CN113592181B - Small hydro-electric group output prediction method and system - Google Patents
Small hydro-electric group output prediction method and system Download PDFInfo
- Publication number
- CN113592181B CN113592181B CN202110879362.2A CN202110879362A CN113592181B CN 113592181 B CN113592181 B CN 113592181B CN 202110879362 A CN202110879362 A CN 202110879362A CN 113592181 B CN113592181 B CN 113592181B
- Authority
- CN
- China
- Prior art keywords
- data
- output
- hydropower station
- small hydropower
- sample 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 46
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 26
- 230000008569 process Effects 0.000 claims abstract description 13
- 230000009467 reduction Effects 0.000 claims abstract description 9
- 238000003066 decision tree Methods 0.000 claims description 17
- 238000007781 pre-processing Methods 0.000 claims description 13
- 230000001502 supplementing effect Effects 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 7
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 238000010248 power generation Methods 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 230000001174 ascending effect Effects 0.000 claims description 2
- 238000010276 construction Methods 0.000 claims description 2
- 238000013528 artificial neural network Methods 0.000 abstract description 3
- 239000013589 supplement Substances 0.000 abstract description 2
- 230000000007 visual effect Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000004931 aggregating effect Effects 0.000 description 1
- 230000008485 antagonism Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000002015 leaf growth Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 239000002699 waste material Substances 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
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- 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)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Primary Health Care (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a method and a system for predicting small hydro-electric group output. In the prediction method, the initial sample is preprocessed, so that the interference of bad data on the prediction accuracy is avoided to a certain extent, and the model is easier to converge to an optimal solution; generating small water power output data by using a deep convolution generation countermeasure network to supplement original sample data, and effectively solving the problem of prediction accuracy reduction caused by lack of sample data such as medium rain, heavy rain and the like; after the sample data are normalized and restored, reasonably reconstructing a sample data set according to the characteristics of the light gradient elevator, so that a prediction model is convenient to train, and the prediction accuracy of the light gradient elevator is improved; compared with the traditional neural network prediction method, the model has the advantages of supporting parallelization learning, low memory occupancy rate, visual training process, high training efficiency, high accuracy and the like, can process mass data, and can be applied to industrial practice faster and better.
Description
Technical Field
The invention relates to the field of hydro-electric group output prediction, in particular to a small hydro-electric group output prediction method and system considering the condition of insufficient data samples.
Background
Hydropower is a clean renewable energy source accepted by the market, water resources in southwest areas of China are rich, and hydropower development in the areas in recent years has a rapid development trend. The installed proportion of the radial-flow hydropower station in the area is larger, because the small hydropower station has weak water storage capacity, when the water is quite much supplied, all units are put into operation, the hydropower station generates electricity in a large quantity, the consumption capacity of local users is not strong, and the demand for electric power is far smaller than the generated energy; when the water flow is small, part of the units are not utilized due to water shortage. The small hydropower station is basically in a state of water, has weak self-regulation capability, is influenced by factors such as rainfall, has strong uncertainty, and the actual output is often deviated from the planned output seriously, so that not only is serious resource waste caused, but also the safe and stable operation of a power grid is influenced, and accurate prediction of the electric power generation capability of the small hydropower station is needed.
The small hydropower stations in China are weak in management and information acquisition capacity, and lack of related weather and hydrologic observation stations leads to insufficient historical hydrologic and meteorological data for prediction and missing data of historical power generation hours, so that the prediction difficulty is high. At present, in the field of small hydropower output prediction, methods with certain effects are proposed, such as weighted modeling according to the degree of similarity of regions by considering the regional cross-correlation of a plurality of small hydropower station groups; the power generation capacity of the small hydropower station group is calculated by referring to the prediction data of the local large hydropower station, but the problems of serious shortage of the historical data of the small hydropower station group cannot be well solved by the methods.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a small hydro-electric group output prediction method and a system considering the condition of insufficient data samples.
In order to achieve the above object, the present invention provides the following solutions:
a small hydro-electric group output prediction method comprises the following steps:
preprocessing original sample data; the raw sample data includes: rainfall data, air temperature, basin where small hydropower station is located, and historical output data;
generating small water electric power output data by adopting a deep convolution generation countermeasure network based on the original sample data, and supplementing the small water electric power output data into the original sample data;
carrying out normalization reduction treatment on the supplemented original sample data to obtain a sample data set, and reconstructing the sample data set according to the characteristics of the light gradient hoisting machine;
training the built light gradient elevator model by adopting the reconstructed sample data set; the light gradient elevator model takes the characteristic label of the small hydropower station as input and takes the output force of the small hydropower station as output;
and obtaining the output of the small hydropower station according to the characteristic label of the small hydropower station at a certain moment on the day to be predicted by adopting the trained light gradient elevator model.
Preferably, the preprocessing the raw sample data specifically includes:
performing real data removal and abnormal data rejection processing on the original sample data to obtain load data;
normalizing the load data;
grouping the normalized load data according to the date, performing independent heat coding on the basin where the small hydropower station is located on the same day, and performing quantized coding on the rainfall condition on the same day.
Preferably, the generating the small water power output data by adopting the deep convolution to generate the countermeasure network and supplementing the small water power output data into the original sample data specifically includes:
constructing a deep convolution to generate an countermeasure network; the deep convolution generating an countermeasure network includes: a generator and a arbiter; the generator consists of an input layer, a deconvolution layer and an output layer; the discriminator consists of an input layer, a convolution layer and an output layer;
the input data is remodeled into a three-dimensional tensor through the deep convolution to generate an input layer full-connection structure in an countermeasure network, then features are extracted through a convolution layer or a deconvolution layer, and finally small hydropower output data is obtained through a sigmoid activation function after the data is reconstructed again through an output layer; the input data includes: generator input data and arbiter input data; the input data of the generator is a data set consisting of noise data randomly and uniformly sampled in the [0,1] interval and sample tag data; when training the generator, the input data of the discriminator is a data set formed by the output data of the generator and sample label data; when training the discriminator, the input data of the discriminator is a data set composed of load data and sample label data.
Preferably, the training process of generating the countermeasure network by the deep convolution is as follows:
iteratively updating the depth convolution by adopting an Adam algorithm to generate super parameters of an countermeasure network;
alternately training the generator and the discriminator until a Nash equilibrium is reached to obtain a trained generator and a trained discriminator; the loss function expression is as follows:
wherein D represents a discriminator, G represents a generator, V represents a loss function, and P data (x) Representing the original sample set, P z (z) represents noise data, y represents label data, E represents cross entropy error, min represents the desire to minimize the loss function when training the generator, and max represents the desire to maximize the loss function when training the arbiter.
Preferably, reconstructing the sample data set according to the characteristics of the light gradient hoist specifically includes:
the single-point prediction method is adopted, and the light gradient elevator is adopted to output small hydropower output values at all times of day to be predicted based on training set data and label data; and combining a rainfall characteristic value of a small hydropower station with a preset river basin at the time to be predicted, a corresponding river basin class number and an output value of the small hydropower station group at the same time on a similar day of historical rainfall as training set data, wherein the output value of the small hydropower station at the current time is tag data.
Preferably, the building process of the light gradient elevator model comprises the following steps:
A. performing optimal feature division on the training set data by adopting a histogram algorithm, and dividing continuous samples into k bin containers;
B. filtering out the container with the smallest sub-set in the k bin containers, and calculating the index values of the rest bin containers to obtain a value;
C. after the rest bin containers are arranged in an ascending order according to the value, traversing is carried out to obtain an optimal splitting threshold value so as to form splitting nodes of a decision tree, an initial decision tree is built, and a first output result is obtained;
D. taking the residual error of the first output result and the original training set as new training set data, executing the steps A-C to form a next decision tree, and combining the next decision tree with the initial decision tree to obtain a second output result;
and D, repeating the steps A-D, and combining all the generated decision trees to obtain a final light gradient elevator model.
Preferably, a leaf-wise strategy with depth limitation is used in the construction process of the light gradient elevator model to form a single tree.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the small hydro-electric group output prediction method provided by the invention preprocesses the initial sample, avoids the interference of bad data on the prediction accuracy to a certain extent, and enables the model to be easier to converge to the optimal solution; generating small water power output data by using a deep convolution generation countermeasure network to supplement original sample data, and effectively solving the problem of prediction accuracy reduction caused by lack of sample data such as medium rain, heavy rain and the like; after the sample data are normalized and restored, reasonably reconstructing a sample data set according to the characteristics of the light gradient elevator, so that a prediction model is convenient to train, and the prediction accuracy of the light gradient elevator is improved; compared with the traditional neural network prediction method, the model has the advantages of supporting parallelization learning, low memory occupancy rate, visual training process, high training efficiency, high accuracy and the like, can process mass data, and can be applied to industrial practice faster and better.
Corresponding to the small hydropower station group output prediction method provided by the invention, the invention also provides the following implementation system:
a small hydro-electric group output prediction system, comprising:
the preprocessing module is used for preprocessing the original sample data; the raw sample data includes: rainfall data, air temperature, basin where small hydropower station is located, and historical output data;
the data supplementing module is used for generating small water power output data by adopting a deep convolution generation countermeasure network based on the original sample data and supplementing the small water power output data into the original sample data;
the data reconstruction module is used for carrying out normalization reduction processing on the supplemented original sample data to obtain a sample data set, and reconstructing the sample data set according to the characteristics of the light gradient hoisting machine;
the model training module is used for training the built light gradient elevator model by adopting the reconstructed sample data set; the light gradient elevator model takes the characteristic label of the small hydropower station as input and takes the output force of the small hydropower station as output;
and the output prediction module is used for obtaining the output of the small hydropower station according to the characteristic label of the small hydropower station at a certain moment on the day to be predicted by adopting the trained light gradient elevator model.
The technical effects achieved by the small hydropower station group output prediction system provided by the invention are the same as those achieved by the small hydropower station group output prediction method provided by the invention, so that the description is omitted here.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for predicting the output of a small hydro-electric group provided by the invention;
FIG. 2 is a quantized code diagram of the current day rainfall condition provided by an embodiment of the invention;
FIG. 3 is a schematic diagram of a deep convolutional generating challenge network (DCGAN) according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a light gradient elevator model (LightGBM) according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of the small hydro-electric group output prediction system provided by the invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention aims to provide a method and a system for predicting the output of a small hydro-electric group, which can improve the accuracy of the output prediction of the small hydro-electric group under the condition of insufficient data samples.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
As shown in fig. 1, the method for predicting the output of the small hydro-electric group provided by the invention comprises the following steps:
step 100: the raw sample data is preprocessed. The raw sample data includes: rainfall data, air temperature, basin in which small hydropower station is located, and historical output data.
Step 101: generating small water power output data by adopting a deep convolution generation countermeasure network based on the original sample data, and supplementing the small water power output data into the original sample data.
Step 102: and carrying out normalization reduction treatment on the supplemented original sample data to obtain a sample data set, and reconstructing the sample data set according to the characteristics of the light gradient elevator.
Step 103: and training the built light gradient elevator model by adopting the reconstructed sample data set. The light gradient elevator model takes the characteristic label of the small hydropower as input and takes the output of the small hydropower as output.
Step 104: and obtaining the output of the small hydropower station according to the characteristic label of the small hydropower station at a certain moment on the day to be predicted by adopting a trained light gradient elevator model. In the step, a rainfall characteristic value of a small hydropower station group in a certain river basin at the moment to be predicted, a corresponding river basin class number and an output value of the small hydropower station at the same moment when rainfall is similar to the day are input, and the output value is the output value of the small hydropower station group at the current moment.
In order to make the model easier to converge to the optimal solution, the preprocessing of the raw sample data in the step 100 specifically includes:
and carrying out processing of truly removing data and eliminating abnormal data on the original sample data to obtain load data. And collecting rainfall in the area where the small hydropower stations are located, river basin where each small hydropower station is located and historical output data of the small hydropower stations as an original sample set.
The load data is normalized, and the normalization processing is as follows:
where x represents the load data in the original sample, x max Represents the maximum value, x, of all raw load data min Represents the minimum value in all the original load data, and x is the normalized load data.
The normalized load data are grouped according to the date, the basin where the small hydropower station is located on the same day is subjected to independent heat coding, and the rainfall on the same day is subjected to quantitative coding, as shown in figure 2.
And finally, using the processed basin and rainfall characteristics as labels, and forming a group of data with the historical output data of the small hydropower group as input of the deep convolution generation countermeasure network.
In order to further improve the prediction accuracy, the sample expansion is performed on the portion lacking in the original sample data such as medium rain, heavy rain and the like to increase the total sample data, in the step 101, the deep convolution is adopted to generate the small water power output data against the network, and the small water power output data is supplemented to the original sample data, which specifically includes:
constructing a deep convolution generates a challenge network (DCGAN). The deep convolution generating an countermeasure network includes: a generator and a arbiter. The generator is composed of an input layer, a deconvolution layer and an output layer. The discriminator is composed of an input layer, a convolution layer and an output layer.
The input data is remodeled into a three-dimensional tensor through deep convolution to generate an input layer full-connection structure in an countermeasure network, features are extracted through a convolution layer or a deconvolution layer, and finally small hydropower output data is obtained through a sigmoid activation function after the data is reconstructed through an output layer. The input data includes: generator input data and arbiter input data. The input data of the generator is a data set consisting of noise data Z and sample tag data y which are randomly and uniformly sampled from the interval of [0,1 ]. When the generator is trained, the input data of the discriminator is a data set consisting of the output data g of the generator and the sample label data y, namely a data set consisting of weather and hydrologic characteristic data. When the discriminator is trained, the input data of the discriminator is a data set consisting of load data and sample label data, namely a data set consisting of small hydropower history output data x and meteorological and hydrological characteristic data y. The output data of the generator is the generated historical output data of the small hydropower station group, and the output of the discriminator is the score of the actual degree judgment of the input data.
The essence of the implementation of step 101 is the data enhancement of the data by using deep convolution to generate small hydro-power against the network.
The training process of the deep convolution generating countermeasure network is as follows:
iterative updating of the depth convolution by Adam algorithm generates super parameters of the countermeasure network.
The generator and the discriminant are trained alternately until a Nash equilibrium is reached to obtain a trained generator and a trained discriminant. The loss function expression is as follows:
wherein D represents a discriminator, G represents a generator, V represents a loss function, and P data (x) Representing the original sample set, P z (z) represents noise data, y represents tag data, E represents cross entropy error, and min represents trainingMinimizing the loss function is desirable when training the generator, and max represents maximizing the loss function when training the arbiter.
The generator and the discriminator mutually resist updating iteration, a better learning effect can be achieved, and the performances of the discriminator and the generator are continuously improved through updating training of parameters.
The trained generator may then be used to augment the original data set. The trained deep convolution generates the antagonism network as shown in fig. 3.
And carrying out normalized reduction on the sample after data enhancement, reasonably reconstructing a sample data set according to the characteristics of a light gradient elevator (LightGBM), and training a prediction model on the data set, so that the prediction accuracy of the light gradient elevator can be improved. Therefore, the specific implementation process of the step 102 is as follows:
and outputting small water power output values at all times of day to be predicted by adopting a light gradient elevator based on training set data and label data by adopting a single-point prediction method. And combining a rainfall characteristic value of a small hydropower station with a preset river basin at the time to be predicted, a corresponding river basin class number and an output value of the small hydropower station group at the same time on a similar day of historical rainfall as training set data, wherein the output value of the small hydropower station at the current time is tag data.
Because the model structure of the LightGBM is different from that of the traditional neural network, the model structure cannot output the small hydropower station output values at all times of the day to be predicted at one time, a single-point prediction method is adopted, namely, the rainfall characteristic value of a small hydropower station in a certain river basin at the time to be predicted, the corresponding river basin class number and the output value of the small hydropower station at the same time when the rainfall of the small hydropower station is similar to the day are combined to serve as training set data, and the output value of the small hydropower station at the current time is tag data.
The light gradient hoister is an improved gradient hoisting decision tree model proposed by microsoft corporation in 2017, and the main idea is to use a weak learner to perform iterative training to obtain an optimal model, namely a strong learner, as shown in the following formula:
wherein h is m (x) Is a regression tree, f M (x) A strong learner that combines M regression trees.
The overall process of constructing a model suitable for small hydropower output prediction light gradient elevator is approximately as follows: first, a histogram algorithm is adopted to efficiently and rapidly find the splitting points. Second, a leaf-growth with depth constraint (leaf-wise) strategy was used to form the regression tree. And finally, through the search integration of a large number of weak classifiers, the result is finally provided through the summation output of the weak classifiers.
And obtaining optimal dividing characteristics by adopting a histogram algorithm. Dividing continuous samples into K bin containers, filtering bin containers with the smallest subset, carrying out formula calculation on index values of the rest bin containers to obtain a value, sorting the bin containers from small to large according to the value, traversing and obtaining an optimal splitting threshold value.
On the basis of dividing the optimal feature points, a single decision tree is formed by using a leaf-wise strategy with depth limitation. The characteristics of the strategy include: when forming the splitting node of the decision tree, the leaf with the maximum gain is found to split and circulate. By limiting the depth and the number of leaves of the tree, the complexity of the model is reduced, and overfitting is prevented.
And (3) using a gradient lifting algorithm to iteratively generate a large number of decision trees, finally aggregating the decision trees into a prediction model, and outputting a prediction result.
The constructed structure of the model suitable for the small water power output prediction light gradient elevator is shown in fig. 4. The LightGBM can improve generalization ability and robustness of a single learner by combining prediction results of a plurality of base learners.
In addition, the invention also provides a small hydropower station group output prediction system corresponding to the small hydropower station group output prediction method. As shown in fig. 5, the prediction system includes: the device comprises a preprocessing module 1, a data supplementing module 2, a data reconstruction module 3, a model training module 4 and an output predicting module 5.
Wherein, the preprocessing module 1 is used for preprocessing the original sample data. The raw sample data includes: rainfall data, air temperature, basin in which small hydropower station is located, and historical output data.
The data supplementing module 2 is used for generating small water power output data by adopting a deep convolution generation countermeasure network based on the original sample data and supplementing the small water power output data into the original sample data.
The data reconstruction module 3 is used for carrying out normalization reduction processing on the supplemented original sample data to obtain a sample data set, and reconstructing the sample data set according to the characteristics of the light gradient elevator.
The model training module 4 is used for training the built light gradient elevator model by adopting the reconstructed sample data set. The light gradient elevator model takes the characteristic label of the small hydropower as input and takes the output of the small hydropower as output.
The output prediction module 5 is used for obtaining the output of the small hydropower station according to the characteristic label of the small hydropower station at a certain moment of the day to be predicted by adopting a trained light gradient elevator model.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.
Claims (6)
1. The method for predicting the output of the small hydropower group is characterized by comprising the following steps of:
preprocessing original sample data; the raw sample data includes: rainfall data, air temperature, basin where small hydropower station is located, and historical output data;
generating small water electric power output data by adopting a deep convolution generation countermeasure network based on the original sample data, and supplementing the small water electric power output data into the original sample data;
carrying out normalization reduction treatment on the supplemented original sample data to obtain a sample data set, and reconstructing the sample data set according to the characteristics of the light gradient hoisting machine;
training the built light gradient elevator model by adopting the reconstructed sample data set; the light gradient elevator model takes the characteristic label of the small hydropower station as input and takes the output force of the small hydropower station as output;
the trained light gradient elevator model is adopted, and the output of the small hydropower station is obtained according to the characteristic label of the small hydropower station at a certain moment on the day to be predicted;
the preprocessing of the original sample data specifically comprises:
performing real data removal and abnormal data rejection processing on the original sample data to obtain load data;
normalizing the load data;
grouping the normalized load data according to the date, performing independent heat coding on the basin where the small hydropower station is located on the same day, and performing quantized coding on the rainfall condition on the same day;
reconstructing the sample data set according to the characteristics of the light gradient elevator, wherein the method specifically comprises the following steps:
the single-point prediction method is adopted, and the light gradient elevator is adopted to output small hydropower output values at all times of day to be predicted based on training set data and label data; and combining a rainfall characteristic value of a small hydropower station with a preset river basin at the time to be predicted, a corresponding river basin class number and an output value of the small hydropower station at the same time on a similar day of historical rainfall as training set data, wherein the output value of the small hydropower station at the current time is tag data.
2. The method for predicting small hydro-electric group output according to claim 1, wherein generating small hydro-electric output data using a deep convolution generation countermeasure network and supplementing the small hydro-electric output data into the raw sample data specifically comprises:
constructing a deep convolution to generate an countermeasure network; the deep convolution generating an countermeasure network includes: a generator and a arbiter; the generator consists of an input layer, a deconvolution layer and an output layer; the discriminator consists of an input layer, a convolution layer and an output layer;
the input data is remodeled into a three-dimensional tensor through the deep convolution to generate an input layer full-connection structure in the countermeasure network, then features are extracted through a convolution layer or a deconvolution layer, and finally after the data is reconstructed again through an output layer, output data is obtained through a sigmoid activation function; the input data includes: generator input data and arbiter input data; the input data of the generator is a data set consisting of noise data randomly and uniformly sampled in the [0,1] interval and sample tag data; when training the generator, the input data of the discriminator is a data set formed by the output data of the generator and sample label data; when training the discriminator, the input data of the discriminator is a data set composed of load data and sample label data.
3. The method of claim 2, wherein the training process of the deep convolution generation countermeasure network is:
iteratively updating the depth convolution by adopting an Adam algorithm to generate super parameters of an countermeasure network;
alternately training the generator and the discriminator until a Nash equilibrium is reached to obtain a trained generator and a trained discriminator; the loss function expression is as follows:
wherein D represents a discriminator, G represents a generator, V represents a loss function, and P data (x) Representative ofOriginal sample set, P z (z) represents noise data, y represents label data, E represents cross entropy error, min represents the desire to minimize the loss function when training the generator, and max represents the desire to maximize the loss function when training the arbiter.
4. The method for predicting the output of a small hydropower station according to claim 1, wherein the process for constructing the light gradient elevator model comprises the following steps:
A. performing optimal feature division on the training set data by adopting a histogram algorithm, and dividing continuous samples into k bin containers;
B. filtering out the container with the smallest sub-set in the k bin containers, and calculating the index values of the rest bin containers to obtain a value;
C. after the rest bin containers are arranged in an ascending order according to the value, traversing is carried out to obtain an optimal splitting threshold value so as to form splitting nodes of a decision tree, an initial decision tree is built, and a first output result is obtained;
D. taking the residual error of the first output result and the original training set as new training set data, executing the steps A-C to form a next decision tree, and combining the next decision tree with the initial decision tree to obtain a second output result;
and D, repeating the steps A-D, and combining all the generated decision trees to obtain a final light gradient elevator model.
5. The method for predicting small hydro-power generation group output according to claim 4, wherein a single tree is formed by using a leaf-wise strategy with depth limitation in the construction process of the light gradient elevator model.
6. A small hydro-electric group output prediction system, comprising:
the preprocessing module is used for preprocessing the original sample data; the raw sample data includes: rainfall data, air temperature, basin where small hydropower station is located, and historical output data;
the data supplementing module is used for generating small water power output data by adopting a deep convolution generation countermeasure network based on the original sample data and supplementing the small water power output data into the original sample data;
the data reconstruction module is used for carrying out normalization reduction processing on the supplemented original sample data to obtain a sample data set, and reconstructing the sample data set according to the characteristics of the light gradient hoisting machine;
the model training module is used for training the built light gradient elevator model by adopting the reconstructed sample data set; the light gradient elevator model takes the characteristic label of the small hydropower station as input and takes the output force of the small hydropower station as output;
and the output prediction module is used for obtaining the output of the small hydropower station according to the characteristic label of the small hydropower station at a certain moment on the day to be predicted by adopting the trained light gradient elevator model.
The preprocessing of the original sample data specifically comprises the following steps:
performing real data removal and abnormal data rejection processing on the original sample data to obtain load data;
normalizing the load data;
grouping the normalized load data according to the date, performing independent heat coding on the basin where the small hydropower station is located on the same day, and performing quantized coding on the rainfall condition on the same day;
reconstructing the sample data set according to the characteristics of the light gradient elevator, wherein the method specifically comprises the following steps:
the single-point prediction method is adopted, and the light gradient elevator is adopted to output small hydropower output values at all times of day to be predicted based on training set data and label data; and combining a rainfall characteristic value of a small hydropower station with a preset river basin at the time to be predicted, a corresponding river basin class number and an output value of the small hydropower station at the same time on a similar day of historical rainfall as training set data, wherein the output value of the small hydropower station at the current time is tag data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110879362.2A CN113592181B (en) | 2021-08-02 | 2021-08-02 | Small hydro-electric group output prediction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110879362.2A CN113592181B (en) | 2021-08-02 | 2021-08-02 | Small hydro-electric group output prediction method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113592181A CN113592181A (en) | 2021-11-02 |
| CN113592181B true CN113592181B (en) | 2023-04-28 |
Family
ID=78253528
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110879362.2A Active CN113592181B (en) | 2021-08-02 | 2021-08-02 | Small hydro-electric group output prediction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113592181B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110414601A (en) * | 2019-07-30 | 2019-11-05 | 南京工业大学 | Photovoltaic module method for diagnosing faults, system and equipment based on depth convolution confrontation network |
| CN110516742A (en) * | 2019-08-28 | 2019-11-29 | 广东电网有限责任公司 | A kind of distribution terminal fault distinguishing method and system based on Combination neural network model |
| CN110909994A (en) * | 2019-11-13 | 2020-03-24 | 广西电网有限责任公司 | Small hydropower station power generation amount prediction method based on big data drive |
| CN112052996A (en) * | 2020-09-03 | 2020-12-08 | 哈尔滨工业大学 | Small hydropower station cluster power prediction method based on convolutional neural network technology |
-
2021
- 2021-08-02 CN CN202110879362.2A patent/CN113592181B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110414601A (en) * | 2019-07-30 | 2019-11-05 | 南京工业大学 | Photovoltaic module method for diagnosing faults, system and equipment based on depth convolution confrontation network |
| CN110516742A (en) * | 2019-08-28 | 2019-11-29 | 广东电网有限责任公司 | A kind of distribution terminal fault distinguishing method and system based on Combination neural network model |
| CN110909994A (en) * | 2019-11-13 | 2020-03-24 | 广西电网有限责任公司 | Small hydropower station power generation amount prediction method based on big data drive |
| CN112052996A (en) * | 2020-09-03 | 2020-12-08 | 哈尔滨工业大学 | Small hydropower station cluster power prediction method based on convolutional neural network technology |
Non-Patent Citations (3)
| Title |
|---|
| 徐磊 等.基于卷积神经网络与LightGBM的短期风电功率预测方法.水电能源科学.2021,第39卷(第2期),第209-212页. * |
| 邓真平 等.基于模拟退火-轻型梯度提升机的电厂出力短时多点预测方法.西北民族大学学报（自然科学版）.2021,第42卷(第121期),第4-10页. * |
| 陈伟龙 等.丽水电网小水电站群出力预测方法.水电自动化与大坝监测.2010,第34卷(第6期),第66-68页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113592181A (en) | 2021-11-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110580543A (en) | Power load prediction method and system based on deep belief network | |
| CN103400210A (en) | Short-term wind-speed combination forecasting method | |
| CN111832812A (en) | Wind power short-term prediction method based on deep learning | |
| CN110909911B (en) | Aggregation method of multidimensional time series data considering space-time correlation | |
| CN111815035A (en) | Short-term load prediction method fusing morphological clustering and TCN-Attention | |
| CN113141008B (en) | Data-driven power distribution network distributed new energy consumption capability assessment method | |
| CN102129511A (en) | System for forecasting short-term wind speed of wind power station based on MATLAB | |
| CN105373856A (en) | Wind electricity power short-term combined prediction method considering run detection method reconstruction | |
| CN113268931B (en) | Method and system for reconstructing photovoltaic array based on multi-target slime optimization algorithm | |
| CN109002926A (en) | The photovoltaic power generation quantity prediction model and its construction method of a kind of high accuracy and application | |
| CN110009141B (en) | Climbing event prediction method and system based on SDAE feature extraction and SVM classification model | |
| CN111401603A (en) | Multi-type energy demand prediction method | |
| CN113708366A (en) | Power distribution network optimal scheduling method and system considering new energy and electric automobile | |
| Chen et al. | Research on wind power prediction method based on convolutional neural network and genetic algorithm | |
| CN115065078B (en) | Energy storage capacity configuration method and system in micro-grid environment | |
| CN114154558A (en) | Distributed energy power generation load prediction system and method based on graph neural network | |
| CN110909994A (en) | Small hydropower station power generation amount prediction method based on big data drive | |
| CN114386718A (en) | Wind power plant output power short-time prediction algorithm combined with particle swarm neural network | |
| Mukilan et al. | Prediction of rooftop photovoltaic solar potential using machine learning | |
| CN113592181B (en) | Small hydro-electric group output prediction method and system | |
| CN103577893B (en) | A kind of new energy and the two-way energy conservation optimizing method powered for high energy load of thermoelectricity | |
| CN113363976A (en) | Scene graph-based mid-term optimized scheduling method for wind, light and water complementary power generation system | |
| Xiaosheng et al. | A deep learning approach for wind power prediction based on stacked denoising auto encoders optimized by bat algorithm | |
| Lv et al. | Research on short-term load forecasting approach for smart grid | |
| Lv et al. | Photovoltaic Power Forecasting in Dunhuang based on Recurrent Neural Network |
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 |