Fig. 9 is a schematic diagram of a correlation between single wind power and regional wind power according to embodiment 1 of the present disclosure.
Fig. 10 is a characteristic pattern of one of the characteristic diagrams described in embodiment 1 of the present disclosure.
Fig. 11 is a schematic time-varying result diagram of a regional wind farm reliability prediction model according to embodiment 1 of the present disclosure.
Fig. 12 is a schematic diagram of a time-varying result of a regional wind farm sharpness prediction model according to embodiment 1 of the present disclosure.
Fig. 13 is a schematic diagram illustrating a result of predicting wind power in a 72-hour future area by using QRNN according to embodiment 1 of the present disclosure.
Fig. 14 is a schematic diagram illustrating a result of predicting wind power in a 72-hour future region by using SQR according to embodiment 1 of the present disclosure.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Example 1:
as shown in fig. 1, an embodiment 1 of the present disclosure provides a regional wind power prediction method based on spatio-temporal quantile regression, which includes the following steps:
collecting operation and numerical weather forecast data in a preset time period of a plurality of wind power plants, converting the collected data into a characteristic diagram, and establishing a training set, a verification set and a test set;
establishing a space-time quantile regression model, training the model by using a training set, diagnosing the adaptability of the model and optimizing the hyper-parameters of the model by using a verification set, evaluating the reliability and the sharpness of the model by using a test set, and further optimizing the model according to an evaluation result;
and acquiring the operation data and the environment data of each wind power plant in real time, and predicting regional wind power generation in a certain period of time in the future according to the optimized spatio-temporal quantile regression model.
The embodiment combines the advantages of a Hybrid Neural Network (HNN) and a Quantile Regression (QR) to provide a space-time quantile regression model (SQR) for short-term unparameterized probability prediction of regional wind power, the method reorganizes input high-dimensional data into a feature map, performs feature extraction by the hybrid neural network, extracts representative features by using the advantages of the hybrid neural network, constructs a nonlinear regression model, obtains quantiles according to the quantile regression rule, performs probability prediction on regional wind power generation, solves the problem of selection of interpretation variables when large input information exists, and provides a specific solution for regional wind power generation probability prediction with large data.
As described in the background art, currently, wind power is mainly predicted by using parameterized probability prediction, and it is generally assumed that random variables obey certain functional distribution, such as beta distribution, lognormal distribution, and gaussian distribution. However, the actual wind power may not meet the assumed distribution, or even any known distribution, and unreasonable distribution assumptions may directly result in a deviation of the probabilistic predictive result. In the embodiment, aiming at the deficiency of the parameter probability prediction, a non-parametric probability prediction method is provided for estimating a probability density function, a confidence interval and a quantile without making any assumption on the type of probability distribution.
Quantile Regression (QR) is a common non-parametric probability prediction method, can effectively realize non-parametric probability prediction of wind power generation, and is widely applied to wind power non-parametric probability prediction.
The HNN is a serial neural network combining CNN and LSTM (long-term memory recurrent neural network), and can process massive data sets and record long-term dependence information.
The method can be used for estimating a typical QR model by HNN based on gradient enhancement, QR can obtain different proportions and completely describes the distribution of dependent variables, and the HNN can extract representative space-time characteristics according to the structures of CNN and LSTM layers.
Therefore, the unique advantage of SQR is that its effectiveness depends only on the framework of HNN, which can overcome the disadvantage that typical QR can only fit linear correlations and lose the ability to process large data. Further, the feature map is constructed based on high-dimensional input data, and feature extraction by HNN is possible. On the basis, the embodiment provides the application of the SQR method in non-parametric probability prediction, and the SQR method is applied to uncertainty behavior estimation of regional wind power prediction to comprehensively obtain different quantiles and dependent variable distribution.
The detailed method comprises the following steps:
(1) input format design
The CNN layer of HNN is a multilayer neural network, which is good at processing larger images, and successfully solves the problem that high-dimensional images with large data volume are difficult to identify through the CNN neural network. Given that HNNs are good at dealing with these problems, high-dimensional input data is reorganized into feature maps. The positions of the wind power plants are arranged by calculating the correlation among the wind power plants, and the potential spatial correlation is embodied. In addition, LSTM is a recurrent neural network that can learn long-term dependent information. The method is suitable for classification, processing and prediction of unknown information with long time lag between major events by using experience in time series. Therefore, it is necessary to reflect the time correlation by using the feature map, and each set of input feature maps must be continuous in time.
Assuming that there are m wind farms in a region, W ═ W1,w2,…,wm) Then, the correlation between each wind power plant and regional wind power is calculated, the wind power plant with the largest correlation coefficient is selected as a base, and the wind power plants are arranged in a descending order according to the distance between each wind power plant and the base wind power plant, so that a characteristic diagram can be generated. A series of time-continuous feature maps form a set of input feature maps, as shown in fig. 2 (m 10). These columns represent wind farms, the rank family represents the type of variable, the numerical weather forecast (NWP) requires historical data 15 minutes ahead, and the specific meaning of the abbreviations in fig. 2 is shown in table 1.
Table 1: specific meanings of abbreviations in FIG. 2
(2) Quantile regression
Based on the traditional conditional mean regression, quantile regression is developed. The quantile regression algorithm is a non-parametric model, does not assume the distribution of observed values, and has stronger robustness on abnormal data in a sample. It covers all features except mean regression that includes different quantiles.
Let Y be the response variable and X be the real-valued parameter (possibly high-dimensional data). Given a conditional distribution function F (y | X ═ X), the equation can be expressed as:
F(y|X=x)=P(Y≤y|X=x) (1)
for the distribution function F, the quantile of α can be defined as qα(x):
qα(x)=inf{y:|F(y|X=x)≥α} (2)
Traditionally, the quantile q of αα(x) Can be converted into a linear optimization problem:
qα(x)=βX′ (3)
wherein β ═ β1,β1,…,βn],,X=[x1,x2,…,xn]。
As with the least squares estimation of conditional mean regression, quantile regression can be achieved by minimizing a squared error test function. The following test functions are given:
where y is a real-valued response variable and the quantile of α is q.
Therefore, the conditional quantiles make the desired loss function E (L)α) And (3) minimizing:
then, establishing a prediction interval by using the quantile regression result, wherein if 90% of prediction intervals are given:
PI(x)=[q0.05(x),q0.95(x)] (6)
(3) convolutional neural network
CNN, a feed forward network of a deep structure, is calculated by convolution operations. Generally, a CNN consists of an input layer, a feature extraction layer consisting of convolutional layers and pooling layers, and a fully-connected layer consisting of fully-linked multi-layer perceptron classifiers.
When the common neural network is applied to regional wind power prediction, the problems of excessive parameters, overlarge model, consumption of computing resources and the like can be caused. Compared with the common neural network, the CNN has the feature layer composed of the convolutional layer and the pooling layer, and the problem of the common neural network is overcome by sharing the conventional kernel in the feature layer, and the convolution and pooling process is shown in fig. 3.
The feature map passes through the input layer into the convolution layer and is convolved with the convolution kernel, which is a weight sharing process. The result of the convolution is the output of the convolutional layer. This process can be described as follows:
wherein, CiIs the output of the convolutional layer, X is the input data, W is the convolution kernel, B is the offset, and f is the active function.
Pooling is a process of further dimension reduction and feature extraction by sampling, and typically, the lower layer samples include a max pool and an average pool. The embodiment adopts a maximum pool method:
Pj=max Ci (8)
where P is the output of the pooling layer.
And extracting and converting input data into high-level information characteristics through the previous processing of the convolutional layer and the pooling layer, and classifying the high-level information characteristics by the connecting layer to obtain a final identification result.
(4) Long and short time memory recursion neural network
Unlike conventional neural networks, RNN (recurrent neural network), as a recursive network, can record historical information through a series of unit rings in hidden layers, and predict future information through historical experience. However, when the historical information is far away from the current information and the time interval is larger and larger, the RNN loses the linking capability with the long-term information, the gradient disappears or explodes, the recursive structure of the RNN is shown in fig. 4, and when data is input into the unit loop a, h can be output through certain calculation, and then the next unit loop is entered. Obviously, as the number of cells increases, the learning ability of RNN becomes worse.
LSTM (long and short memory recurrent neural network), proposed in 1997 by Hochreiter and Schmidhuber, is an improved RNN that overcomes RNN's shortfalls in learning long-term association information. By setting the forgetting gate, the LSTM solves the problem of RNN gradient disappearance or explosion on the time axis. As shown in fig. 5, the LSTM network consists of an input gate, a forgetting gate, and an output gate, and three thresholds cooperatively process feature selection, forgetting, and memory.
The forget gate decides to select some messages to serve the latter, while the other is discarded.
ftIs done by an s-shaped function:
ft=σ(Wf·[ht-1,xt]+bf) (9)
the useful data is updated in the input gate. ComputingThe value is to update the old state and then get the new state.
This process can be described as follows:
it=σ(Wi·[ht-1,xt]+bi) (10)
the desired output is finally achieved in the output gate:
ot=σ(Wo·[ht-1,xt]+b0) (13)
ht=ot·tanh(Ct) (14)
(5) spatio-temporal quantile regression
This example presents an SQR method based on regional wind power non-parametric probability prediction of QR and HNN, which are formed by the serial connection between CNN and LSTM. That is, the HNN may be divided into two layers, a CNN layer and an LSTM layer. Therefore, the HNN can not only effectively process high-dimensional data, but also have long-term memory ability. Therefore, the SQR of the QR and HNN combination can well predict the non-parametric probability of regional wind power generation, and the SQR structure is shown in FIG. 6.
(6) Evaluation of
1) Reliability of
Coverage probability is generally considered to be the most important index for evaluating reliability of a prediction interval. If the probability prediction interval is [ L ]i,Ui]Then the coverage probability predicted by the model can be expressed as:
2) Sharpness
The sharpness of the model can be evaluated by predicting the interval normalized mean width (PINAW), as follows:
wherein R is the difference between the maximum value and the minimum value of the target value.
The effectiveness of the proposed SQR in regional wind power plant prediction is verified through tests on 10 wind power plants in Jiangsu province.
Data preprocessing
The data set contains the operating data and the numerical weather forecast of 10 wind farms with the time range of 2016 1/2016 to 2016 12/31/2016 and the time resolution of 15 minutes, and the spatial distribution of the 10 wind farms is shown in FIG. 7.
Due to the influence of atmospheric environment, the wind power generation presents stronger uncertainty and volatility, and the nonlinearity of the wind power generation is obvious. The 72-hour normalized regional wind power time series is shown in fig. 8, and it can be seen that regional wind power exhibits significant uncertainty, volatility and nonlinearity.
To find the basis in regional wind farms, the relationship of single wind to regional wind is calculated, as shown in fig. 9. The wind power plant with the maximum correlation coefficient is selected as a foundation, and meanwhile, as a data driving method, the SQR emphasizes the continuity of input data so as to ensure that the model can generate more detailed description on the system, and therefore the accuracy of the model is improved. Thus, reasonable explanatory variables can further improve accuracy. Theoretically, any state variable (such as temperature, humidity, air pressure, wind speed and wind direction) related to wind power can be added into the model as an explanatory variable, but the explanatory variable with strong correlation with wind power is more valuable, and the embodiment also utilizes sine and cosine values of wind direction, wind speed, temperature, atmosphere and pressure as the explanatory variable besides power; a profile was generated using an Excel tool as shown in fig. 10.
(II) model construction and description
Strictly speaking, the SQR consists of three parts: CNN layer, LSTM layer, and QR. The method comprises the steps of forming a predictor for feature extraction and time memory by a CNN layer and an LSTM layer, and estimating a quantile regression model through gradient enhancement to obtain a probability prediction result. The CNN layer is modeled by two convolutional layers and two pooling layers. The number of convolution kernels is 16 and 32 in order, and the format is 4 x 4. At the pooling level, the kernel is designed to be 2 x 2. Extracting the key features into the LSTM layer through convolution and pooling, and then optimizing to finally obtain a result.
(III) analysis of the results of the examples
In order to verify the prediction performance of the proposed regional wind power prediction SQR model, a QRNN model (fractional bit regression + recurrent neural network) is used for estimating a prediction interval based on the same original data to perform benchmark test comparison. In order to fully estimate regional wind power fluctuations, the present embodiment calculates 0.025, 0.05, 0.09, and 0.975 quantiles, which may constitute prediction intervals of 95% and 85%.
In order to verify the reliability of the proposed model, the wind power test set of 11 months is predicted by using an SQR model and a QRNN model respectively, and the prediction time is 72 hours. Fig. 11 is a reliability index PICP of the SQR model and the QRNN model at 95% and 85% confidence levels for evaluating the reliability of the models, where the smaller the PICP, the higher the reliability. As can be seen from fig. 11, the reliability index of the SQR model is smaller than that of the QRNN model, and the results show that the SQR model is more reliable than the QRNN model in the prediction intervals at 85% and 95% confidence levels.
Figure 12 shows the sharpness of the SQR model and the QRNN model 72 hours ago at 85% and 95% confidence levels. It is used to evaluate the sharpness of the model, in which the smaller the sharpness, the better the sharpness. The resulting quantiles of SQR have a higher level of sharpness compared to the QRNN model, which indicates that the SQR has better resolution for different look-ahead times within the two confidence intervals. The number of sharpness for SQR is on average about 4.5% less than QRNN.
Fig. 13 and 14 are the prediction results of the QRNN model and the SQR model applied to the wind power test set of the 11-month region, respectively. It can be seen that the actual wind power can be well surrounded by the predicted interval, and 85% of the predicted interval can be encapsulated by 95% of the predicted interval, thus proving the high reliability of the method. Fig. 13 and fig. 14 compare, the confidence interval predicted by the SQR model is narrower, which indicates that the SQR has better definition performance. The results are consistent with the comparative analysis results of fig. 11 and 12. Furthermore, the prediction interval becomes wider as the prediction time increases, demonstrating an increase in uncertainty. The SQR is used as a probability prediction model with high learning capacity and has strong self-adaption capacity to the characteristics of a regional wind power plant.
This example investigated the proposed computation speeds of SQR and QRNN, as shown in table 2. Both methods were performed on computers equipped with Intel (R) core (TM) i7-8550U CPU @1 and 8GHz 2.00GHz and 8GB RAMs. The average training time of each model described in this example was counted based on the training set from 2016 month 1 to 2016 month 6 and 1. As can be seen from table 2, the SQR is more than 7 times faster than the QRNN training speed. The SQR model is constructed quickly, and only 2.2 hours of training time are needed. The result shows that the SQR prediction model has higher computational efficiency. Therefore, the method based on the SQR model provided by the embodiment has high practical application potential in the regional wind power probability prediction system.
Table 2: construction of computational models
Due to the high penetrability and the high volatility of wind power, accurate prediction of wind power is the key to improving the efficiency of new energy and economic and safe operation of a power system. Compared with a single wind power plant, the regional wind power prediction is reasonably integrated into a decision model, so that an operator can effectively make decisions, such as making a power generation plan, reserving electric quantity and the like. Regional wind power prediction has a large impact on the operation and control of the power system. In addition, wind power probability prediction is more popular with the unavailability of wind power prediction errors. The SQR composed of QR and HNN is used as non-parametric probability prediction of regional wind power generation, the SQR can explore space-time and non-linear relations among regional wind power generation fields to obtain different prediction quantiles, in order to further improve precision, a time sequence of a characteristic diagram is constructed, and finally a more accurate prediction result is obtained.
The effectiveness of the SQR in the future 72 hours is verified on the basis of the measured data of 10 wind power plants in Jiangsu province. The SQR is proved to be more effective and advanced than the QRNN through comprehensive comparison with the QRNN. Experimental results show that the SQR has good global and local self-adaptive capacity and high calculation efficiency. Therefore, the SQR provides an efficient and flexible framework for regional wind power probability prediction and has high reliability.
Example 2:
the embodiment 2 of the present disclosure provides a regional wind power prediction system based on spatio-temporal quantile regression, including:
a data acquisition and pre-processing module configured to: collecting operation and numerical weather forecast data in a preset time period of a plurality of wind power plants, converting the collected data into a characteristic diagram, and establishing a training set, a verification set and a test set;
a model building module configured to: establishing a space-time quantile regression model, training the model by using a training set, diagnosing the adaptability of the model and optimizing the hyper-parameters of the model by using a verification set, evaluating the reliability and the sharpness of the model by using a test set, and further optimizing the model according to an evaluation result;
a prediction module configured to: and acquiring the operation data and the environment data of each wind power plant in real time, and predicting regional wind power generation within a certain period of time in the future according to the trained and optimized spatio-temporal quantile regression model.
Example 3:
the embodiment 3 of the present disclosure provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps in the regional wind power prediction method based on spatio-temporal quantile regression according to the embodiment 1 of the present disclosure.
Example 4:
the embodiment 4 of the present disclosure provides a computer device, which includes a memory, a processor, and a computer program that is stored in the memory and can be run on the processor, where the processor implements the steps in the regional wind power prediction method based on spatio-temporal quantile regression according to embodiment 1 of the present disclosure when executing the program.
The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure, and various modifications and changes may be made to the present disclosure by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.
Claims (9)
1. A regional wind power prediction method based on spatio-temporal quantile regression is characterized by comprising the following steps:
collecting operation data and weather forecast data in a preset time period of a plurality of wind power plants, converting the collected data into a characteristic diagram, and establishing a training set, a verification set and a test set; wherein, convert the data of gathering into the characteristic map, specifically be: calculating the correlation between each wind power plant and regional wind power, selecting the wind power plant with the maximum correlation coefficient as a basis, and arranging the wind power plants in a descending order according to the distance between each wind power plant and the basic wind power plant to generate a characteristic diagram;
establishing a space-time quantile regression model, training the model by using a training set, diagnosing the adaptability of the model and optimizing the hyper-parameters of the model by using a verification set, evaluating the reliability and the sharpness of the model by using a test set, and further optimizing the model according to an evaluation result;
and acquiring the operation data and the environment data of each wind power plant in real time, and predicting regional wind power generation in a certain period of time in the future according to the optimized spatio-temporal quantile regression model.
2. The regional wind power prediction method based on spatio-temporal quantile regression as claimed in claim 1, wherein the operation data and weather forecast data in the preset time period of the plurality of wind farms are any data related to wind power generation, including power generation, wind direction, wind speed, temperature, humidity and air pressure of each wind farm.
3. The regional wind power prediction method based on spatio-temporal quantile regression as claimed in claim 1, characterized in that a series of continuous time feature maps are used as a set of input feature maps, and the time continuous feature maps are used for reflecting the correlation between the relative position of space and time between wind power plants.
4. The regional wind power prediction method based on spatio-temporal quantile regression of claim 1, wherein the spatio-temporal quantile regression model comprises a mixed neural network and a fractional regression algorithm, and a non-linear regression model is constructed by extracting a representative feature map by using the mixed neural network; and obtaining quantiles of the constructed nonlinear regression model according to a quantile regression algorithm, and estimating a space-time quantile regression model through gradient enhancement to realize probability prediction of regional wind power generation.
5. The regional wind power prediction method based on spatio-temporal quantile regression of claim 4, wherein the hybrid neural network is a serial neural network combining a convolutional neural network and a long-and-short-term memory neural network, and the convolutional neural network comprises an input layer, a convolutional layer, a pooling layer, a connection layer and an output layer;
the feature graph penetrates through the input layer and enters the convolution layer, convolution is carried out on the feature graph and a convolution kernel, the convolution layer outputs a convolution result to the pooling layer, and the pooling layer adopts a maximum pooling method to further reduce the dimension and extract features of the convolution result through sampling;
and after the processing of the convolutional layer and the pooling layer, extracting and converting the input features into high-level information features, classifying the high-level information features by using the connecting layer, and outputting the classified high-level information features to the long-time memory neural network for optimization through the output layer to obtain a final recognition result.
6. The regional wind power prediction method based on spatio-temporal quantile regression as claimed in claim 5, wherein the convolutional neural network is modeled by two convolutional layers and two pooling layers, the number of convolutional kernels of the two convolutional layers is 16 and 32 in sequence, the format is 4 x 4, the kernel in the pooling layer is designed to be 2 x 2, high-level information features are extracted into the long-term and short-term memory neural network through convolution and pooling, and a final prediction result is obtained after optimization.
7. A regional wind power prediction system based on spatio-temporal quantile regression is characterized by comprising the following components:
a data acquisition and pre-processing module configured to: collecting operation data and weather forecast data in a preset time period of a plurality of wind power plants, converting the collected data into a characteristic diagram, and establishing a training set, a verification set and a test set; wherein, convert the data of gathering into the characteristic map, specifically be: calculating the correlation between each wind power plant and regional wind power, selecting the wind power plant with the maximum correlation coefficient as a basis, and arranging the wind power plants in a descending order according to the distance between each wind power plant and the basic wind power plant to generate a characteristic diagram;
a model building module configured to: establishing a space-time quantile regression model, training the model by using a training set, diagnosing the adaptability of the model and optimizing the hyper-parameters of the model by using a verification set, evaluating the reliability and the sharpness of the model by using a test set, and further optimizing the model according to an evaluation result;
a prediction module configured to: and acquiring the operation data and the environment data of each wind power plant in real time, and predicting regional wind power generation within a certain period of time in the future according to the trained and optimized spatio-temporal quantile regression model.
8. A computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, carries out the steps of the spatio-temporal quantile regression-based regional wind power prediction method according to any one of claims 1 to 6.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements the steps in the spatio-temporal quantile regression based regional wind power prediction method according to any one of claims 1 to 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910802636.0A CN110648014B (en) | 2019-08-28 | 2019-08-28 | Regional wind power prediction method and system based on space-time quantile regression |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910802636.0A CN110648014B (en) | 2019-08-28 | 2019-08-28 | Regional wind power prediction method and system based on space-time quantile regression |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110648014A CN110648014A (en) | 2020-01-03 |
CN110648014B true CN110648014B (en) | 2022-04-15 |
Family
ID=68991053
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910802636.0A Active CN110648014B (en) | 2019-08-28 | 2019-08-28 | Regional wind power prediction method and system based on space-time quantile regression |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110648014B (en) |
Families Citing this family (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111612262A (en) * | 2020-01-15 | 2020-09-01 | 长沙理工大学 | Wind power probability prediction method based on quantile regression |
CN112037106B (en) * | 2020-08-07 | 2023-12-15 | 汉威科技集团股份有限公司 | Data anomaly analysis method based on feature cross-correlation and probability density |
CN112215392A (en) * | 2020-08-27 | 2021-01-12 | 国电电力云南新能源开发有限公司 | Method for predicting power generation capacity of wind power medium and long term region based on equipment state and environmental factors |
CN114186469A (en) * | 2020-09-14 | 2022-03-15 | 青岛海信日立空调系统有限公司 | Air conditioning system energy efficiency prediction method and air conditioning system |
CN112257928A (en) * | 2020-10-22 | 2021-01-22 | 国网山东省电力公司潍坊供电公司 | Short-term power load probability prediction method based on CNN and quantile regression |
CN112308322B (en) * | 2020-11-03 | 2023-11-24 | 华翔翔能科技股份有限公司 | Multi-wind-field space-time wind speed prediction method and device and electronic equipment |
CN112990536B (en) * | 2021-01-27 | 2022-06-03 | 广东工业大学 | Bus arrival time prediction method based on quantile convolution network |
CN112580899A (en) * | 2021-02-23 | 2021-03-30 | 国能日新科技股份有限公司 | Medium-and-long-term wind power generation prediction method and system fused with machine learning model |
CN114021853A (en) * | 2021-11-26 | 2022-02-08 | 广东电网有限责任公司广州供电局 | Wind power prediction method and device, electronic equipment and storage medium |
CN114298444B (en) * | 2022-03-09 | 2022-05-27 | 广东电网有限责任公司佛山供电局 | Wind speed probability prediction method, device, equipment and storage medium |
CN115907233B (en) * | 2023-01-06 | 2023-05-09 | 北京科技大学 | Wind power generation quantile prediction method based on machine mental model and self-attention |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106803129A (en) * | 2015-11-26 | 2017-06-06 | 中国电力科学研究院 | A kind of wind power ensemble prediction method based on multi-source numerical weather forecast |
CN109657839A (en) * | 2018-11-22 | 2019-04-19 | 天津大学 | A kind of wind power forecasting method based on depth convolutional neural networks |
CN109840308A (en) * | 2017-11-29 | 2019-06-04 | 中国电力科学研究院有限公司 | A kind of region wind power probability forecast method and system |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2010016110A1 (en) * | 2008-08-05 | 2010-02-11 | 富士通株式会社 | Method of creating estimation model by means of multiple regression analysis, creation system and creation program |
-
2019
- 2019-08-28 CN CN201910802636.0A patent/CN110648014B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106803129A (en) * | 2015-11-26 | 2017-06-06 | 中国电力科学研究院 | A kind of wind power ensemble prediction method based on multi-source numerical weather forecast |
CN109840308A (en) * | 2017-11-29 | 2019-06-04 | 中国电力科学研究院有限公司 | A kind of region wind power probability forecast method and system |
CN109657839A (en) * | 2018-11-22 | 2019-04-19 | 天津大学 | A kind of wind power forecasting method based on depth convolutional neural networks |
Non-Patent Citations (2)
Title |
---|
基于时空关联混合模型的交通流预测方法研究;朱勇;《中国优秀博硕士学位论文全文数据库(电子期刊) 工程科技II辑》;20190215;第8-15、26-32页 * |
考虑时空相关性的源荷功率概率建模和概率预测方法;李丹;《中国优秀博硕士学位论文全文数据库(电子期刊) 工程科技II辑》;20180615;第47、96-104页 * |
Also Published As
Publication number | Publication date |
---|---|
CN110648014A (en) | 2020-01-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110648014B (en) | Regional wind power prediction method and system based on space-time quantile regression | |
CN109508360B (en) | Geographical multivariate stream data space-time autocorrelation analysis method based on cellular automaton | |
CN110348624B (en) | Sand storm grade prediction method based on Stacking integration strategy | |
CN108445752B (en) | Random weight neural network integrated modeling method for self-adaptively selecting depth features | |
Hu et al. | Research and application of a hybrid model based on Meta learning strategy for wind power deterministic and probabilistic forecasting | |
CN113554466B (en) | Short-term electricity consumption prediction model construction method, prediction method and device | |
CN112598248B (en) | Load prediction method, load prediction device, computer equipment and storage medium | |
CN113255986B (en) | Multi-step daily runoff forecasting method based on meteorological information and deep learning algorithm | |
CN110689183B (en) | Cluster photovoltaic power probability prediction method, system, medium and electronic device | |
Chen et al. | Probabilistic forecasting of drought: a hidden Markov model aggregated with the RCP 8.5 precipitation projection | |
CN113449919B (en) | Power consumption prediction method and system based on feature and trend perception | |
CN113507118B (en) | Wind power prediction method and system | |
CN114065996A (en) | Traffic flow prediction method based on variational self-coding learning | |
CN117494034A (en) | Air quality prediction method based on traffic congestion index and multi-source data fusion | |
CN116187835A (en) | Data-driven-based method and system for estimating theoretical line loss interval of transformer area | |
CN114357670A (en) | Power distribution network power consumption data abnormity early warning method based on BLS and self-encoder | |
CN115423146A (en) | Self-adaptive runoff forecasting method based on multi-factor nearest neighbor sampling regression and support vector machine | |
CN108830405B (en) | Real-time power load prediction system and method based on multi-index dynamic matching | |
Li et al. | Short-term probabilistic load forecasting method based on uncertainty estimation and deep learning model considering meteorological factors | |
CN117037493A (en) | Short-time traffic flow prediction method based on space-time convolution network | |
CN116341391A (en) | Precipitation prediction method based on STPM-XGBoost model | |
CN116663404A (en) | Flood forecasting method and system coupling artificial intelligence and Bayesian theory | |
CN116561569A (en) | Industrial power load identification method based on EO feature selection and AdaBoost algorithm | |
CN114254828B (en) | Power load prediction method based on mixed convolution feature extractor and GRU | |
CN112446550B (en) | Short-term building load probability density prediction method |
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 |