CN110728391B - Depth regression forest short-term load prediction method based on expandable information - Google Patents
Depth regression forest short-term load prediction method based on expandable information Download PDFInfo
- Publication number
- CN110728391B CN110728391B CN201810781666.3A CN201810781666A CN110728391B CN 110728391 B CN110728391 B CN 110728391B CN 201810781666 A CN201810781666 A CN 201810781666A CN 110728391 B CN110728391 B CN 110728391B
- Authority
- CN
- China
- Prior art keywords
- data
- forest
- regression forest
- stage
- dimensional
- 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 38
- 239000013598 vector Substances 0.000 claims description 54
- 238000012545 processing Methods 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 13
- 230000000694 effects Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000013277 forecasting method Methods 0.000 abstract 1
- 238000004422 calculation algorithm Methods 0.000 description 33
- 238000003066 decision tree Methods 0.000 description 17
- 238000013528 artificial neural network Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 3
- 238000012512 characterization method Methods 0.000 description 3
- 238000002790 cross-validation Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000005192 partition Methods 0.000 description 3
- 238000007637 random forest analysis Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000005728 strengthening 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Marketing (AREA)
- Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Game Theory and Decision Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Primary Health Care (AREA)
- Development Economics (AREA)
- Water Supply & Treatment (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a deep regression forest short-term load forecasting method based on expandable information. The trained deep regression forest model can obtain a high-precision predicted value without manually debugging hyper-parameters, and the method has good generalization capability, does not need deep understanding of multiple specific objects, and only needs to serialize the data.
Description
Technical Field
The invention belongs to the field of short-term load prediction of an electric power system, and relates to an electric power load prediction method based on fusion of multiple information of climate, weather, electric power load, date and large-scale activities, which is suitable for short-term load prediction of the electric power system.
Background
As one of the important daily works of the power dispatching department, load prediction can guide the power production department to economically make a power generation plan and a power system operation mode. Accurate load prediction is beneficial to improving the safety and stability of the power system, reducing the power generation cost and improving the overall benefit of power enterprises. For a long time, scholars at home and abroad make a great deal of research on the theory and method of load prediction. The traditional time series method is taken as a representative in a classical load prediction method, and has the advantages of simple prediction model and small data size required by prediction. However, the method emphasizes the role of the time factor in the prediction, and fades the influence of other external factors, so that the prediction error is large.
Since the 20 th century and the 80 th era, with the development of computer technology and the advent of the big data era, many machine learning-based intelligent prediction algorithms were developed. Meanwhile, in order to improve the accuracy of the load prediction of the power system, power system experts and scholars try to apply an intelligent prediction algorithm to the load prediction of the power system and gradually put forward the theory of modern load prediction. The modern load prediction theory mainly comprises: grey mathematical theory, expert system methods, fuzzy load prediction, artificial neural networks, and the like. Among them, Artificial Neural Networks (ANN) and optimization algorithms thereof are widely used in load prediction. The ANN has the capability of self-learning and associative memory, can fully approximate any complex nonlinear relation, and has strong robustness and fault tolerance. Therefore, the ANN can effectively predict the complex nonlinear problem of the power system load prediction. However, the ANN has the defects of slow training speed and the need of manually setting and adjusting a large number of hyper-parameters. Meanwhile, the ANN is prone to fall into a local optimal solution and even fails to converge on the optimal solution, resulting in prediction misalignment. As the concept of Deep Learning (Deep Learning) is proposed by machine Learning experts and scholars, Deep Neural Networks (DNNs) and their optimization algorithms begin to gradually become representative of Deep Learning at present. The DNN has the capability of characterizing learning in a deep learning theory, and a complex model can be processed by using fewer hyper-parameters. Meanwhile, the DNN adopts a pre-training method to relieve the problem that the algorithm is easy to fall into local optimum. Therefore, DNN has load prediction capability as well, and is also applied to the field of load prediction. However, the DNN still has the disadvantages of slow training speed and the training effect depending on the setting and adjustment of the hyper-parameters by human.
Recently, professor of the university of Nanjing at China's machine Learning And DatA Mining institute (Learning And Mining from DatA, LAMDA) has proposed a Deep Forest algorithm (Deep Forest), also known as a multi-granular Cascade Forest algorithm (gcForest). As an integrated classification algorithm based on decision trees, the experimental prediction capability of the deep forest algorithm can be comparable to that of a deep neural network algorithm. Meanwhile, the deep forest algorithm only needs to set a small amount of hyper-parameters, and does not need to manually adjust a large amount of hyper-parameters in the training process. Experiments show that the default hyper-parameter setting of the deep forest algorithm is suitable for processing different tasks in different fields. Even if the hyper-parameter is not adjusted, an effective prediction effect can be obtained.
Disclosure of Invention
The invention provides a deep regression forest short-term load prediction method based on expandable information. Firstly, establishing a prediction model capable of expanding information; then, extracting historical data, and training the deep regression forest; and finally, predicting the power load based on the real-time data.
The method for predicting the short-term load of the deep regression forest based on the expandable information comprises the following steps:
(1) performing sequence processing on collectable historical data to form a column of data;
(2) in the calculation method, new types of data can be added at any time and can be subjected to serialization processing;
(3) training the deep regression forest by using the data obtained in the step (1) and the step (2);
(4) and predicting the short-term power load by using the real-time data.
The data in step (1) may include climate, weather, historical power load, date and large activity data. If certain data is matrix data, the certain data is serialized, and the method for serializing the matrix A is as follows:
After serialization treatment A' ═ a1,1a1,2…a1,n a2,1a2,2…a2,nam,1am,2…am,n]。
The steps areThe data in step (2) may be emergency data, disaster data, or the like, and therefore, it is necessary to continue to expand the training data to form a ″ ═ a' a1a2…am′]。
The step (3) comprises a multi-granularity Scanning stage (multi-granular Scanning) and a Cascade Regression Forest stage (Cascade Regression Forest).
And (4) extracting sample features in the multi-granularity scanning stage in the step (3), and mining the sequence relation of the sequence data features as much as possible. And setting the total length of the serialized data to be m '-m' + m multiplied by n, setting an n 'dimensional vector window to perform sliding value taking on the original feature vector, and obtaining m' -n '+ 1 n' dimensional vectors when the step length is defaulted to 1. Then, the obtained vectors are classified by two different types of forest models respectively to obtain m '-n' +1 2-dimensional classification vectors respectively. Finally, all the classification vectors are spliced into a 4(m '-n' +1) dimensional feature vector in sequence to be used as the input of the cascade regression forest.
And (4) each level of the cascade regression forest stage in the step (3) is composed of a plurality of forest models of different types. The deep regression forest algorithm utilizes the cascade regression forest stage to process the data characteristics layer by layer, thereby enhancing the characterization learning capability of the algorithm and being beneficial to improving the prediction accuracy. In the cascade regression forest stage, each stage acquires the processed characteristic information from the previous stage, and generates new characteristic information by using the characteristic information to transmit the new characteristic information to the next stage. Except that the first stage directly adopts the feature vector after the multi-granularity scanning processing as the input, each subsequent stage splices the feature result vector output by the previous stage with the original input feature vector as the input of the stage. Firstly, after the feature vectors are classified and processed by two different types of forest models, two 2-dimensional class vectors are obtained. The two 2-dimensional class vectors in the deep regression forest can effectively reflect the characteristics of the sample and are called enhancement characteristic vectors. Next, the enhanced feature vector is spliced with the original feature vector of 4(m "-n" +1) dimensions to form a feature vector of 4(m "-n" +2) dimensions. Then, the 4(m "-n" +2) -dimensional feature vector with the enhanced features is taken as the input vector of the next stage. And the method is carried out until the last stage of the cascade regression forest is reached. And performing regression processing on the category generated by the last stage and the category value thereof.
And (4) predicting the short-term power load in the step (4), namely, using the real-time information data as output and using the output of the over-depth regression forest as predicted power load.
Compared with the prior art, the invention has the following advantages and effects:
(1) the invention is inspired by deep learning theories such as a deep neural network and a deep forest algorithm, and sets two stages of multi-granularity scanning and cascade regression forest. The algorithm has the capability of processing the characterization relation and the capability of strengthening the characterization learning layer by layer. The deep regression forest algorithm is used as an integrated algorithm based on a decision tree, and the problem of high difficulty in determining the hyper-parameters of the deep neural network is solved.
(2) The invention adopts the real load value and weather data of a certain area and tests the short-term load prediction capability of the deep regression forest algorithm. As a prediction algorithm, the deep regression forest algorithm can effectively predict the specific value of the load and the change trend of the load, and has a low prediction error.
(3) Under the condition of keeping the super-parameter setting unchanged, the invention proves that the deep regression forest algorithm has effective short-term load prediction capability under different data scales through experiments. The deep regression forest algorithm can achieve high load prediction accuracy by using small-scale prediction samples. Therefore, the deep regression forest algorithm can effectively process data sets with different scales, excavate the relation among all data of the power system and improve the short-term load prediction effect.
Drawings
FIG. 1 is a diagram of a multi-granularity scan process of the method of the present invention.
FIG. 2 is a diagram of a cascaded regression forest of the method of the present invention.
FIG. 3 is a depth regression forest ensemble prediction process diagram of the method of the present invention.
Detailed Description
The invention provides a depth regression forest short-term load prediction method based on expandable information, which is described in detail in the following steps in combination with the attached drawings:
FIG. 1 is a diagram of a multi-granularity scan process of the method of the present invention. It is assumed in fig. 1 that there is one sample with a 200-dimensional feature vector without multi-granular scanning. Deep regression forest algorithms hope to solve the binary problem. The specific steps of the multi-granularity scanning are as follows: firstly, a 50-dimensional vector window is set to perform sliding value taking on an original feature vector, and the default step length is 1, so that 151 50-dimensional vectors can be obtained. Then, the obtained vectors are classified by two different types of forest models respectively, and 151 2-dimensional classification vectors are obtained respectively. Finally, all the classification vectors are spliced in sequence to form a feature vector with 604 dimensions, and the feature vector is used as the input of the cascade regression forest.
FIG. 2 is a diagram of a cascaded regression forest of the method of the present invention. The cascaded regression forest in fig. 2 uses the 604-dimensional feature vector obtained after processing by the multi-granularity scanning process in fig. 1 as input. Firstly, after the feature vectors are classified and processed by two different types of forest models, two 2-dimensional class vectors are obtained. The enhanced feature vector is then concatenated with the 604-dimensional original feature vector to form a 608-dimensional feature vector. Then, the 608-dimensional feature vector with enhanced features is taken as the input vector of the next stage. And the method is carried out until the last stage of the cascade regression forest is reached. And finally, averaging the category vectors generated by the last stage, and accumulating and processing the classification values and the obtained average value to obtain an accumulated and predicted power load value. In the cascade regression forest stage processing process, in order to reduce the overfitting risk, the category vector generated by each forest is generated through k-fold cross validation (k-fold cross validation). Each sample will be trained k-1 times as training data, resulting in k-1 class vectors. Then, the average value is taken as the enhanced feature vector of the next stage. The depth regression forest algorithm adopts 3-fold cross validation by default.
FIG. 3 is a depth regression forest ensemble prediction process diagram of the method of the present invention. A variety of different types of forest models are typically employed in deep regression forest modelsThe forest model is used for ensuring and improving the generalization degree of the prediction model. The complexity and number of the forest models can be configured differently. And the deep regression forest model processes the data by default by selecting a complete-random tree forest model (complete-random tree forest) and a random forest model. Each fully random tree forest model contains 500 fully random decision trees. The features employed for node splitting of each decision tree are randomly selected. When the node reaches full purity, the decision tree stops growing. The random forest model also comprises 500 decision trees and is randomly selectedOne feature for node splitting, d is the number of features of the input sample,is rounded up. And selecting the characteristics with the optimal Gini index for splitting by the decision tree. When the node reaches full purity, the decision tree stops growing. In the deep Regression forest algorithm, a Classification Regression decision Tree (CART) is generally used as a decision Tree of a random forest. CART is a typical binary decision tree proposed by Breiman et al, which can effectively process large data samples and solve the problem of nonlinear classification. Hence, CART is suitable for solving the classification problem of unclear classification mechanism. The core of the decision tree generation algorithm lies in how to select the attribute to be tested on each node and how to divide the data purity according to different data measurement methods. CART takes a Gini index (Gini) as an attribute measurement standard, and the smaller the Gini index is, the more accurate the partitioning effect is. The Gini index is defined as shown in formula (1).
Where p (i | t) is the probability that the test variable t belongs to a sample of class i; c is the number of samples. When Gini is 0, all samples belong to the same class.
If the attribute satisfies a certain valuePurity, the decision tree generation algorithm partitions the sample into the left sub-tree, otherwise partitions the sample into the right sub-tree. The CART decision tree generation algorithm selects the split attribute rule according to the principle that the Gini index is minimum. Suppose attribute A in training set C divides C into C1And C2The Gini index of the given partition C is shown in equation (2).
The growth depth of the decision tree is limited by conditions and can not grow without limitation. The conditions under which the decision tree stops growing are as follows:
(i) the data volume of the node is less than a specified value;
(ii) the Gini index is less than a threshold;
(iii) the depth of the decision tree reaches a specified value;
(iv) all features have been used.
And the default hyper-parameters for the depth regression forest in the present invention can be set as the following table.
The deep regression forest prediction model is trained by utilizing a training sample set, and the load of 96 moments on the day of 1 month and 10 days in 2015 year in a certain area is predicted. The prediction error obtained by the deep regression forest algorithm is 0.7445% (in other comparison algorithms, the prediction error obtained by the random regression forest algorithm is the smallest, 1.1085% and 32.84% larger than that of the deep regression forest algorithm). It can be seen that the deep regression forest of the present invention can actually perform accurate short-term power load prediction.
Claims (1)
1. A method for predicting the short-term load of a deep regression forest based on expandable information is characterized in that training can be carried out after data are serialized, large data are not relied on, and the calculation process comprises a multi-granularity scanning stage and a cascade regression forest stage; the method comprises the following steps:
(1) carrying out sequence processing on the collected historical data to form a column of data; the data comprises climate, weather, historical power load, date and large-scale activity data after serialization processing;
(2) in the calculation method, new types of data are added at any time, and serialization processing is carried out; the serialized data are emergency data and disaster data;
(3) training the deep regression forest by using the data obtained in the step (1) and the step (2); the deep regression forest method comprises a multi-granularity scanning stage and a cascading regression forest stage; setting the total length of the serialized data to be m '-m' + m multiplied by n, setting an n 'dimensional vector window to carry out sliding value taking on the original characteristic vector, and obtaining m' -n '+ 1 n' dimensional vectors if the step length is defaulted to be 1; classifying the obtained vectors respectively through two different types of forest models to respectively obtain m '-n' +1 2-dimensional classified vectors; then all the classified vectors are spliced into a 4(m '-n' +1) dimensional feature vector in sequence; in the cascade regression forest stage in the step (3), the cascade regression forest stage is used for processing the data characteristics layer by layer to obtain two 2-dimensional category vectors; the two 2-dimensional class vectors in the depth regression forest can effectively reflect the characteristics of the sample, and the enhanced characteristic vector is spliced with the original characteristic vector with the dimension of 4(m '-n' +1) to form a characteristic vector with the dimension of 4(m '-n' + 2); taking a 4(m '-n' +2) dimensional feature vector with enhanced features as an input vector of a next stage; the method is carried out until the last stage of the cascade regression forest is reached; carrying out regression processing on the category generated at the last stage and the category value thereof;
(4) predicting the short-term power load by utilizing real-time data; the real-time data are climate, weather, historical power load, date, large-scale activities, emergency data and disaster data after serialization processing; and the data is expanded at any time and trained on line.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810781666.3A CN110728391B (en) | 2018-07-17 | 2018-07-17 | Depth regression forest short-term load prediction method based on expandable information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810781666.3A CN110728391B (en) | 2018-07-17 | 2018-07-17 | Depth regression forest short-term load prediction method based on expandable information |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110728391A CN110728391A (en) | 2020-01-24 |
CN110728391B true CN110728391B (en) | 2022-07-05 |
Family
ID=69217028
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810781666.3A Active CN110728391B (en) | 2018-07-17 | 2018-07-17 | Depth regression forest short-term load prediction method based on expandable information |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110728391B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113759838A (en) * | 2020-11-04 | 2021-12-07 | 蕴硕物联技术(上海)有限公司 | Method and device for predicting shot blasting quality |
CN113380345A (en) * | 2021-06-08 | 2021-09-10 | 河南大学 | Organic chemical coupling reaction yield prediction and analysis method based on deep forest |
CN113610328A (en) * | 2021-08-25 | 2021-11-05 | 浙江浙能台州第二发电有限责任公司 | Power generation load prediction method |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107563539A (en) * | 2017-07-24 | 2018-01-09 | 佛山市顺德区中山大学研究院 | Short-term and long-medium term power load forecasting method based on machine learning model |
CN107944495A (en) * | 2017-12-04 | 2018-04-20 | 贵州电网有限责任公司电力科学研究院 | A kind of household electricity load classification recognition methods based on deep layer forest algorithm |
CN108011367A (en) * | 2017-12-04 | 2018-05-08 | 贵州电网有限责任公司电力科学研究院 | A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms |
-
2018
- 2018-07-17 CN CN201810781666.3A patent/CN110728391B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107563539A (en) * | 2017-07-24 | 2018-01-09 | 佛山市顺德区中山大学研究院 | Short-term and long-medium term power load forecasting method based on machine learning model |
CN107944495A (en) * | 2017-12-04 | 2018-04-20 | 贵州电网有限责任公司电力科学研究院 | A kind of household electricity load classification recognition methods based on deep layer forest algorithm |
CN108011367A (en) * | 2017-12-04 | 2018-05-08 | 贵州电网有限责任公司电力科学研究院 | A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms |
Non-Patent Citations (8)
Title |
---|
"‘智能电网+’研究综述";鞠平等;《电力自动化设备》;20180531;第38卷(第5期);全文 * |
"A Siamese Deep Forest";Lev V.Utkin等;《Knowledge-Based Systems》;20180101;第139卷;全文 * |
"Deep forest: towards an alternative to deep neural networks";Zhou ZhiHua等;《https://arxiv.org/abs/1702.08835v1》;20170228;全文 * |
"Short-term load forecasting based on human body amenity indicator and arithmetic of random forests";Yongjian Sun等;《International Conference on Renewable Power Generation (RPG 2015)》;20160407;全文 * |
"基于LSTM 模型的短期负荷预测";张亚宁;《科技与创新》;20180630;摘要,正文第1-5节 * |
"基于Spark 平台和并行随机森林回归算法的短期电力负荷预测";刘琪琛等;《电力建设》;20171031;第38卷(第10期);全文 * |
"深度森林与人工神经网络在光伏出力预测的比较";沈文博等;《信息技术与网络安全》;20180430;摘要,引言,正文第1-4节 * |
张亚宁."基于LSTM 模型的短期负荷预测".《科技与创新》.2018, * |
Also Published As
Publication number | Publication date |
---|---|
CN110728391A (en) | 2020-01-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Yin et al. | Deep forest regression for short-term load forecasting of power systems | |
Wang et al. | Research on Web text classification algorithm based on improved CNN and SVM | |
CN110728391B (en) | Depth regression forest short-term load prediction method based on expandable information | |
CN110580543A (en) | Power load prediction method and system based on deep belief network | |
CN111461463B (en) | Short-term load prediction method, system and equipment based on TCN-BP | |
CN110751318A (en) | IPSO-LSTM-based ultra-short-term power load prediction method | |
CN111898828A (en) | Hydroelectric power generation prediction method based on extreme learning machine | |
CN109615124B (en) | SCADA master station load prediction method based on deep learning | |
Cao et al. | Multi-step wind power forecasting model using LSTM networks, similar time series and LightGBM | |
CN110717610B (en) | Wind power prediction method based on data mining | |
Cardoso et al. | Improve irrigation timing decision for agriculture using real time data and machine learning | |
CN112861992B (en) | Wind power plant ultra-short term power prediction method based on independent sparse stacking self-encoder | |
Chen et al. | Wind speed forecasting using multi-scale feature adaptive extraction ensemble model with error regression correction | |
Hussain et al. | A dynamic neural network architecture with immunology inspired optimization for weather data forecasting | |
Wang et al. | Short-term load forecasting of power system based on time convolutional network | |
Wen et al. | MapReduce-based BP neural network classification of aquaculture water quality | |
Sang et al. | Ensembles of gradient boosting recurrent neural network for time series data prediction | |
CN117132132A (en) | Photovoltaic power generation power prediction method based on meteorological data | |
CN117374941A (en) | Photovoltaic power generation power prediction method based on neural network | |
CN115481788B (en) | Phase change energy storage system load prediction method and system | |
Zhao | The water potability prediction based on active support vector machine and artificial neural network | |
CN115759343A (en) | E-LSTM-based user electric quantity prediction method and device | |
Prashanthi et al. | A comparative study of the performance of machine learning based load forecasting methods | |
CN114492945A (en) | Short-term photovoltaic power prediction method, medium and equipment in electric power market background | |
Kang et al. | Research on forecasting method for effluent ammonia nitrogen concentration based on GRA-TCN |
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 |