CN113361761A - Short-term wind power integration prediction method and system based on error correction - Google Patents
Short-term wind power integration prediction method and system based on error correction Download PDFInfo
- Publication number
- CN113361761A CN113361761A CN202110608220.2A CN202110608220A CN113361761A CN 113361761 A CN113361761 A CN 113361761A CN 202110608220 A CN202110608220 A CN 202110608220A CN 113361761 A CN113361761 A CN 113361761A
- Authority
- CN
- China
- Prior art keywords
- wind power
- prediction
- model
- data
- error
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000012937 correction Methods 0.000 title claims abstract description 26
- 230000010354 integration Effects 0.000 title claims description 16
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000010248 power generation Methods 0.000 claims abstract description 13
- 238000004422 calculation algorithm Methods 0.000 claims description 61
- 230000002068 genetic effect Effects 0.000 claims description 31
- 238000007637 random forest analysis Methods 0.000 claims description 27
- 210000000349 chromosome Anatomy 0.000 claims description 17
- 238000010276 construction Methods 0.000 claims description 7
- 230000010076 replication Effects 0.000 claims description 5
- 238000003066 decision tree Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 238000003860 storage Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 description 23
- 230000008569 process Effects 0.000 description 11
- 108090000623 proteins and genes Proteins 0.000 description 9
- 230000000875 corresponding effect Effects 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 7
- 238000005457 optimization Methods 0.000 description 7
- 230000035772 mutation Effects 0.000 description 6
- 238000005070 sampling Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000007547 defect Effects 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000002596 correlated effect Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000010187 selection method Methods 0.000 description 2
- 230000004083 survival effect Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 208000001613 Gambling Diseases 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000000126 substance 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
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- 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)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Evolutionary Biology (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Development Economics (AREA)
- Computing Systems (AREA)
- Game Theory and Decision Science (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Entrepreneurship & Innovation (AREA)
- Genetics & Genomics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Physiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a short-term wind power integrated prediction method and a short-term wind power integrated prediction system based on error correction, wherein the method comprises the following steps: acquiring operation data and weather forecast data of a wind power plant; inputting the data into a trained wind power prediction model and a wind power error prediction model respectively for prediction; adding the prediction results output by the two models to obtain a final short-term wind power prediction result; the training data set of the wind power prediction model is historical operation data and weather data in a preset time period of the wind power plant; and taking the power error data set, historical operating data and weather data in a preset time period of the wind power plant as a training data set of the wind power error prediction model. The method improves the short-term wind power prediction precision and improves the consumption capacity of the power grid to new energy power generation.
Description
Technical Field
The invention relates to the technical field of wind power prediction in a new energy power generation process, in particular to a short-term wind power integrated prediction method and system based on error correction.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
With the increasingly prominent environmental problems and the increasing demand for energy, the development of new energy represented typically by wind power has become a common consensus in countries around the world. Wind energy has the advantages of no pollution, reproducibility, wide resources and the like, and is vigorously developed and applied by many countries, but wind power output power has strong randomness and volatility, large-scale wind power is accessed into a power grid to cause the difficulty of distribution and utilization balance to be increased, the uncertainty of operation of a power system is obviously increased, and the contradiction between safe operation of the system and efficient consumption of new energy is increasingly prominent. Therefore, it becomes important to improve the wind power prediction accuracy.
Based on the division of time scales, wind power prediction is mainly divided into ultra-short-term prediction, short-term prediction and medium-long-term prediction. The short-term wind power prediction result can adjust the unit combination scheme, optimize the conventional power supply unit power generation plan and improve the consumption capacity of the power grid on new energy power generation, so the short-term wind power prediction is a hotspot of current research.
At present, most researches on short-term wind power prediction adopt a single model for prediction, but the single prediction model is used for prediction based on a certain assumed space, so that the wind power prediction by using the single model inevitably has prediction errors; in addition, most of the current research methods adopt a certain prediction model for prediction and directly use the model prediction result as the final prediction result, but the research methods do not consider that the results obtained by prediction based on any model have errors.
Disclosure of Invention
In order to solve the problems, the invention provides a short-term wind power integrated prediction method and system based on error correction, which comprehensively utilize an XGboost model and a random forest model in an integrated learning algorithm to predict the wind power, so that the defect that a single prediction model has larger prediction error at some points is avoided, and the short-term wind power prediction precision is improved by a residual error learning method.
In some embodiments, the following technical scheme is adopted:
a short-term wind power integration prediction method based on error correction comprises the following steps:
acquiring operation data and weather forecast data of a wind power plant;
inputting the data into a trained wind power prediction model and a wind power error prediction model respectively for prediction;
adding the prediction results output by the two models to obtain a final short-term wind power prediction result;
the training data set of the wind power prediction model is historical operation data and weather data in a preset time period of the wind power plant; inputting historical operating data and weather data in a preset time period of a wind power plant into a wind power prediction model, outputting a wind power prediction value, and obtaining a power error data set based on the wind power prediction value; and the power error data set, historical operating data in a preset time period of the wind power plant and weather data are used as a training data set of the wind power error prediction model.
As a further approach, the wind farm operational data and weather forecast data include, but are not limited to, generated power, wind direction, wind speed, temperature, humidity, and barometric pressure data for the wind farm.
As a further scheme, the wind power prediction model is constructed by adopting an XGboost model optimized by an improved genetic algorithm.
As a further scheme, during construction of the wind power prediction model, binary coding is adopted, the minimum leaf node weight, the maximum depth, the learning rate and the gamma value of the XGboost model are selected as independent variable parameters, a population is initialized in a random mode, each obtained chromosome contains the independent variable parameters of the XGboost model, the minimum normalized mean absolute error NMAE of the model prediction result is used as the model fitness, and then the optimal solution of the XGboost model independent variable parameters is found through replication, timely-changing improved intersection and variation.
As a further scheme, the wind power error prediction model is constructed by utilizing a random forest model optimized by an improved genetic algorithm.
As a further scheme, when the wind power error prediction model is constructed, binary coding is adopted, the number of trees of a random forest model, the randomly selected characteristic number of each decision tree and the deepest depth of the tree are used as parameters, a population is initialized in a random mode, each obtained chromosome contains three parameters of the random forest model, the normalized mean absolute error NMAE minimum of a model prediction result is used as the model fitness, and then the optimal solution of the three parameters of the random forest model is found through copying, timely-changing improved intersection and variation.
As a further scheme, operation and numerical weather forecast data of the wind power plant are collected in real time, the trained XGboost and random forest models are used for forecasting respectively, a wind power preliminary forecasting result and a wind power error forecasting result are correspondingly obtained, and the two model forecasting results are added to obtain a final wind power forecasting result.
In other embodiments, the following technical solutions are adopted:
a short-term wind power integrated prediction system based on error correction comprises:
the data acquisition module is used for acquiring operation data and weather forecast data of the wind power plant;
the model prediction module is used for inputting the data into a trained wind power prediction model and a wind power error prediction model respectively for prediction;
the data output module is used for adding the prediction results output by the two models to obtain a final short-term wind power prediction result;
the training data set of the wind power prediction model is historical operation data and weather data in a preset time period of the wind power plant; inputting historical operating data and weather data in a preset time period of a wind power plant into a wind power prediction model, outputting a wind power prediction value, and obtaining a power error data set based on the wind power prediction value; and the power error data set, historical operating data in a preset time period of the wind power plant and weather data are used as a training data set of the wind power error prediction model.
In other embodiments, the following technical solutions are adopted:
a terminal device comprising a processor and a memory, the processor being arranged to implement instructions; the memory is used for storing a plurality of instructions which are suitable for being loaded by the processor and executing the short-term wind power integration prediction method based on error correction.
In other embodiments, the following technical solutions are adopted:
a computer readable storage medium, wherein a plurality of instructions are stored, the instructions are suitable for being loaded by a processor of a terminal device and executing the short-term wind power integration prediction method based on error correction.
Compared with the prior art, the invention has the beneficial effects that:
1. the XGboost model and the random forest model in the integrated learning algorithm are comprehensively utilized to establish the wind power
The prediction model avoids the defect that a single prediction model has large prediction errors at certain measuring points.
2. The invention optimizes the parameters of the XGboost model and the random forest model by using the improved genetic algorithm, and overcomes the defect of using the improved genetic algorithm
The traditional genetic algorithm carries out parameter optimization and falls into the defect of local optimization.
3. According to the method, the wind power preliminary prediction result is corrected by using the information of the wind power prediction error, the short-term wind power prediction precision is improved by using a residual error learning method, the power generation plan of a conventional power supply unit is optimized, and the consumption capacity of a power grid on new energy power generation is improved.
Additional features and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
FIG. 1 is a wind power prediction flow chart based on error correction in an embodiment of the present invention;
FIG. 2 is a correlation coefficient diagram between different meteorological elements and wind power in the embodiment of the invention;
FIG. 3 is a flow chart of an improved genetic algorithm in an embodiment of the present invention;
FIG. 4 is a schematic diagram of a Bagging algorithm in an embodiment of the invention;
FIG. 5 is a graph of NMAE values for each model in an embodiment of the present invention;
FIG. 6 is a graph of NRMSE values for various models in an embodiment of the present invention;
FIG. 7 shows the predicted result of wind power No. 4/month 1 in 2019 of each model in the embodiment of the present invention;
FIG. 8 shows the predicted results of wind power generation No. 7/month 1 in 2019 for each model in the embodiment of the present invention;
FIG. 9 shows the wind power prediction results of model No. 10/month 1 in 2019 in the embodiment of the present invention;
FIG. 10 shows the wind power prediction results of model No. 1/2019 in the embodiment of the present invention;
FIG. 11 is a No.6 wind farm relative prediction error variation curve in the embodiment of the invention.
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 application 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 application. 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.
The embodiments and features of the embodiments of the present invention may be combined with each other without conflict.
Example one
According to the embodiment of the invention, a short-term wind power integration prediction method based on error correction is disclosed, and with reference to fig. 1, the method comprises the following steps:
(1) acquiring real-time operation data and weather forecast data of the wind power plant;
specifically, wind farm operation and numerical weather forecast data refer to any data related to wind power generation, including but not limited to: and generating power, wind direction, wind speed, temperature, humidity, air pressure and other data of the wind power plant.
(2) Inputting the data into a trained wind power prediction model and a wind power error prediction model respectively for prediction;
specifically, operation and numerical weather forecast data in a preset time period of the wind power plant are collected, and data of meteorological elements such as 10-meter wind speed and 100-meter wind speed are extracted.
The historical data set is divided into a training set A and a training set B, and a wind power prediction model is initially established by utilizing the XGboost optimized by the improved genetic algorithm based on the training set A.
Inputting meteorological data of the training set B into a trained XGboost model for prediction to obtain a wind power prediction value, further obtaining a power error data set D, and establishing a wind power error prediction model by utilizing a random forest optimized by an improved genetic algorithm based on the training set B and the prediction error set D.
The embodiment makes full use of the advantages that integrated learning can be combined with a plurality of weak supervision models to obtain a comprehensive strong supervision model, and provides a short-term wind power integrated prediction model based on error correction based on a residual error learning method. Firstly, preliminarily establishing a wind power prediction model by using an XGboost model optimized by an improved genetic algorithm, then establishing a wind power error prediction model by using a random forest optimized by the improved genetic algorithm based on a prediction error of the XGboost model, and finally adding prediction results of the XGboost model and the random forest to obtain a final prediction result. The method improves the short-term wind power prediction precision by utilizing the advantages of the integrated learning model and through a residual learning method, further optimizes the power generation plan of the conventional power supply unit and improves the consumption capacity of the power grid to new energy power generation.
The detailed method comprises the following steps:
(1) data preparation
To determine the input amount of the prediction model, the present embodiment first uses the correlation coefficient ρxyTo analyze meteorological factors such as wind speed and wind direction
Correlation between element and wind power, ρxyThe larger the value of (A), the larger is xtAnd ytThe stronger the correlation between two time series, ρxyIs defined as:
wherein the content of the first and second substances,
and
are the average of the two time series, respectively, and N is the number of samples tested.
Table 1 shows the relationship between the correlation coefficient and the degree of correlation between two time series, and thus the correlation coefficient can be based on
To determine the degree of correlation between the two time series.
TABLE 1 correlation coefficient and degree of correlation
Fig. 2 shows a correlation coefficient diagram between the wind power (P) and seven meteorological quantities including the wind speed of 10 meters (WS10), the wind direction of 10 meters (WD10), the wind speed of 100 meters (WS100), the wind direction of 100 meters (WD100), the ground temperature (Tem), the ground pressure (AP) and the ground humidity (Hum), from which it can be seen that the wind speed of 10 meters and the wind speed and power of 100 meters are significantly correlated, the wind speed of 10 meters and the wind direction of 100 meters and power are low correlated, and the temperature, air pressure and humidity and wind power are weakly correlated, so that the two meteorological quantities of the wind speed of 10 meters and the wind speed of 100 meters at the target time are selected as model inputs in the present embodiment.
(2) Preliminary wind power prediction model based on XGboost
The integrated learning algorithm is mainly divided into a Bagging algorithm and a Boosting algorithm, the XGboost algorithm belongs to one of the Boosting algorithms, a new classification regression tree is generated in each iteration of the algorithm, the new tree can continuously fit the residual error of the previous tree, and the previous experimental result is continuously repaired, so that a learning model with higher accuracy is constructed. When a sample needs to be predicted, corresponding scores are generated on the leaf nodes corresponding to each tree, and thus the sum of all the scores is the predicted value of the sample. The XGboost model can realize parallel processing, and compared with the GBDT algorithm of the Boosting algorithm, the XGboost model has great improvement on the processing speed.
1) Brief introduction to the model
The XGboost algorithm utilizes a Hessian matrix to expand a loss function Taylor to the second order, the original optimization problem is converted into a convex function to solve the optimal solution, and the problem of distributed calculation which is difficult to implement in the GBDT algorithm is solved. Meanwhile, the XGboost algorithm carries out regularization constraint on the complexity of the tree, and the probability of overfitting of the model is reduced. Assuming that the model has K trees, the XGboost model can be expressed as:
wherein, yi' As model predictor, xiFor the ith sample, K is the number of trees and F is the space of the regression tree.
The XGboost model generates a new tree in each iteration for fitting the residual error of the previous tree, and the predicted value in the t iteration is assumed to be yi(t)', then
The optimization objective function can now be expressed as:
the objective function is thus partially composed of a loss value, a regularization term, and a constant term 3, expanding the loss function to the second order, and removing the constant term, the objective function becomes:
the canonical terms that define the model are:
where γ and λ are model parameters, T is the total number of leaf nodes in the tree, wjThe weight of the jth leaf node in the tree.
Definition of
IjIs the sample set on the jth leaf node, the objective function is further simplified as:
therefore, the objective function of the model is converted into a quadratic equation of a single element related to the leaf node weight, and the optimal and immediate change of the tree structure is solved
To solve the function optimal solution problem, the final objective function becomes:
fobjthe smaller the value of (A), the better the structure of the whole tree.
2) Model parameter setting
The XGboost model parameters mainly comprise four parameters, namely a minimum leaf node weight min _ child _ weight, a maximum depth max _ depth, a learning rate eta and a gamma value of a combination of a penalty term and a leaf node. The selection of the model parameters greatly affects the model prediction effect, and the XGboost model parameters are optimized by using an improved genetic algorithm in the embodiment in consideration of the fact that the value of each parameter is random.
The genetic algorithm simulates chromosomes by using a solution of a problem, whether the chromosomes are suitable for the environment is judged by using the numerical value of a fitness function, the larger the numerical value of the fitness function is, the more the chromosomes can be suitable for the environment, the survival principle of fitters in the biological evolution theory is used, the chromosomes with larger fitness function values are copied from a parent to be crossed and mutated to form new chromosomes, and the diversity of the solution is maintained.
Coding mode
The coding mode is determined as the basic step of the genetic algorithm, and the binary coding mode and the decimal coding mode are two commonly used coding modes. Each chromosome in the binary coding mode is represented by 0 or 1 to represent each gene position, the coding mode is simple and convenient in decoding and decoding, and steps in crossing and mutation are easy, so that the binary coding is simple and easy to understand. However, when the problem of complex processing and high precision requirement is processed by using binary coding, certain error exists because of discretization of the problem, and when the problem of high dimensionality is processed, the calculation speed is influenced because the binary coding is longer. The decimal code, i.e. decimal number, represents each gene position on each chromosome, the length of the chromosome is determined by the control variable of the model, obviously, for the same model, the length of the chromosome is shorter by adopting the decimal code than by adopting the binary code, and the coding by adopting the decimal code does not need decoding and decoding, so the calculated amount is smaller by adopting the decimal code in general, but each gene position in the chromosome can not be out of range in the crossing and variation operation process.
Generation of initial population
A population consists of many chromosomes and a solution corresponds to one chromosome of the population. The size of the population represents the space size of the model solution, for a large population, although the calculation speed is relatively slow, the optimal solution of the model is easy to find, and conversely, if the population is small, although the calculation speed is relatively fast, the solution of the optimization model is relatively single and is easy to fall into local convergence, so that the initial population size is generally 30-100. The initial population can be obtained by a random method and can also be obtained according to the actual distribution of the population.
Calculating fitness function value
The fitness function is the basis of the genetic algorithm, and is related to the realization of the optimal solution and the calculation speed of the genetic algorithm. In genetic algorithms, the value of the fitness function reflects whether an individual can adapt to the environment and decides whether a parent can be inherited to children. If the fitness function value of an individual is larger, the individual has larger probability to be transmitted to the next generation, otherwise, the individual has high probability to be eliminated. The method for constructing the fitness function mainly comprises a direct construction and an indirect construction, wherein the direct construction is to directly use the objective function as the fitness function without conversion, and the indirect construction is to construct the fitness function by changing the objective function to a certain degree. In this embodiment, the minimum value of the normalized Mean Absolute error NMAE (normalized Mean Absolute error) of the prediction result of the model is used as the fitness, where NMAE is defined as:
where n is the number of predicted sample points, PiTo predict the power value, Pi *As power observations, PNIs the rated installed capacity.
Reproduction
The replication is a process of simulating natural selection, so the replication process is a selection process, that is, an individual with strong adaptability is selected from a population, and an individual with a large fitness function value is selected by generally adopting a proportional method, a sorting method and a proportional sorting method. The proportion method is widely applied and belongs to one of random selection methods, and is also called as gambling plate selection. If n individuals exist, the circular game board is divided into n sectors, the area of each sector respectively represents the probability of the corresponding individual being selected, and the individual corresponding to the sector with the large area is easy to select, which accords with the principle of survival of the opponent. However, in the case that the population has only few individuals with high fitness, the population may propagate too fast by using the proportional selection method, so that the final result is a locally optimal solution.
Fifth, cross
The cross simulation is a gene recombination process in a biological evolution process, excellent characteristics in individuals are combined to generate filial generations different from parents, the crossing selects points at the same position of the two parents to exchange genes according to a certain probability, and then new filial generations are obtained, and in a traditional genetic algorithm, single-cut-point crossing with small calculation amount is usually utilized, namely, a cut point is selected to exchange genes behind the corresponding cut points of the two parents P1 and P2, so that the filial generations C1 and C2 are obtained. For example, P1 is 10101011, P2 is 10001001 in the parent, and when the position of the cut point is 6, the child C1 is 10101001 and C2 is 10001011 after the intersection. The generation of the new individuals is mainly completed through crossing, the speed of crossing the chromosomes depends on the size of the crossing probability, the search is slowed down or even stopped if the crossing probability is too small, so the value of the crossing probability is generally large, but the new individuals are high in speed and easy to damage if the value is too large, and therefore the crossing probability is generally 0.6-0.95.
Variation of
The variant similar organisms are mutated during evolution, if the algorithm adopts binary coding, the corresponding gene positions are inverted, namely 1 is changed into 0, and 0 is changed into 1, and if the coding adopts decimal coding, the numerical values of the corresponding gene positions can be changed to a certain extent during variation, new individuals can be generated due to variation, the gene types of the population can be increased, and thus the genetic algorithm can not obtain the local optimal solution during solving. The genetic algorithm is changed into a simple random search process due to the large numerical value of the variation probability, but if the variation probability is too small, new individuals are difficult to appear in the population, so the value range of the variation probability is generally 0.0001-0.1.
Seventhly, outputting the optimal solution
In general, convergence criteria of a mathematical optimization algorithm are strict, and the judgment basis of a genetic algorithm is heuristic, and there are two common judgment methods: one is judged according to the quality of the solution, namely the iteration process is considered to be finished when the arrangement sequence of the group solution sets does not change any more; another approach is to allow the maximum number of iterations possible as a convergence criterion.
However, the mutation method in the traditional genetic algorithm has the problem of early maturity because of the cross probability PcAnd the probability of variation PmThe convergence characteristic of the genetic algorithm is determined, so that the method for timely changing the crossover and mutation probabilities is adopted in the embodiment for ensuring the global optimal solution, even if the crossover probability and the mutation probability in the algorithm are timely changed along with the value of the individual fitness, the crossover and mutation probabilities are reduced when the solution space of the algorithm cannot be converged, and the crossover probability P is increased when the algorithm is in local optimalcAnd the probability of variation PmTo jump out of the algorithm. If the fitness of an individual is lower than the average fitness, the individual is made to have a cross probability PcAnd the probability of variation PmLarger values eliminate them, whereas individuals have a lower cross probability P if their fitness is higher than the average fitnesscAnd the probability of variation PmSo that it can enter belowOne generation, the specific calculation is shown in equations (12) and (13):
where f' is the maximum of the two volume fitness values for the crossover operation, and favgIs an average fitness value, fmaxIs the maximum fitness value, fminThe minimum fitness value is f, and the variant individual fitness value is f.
The flow chart of the improved genetic algorithm provided by the embodiment is shown in fig. 3, and the adaptive cross probability and the mutation probability enable the population to have good diversity on one hand, and enable the genetic algorithm to have good convergence on the other hand, so that the convergence speed of the genetic algorithm can be obviously improved, and the algorithm is effectively prevented from falling into local optimum. The present embodiment employs a double-point crossover operation, and sets the parameters as: pc1＝0.9，Pc2＝0.6，Pm1＝0.1，Pm2＝0.001。
(3) Wind power error prediction model based on random forest
The random forest algorithm belongs to one of Bagging algorithms, a self-help sampling technology is adopted in the algorithm, a regression tree is utilized to form a combined model, and a schematic diagram of the Bagging algorithm is shown in FIG. 4.
Bagging is characterized by random sampling, wherein samples with the same number as that of a training set are randomly collected and collected for T times to obtain a sampling set. For a sample, in a random sampling of m samples, the probability of being acquired each time is 1/m, and the probability of not being acquired in m samples is:
taking the limit on m can obtain
Therefore, for each round of random sampling of Bagging, about 36.8% of the data is not collected and is called "out-of-bag data", and the data does not participate in fitting of the training set model, but can be used as a test set to test the generalization capability of the model, and the test result is called "outsource estimation". Bagging does not limit the weak learners, decision trees and neural networks are usually selected, and for the regression problem, Bagging usually adopts a simple average method to perform arithmetic average on regression results obtained by the T weak learners to obtain final model output.
1) Brief introduction to the model
The random forest algorithm is improved on the basis of Bagging algorithm, and the algorithm generates a series of mutually independent regression trees
The model is combined by averaging the output of the model, and the mathematical expression is as follows:
wherein N is the number of regression tree models.
In the process of establishing each regression tree, random sampling is carried out on a sample space and a feature space, random attributes are introduced, further, the correlation among regression tree models is reduced, and the generalization of the models is improved by combining a large number of regression trees, so that the algorithm has the characteristics of high efficiency, accuracy and the like. In the process of establishing the model, the number N of the regression tree models is a main model parameter, and the gradual increase of N can cause the generalization error of the random forest to be converged to a limit without an overfitting phenomenon, so that the number of the generated regression tree models can be increased for data containing a large amount of noise, the random forest algorithm can be ensured to show good performance, and the value range of N is generally hundreds to thousands.
2) Model parameter optimization
The random forest model parameters mainly comprise three parameters of tree number trees _ num of the forest, feature number max _ features randomly selected by each decision tree and deepest depth max _ depth of the tree, and similarly, the XGboost model parameters are determined, and the model parameters of the random forest are optimized by using an improved genetic algorithm.
(3) Adding the prediction results output by the two models to obtain a final short-term wind power prediction result;
specifically, operation and numerical weather forecast data of a wind power plant are collected in real time, the XGboot model and the random forest model are input to obtain a wind power preliminary prediction result and a wind power error prediction result respectively, and finally the wind power preliminary prediction result and the power error prediction value are added to obtain a final wind power prediction value.
The embodiment also provides corresponding prediction and evaluation indexes, which are specifically as follows:
in this embodiment, the normalized Mean absolute error NMAE and the normalized Root Mean Square error NRMSE (normalized Root Mean Square error) are selected as the evaluation indexes of the prediction result, and the NRMSE is defined as follows:
where n is the number of predicted sample points, PiFor the wind power prediction, Pi *For the actual wind power value, PNIs the rated installed capacity.
The relative prediction error RE of the wind power is defined as follows:
wherein, PiFor the wind power prediction, Pi *And the actual wind power value is obtained.
The predicted results are as follows:
the analysis of the embodiment utilizes 10 wind power plants in Jilin province to simulate the wind power prediction with a prospective 72 hours. The data set comprises NWP data and historical wind power data of each wind power plant in 2017 and 2019, the time resolution is 15 minutes, and the data of the data set in 2017, 2018 and 2019 are divided into a training set A, a training set B and a testing set respectively.
In order to verify the effectiveness of the proposed model, three models, namely M1, M2 and M3, are selected as reference models for comparison in the present embodiment, and the model proposed in the present embodiment is denoted as M4.
M1: and predicting by using an XGboost model optimized by an improved genetic algorithm.
M2: and (4) predicting by using a random forest model optimized by an improved genetic algorithm.
M3: and predicting by using the BP neural network optimized by the improved genetic algorithm.
Table 2 shows the NMAE value of each wind farm when each model is used for prediction and the average NMAE value of 10 wind farms, and it can be seen from the data in the table that when each wind farm is predicted by using the model provided in this embodiment, the prediction error is smaller than that of the other three models, and when 10 wind farms are predicted by using the model provided in this embodiment, the average NMAE value is 9.26%, whereas when 3 models, M1, M2 and M3, are used for prediction, the average NMAE values are 10.35%, 11.92% and 12.13%, respectively. Particularly, the prediction accuracy of the model provided by the embodiment is improved by 2.87% compared with that of the commonly-used BP neural network.
Table 2 NMAE value (%) -of each model prediction result
Fig. 5 and 6 show the NMAE value and the NRMSE value of 10 wind farms respectively predicted by using the models in four different seasons, namely spring, summer, autumn and winter, respectively, and it can be seen from the data in the graphs that the values of the NMAE and the NRMSE of the prediction results obtained by using the model provided in the present embodiment for prediction are the minimum in any season, such as spring, summer, autumn and winter, and the prediction error fluctuation is small when using the model provided in the present embodiment for prediction in four seasons, namely spring, summer, autumn and winter.
In order to intuitively explain the prediction effect, taking a No.6 wind power plant as an example, fig. 7-10 show the wind power prediction results of the wind power plant in four seasons, namely spring, summer, autumn and winter, respectively, and it can be seen from the graph that the wind power prediction result obtained by using the wind power integrated prediction model based on error correction provided by the embodiment is closer to the real power, and the result verifies that the model provided by the embodiment has good deterministic prediction performance. Fig. 11 shows a variation curve of the relative prediction error of the No.6 wind farm with time, and it can be seen from the diagram that the relative prediction error value of the wind power is close to 0, although the absolute value of the relative prediction error of the wind power increases with the increase of the prediction time, the absolute value of the prediction error is always kept within 15%, and the prediction result verifies that the proposed model can obtain a good wind power prediction result.
In general, the model provided by the embodiment can improve the short-term wind power prediction precision. The method fully utilizes the advantages that integrated learning can be combined with a plurality of weak supervision models to obtain a comprehensive strong supervision model, the XGboost is utilized to establish a wind power preliminary prediction model, an error prediction model based on a random forest is established according to a residual error learning method, the prediction results of the XGboost and the random forest are added to serve as a final prediction result, the effectiveness of the method is proved through example result analysis, and the comparison result shows that the model provided by the example can improve the short-term wind power prediction precision, further optimize the power generation plan of a conventional power unit and improve the consumption and acceptance of a power grid for new energy power generation.
Example two
According to the embodiment of the invention, the short-term wind power integrated prediction system based on error correction is disclosed, and comprises the following components:
the data acquisition module is used for acquiring operation data and weather forecast data of the wind power plant;
the model prediction module is used for inputting the data into a trained wind power prediction model and a wind power error prediction model respectively for prediction;
the data output module is used for adding the prediction results output by the two models to obtain a final short-term wind power prediction result;
the training data set of the wind power prediction model is historical operation data and weather data in a preset time period of the wind power plant; inputting historical operating data and weather data in a preset time period of a wind power plant into a wind power prediction model, outputting a wind power prediction value, and obtaining a power error data set based on the wind power prediction value; and the power error data set, historical operating data in a preset time period of the wind power plant and weather data are used as a training data set of the wind power error prediction model.
It should be noted that specific implementation manners of the modules are already described in detail in the first embodiment, and are not described again.
EXAMPLE III
According to an embodiment of the present invention, an embodiment of a terminal device is disclosed, which includes a processor and a memory, the processor being configured to implement instructions; the memory is used for storing a plurality of instructions, and the instructions are suitable for being loaded by the processor and executing the short-term wind power integration prediction method based on error correction in the first embodiment.
In other embodiments, a computer-readable storage medium is disclosed, wherein a plurality of instructions are stored, and the instructions are suitable for being loaded by a processor of a terminal device and executing the short-term wind power integration prediction method based on error correction in the first embodiment.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.
Claims (10)
1. A short-term wind power integration prediction method based on error correction is characterized by comprising the following steps:
acquiring operation data and weather forecast data of a wind power plant;
inputting the data into a trained wind power prediction model and a wind power error prediction model respectively for prediction;
adding the prediction results output by the two models to obtain a final short-term wind power prediction result;
the training data set of the wind power prediction model is historical operation data and weather data in a preset time period of the wind power plant; inputting historical operating data and weather data in a preset time period of a wind power plant into a wind power prediction model, outputting a wind power prediction value, and obtaining a power error data set based on the wind power prediction value; and the power error data set, historical operating data in a preset time period of the wind power plant and weather data are used as a training data set of the wind power error prediction model.
2. The short-term wind power integration prediction method based on error correction as claimed in claim 1, wherein the wind farm operational data and weather forecast data include but are not limited to power generation, wind direction, wind speed, temperature, humidity and barometric pressure data of a wind farm.
3. The short-term wind power integration prediction method based on error correction as claimed in claim 1, wherein the wind power prediction model is constructed by using an XGboost model optimized by an improved genetic algorithm.
4. The short-term wind power integration prediction method based on error correction as claimed in claim 3, characterized in that during construction of the wind power prediction model, binary coding is adopted, the minimum leaf node weight, the maximum depth, the learning rate and the gamma value of the XGboost model are selected as independent variable parameters, population is initialized in a random mode, each obtained chromosome contains the independent variable parameters of the XGboost model, the normalized mean absolute error NMAE minimum of the model prediction result is used as the model fitness, and then the optimal solution of the independent variable parameters of the XGboost model is found through replication, timely change, improved intersection and variation.
5. The short-term wind power integration prediction method based on error correction as claimed in claim 1, wherein the wind power error prediction model is constructed by using a random forest model optimized by an improved genetic algorithm.
6. The short-term wind power integration prediction method based on error correction as claimed in claim 1, characterized in that, during construction of the wind power error prediction model, binary coding is adopted, the number of trees of the random forest model is selected, the feature number randomly selected by each decision tree, the deepest depth of the trees is used as a parameter, the population is initialized in a random manner, each obtained chromosome contains three parameters of the random forest model, the normalized mean absolute error NMAE of the model prediction result is minimized as the model fitness, and then the optimal solution of the three parameters of the random forest model is found through replication, timely-varying improved intersection and variation.
7. The short-term wind power integrated prediction method based on error correction as claimed in claim 1, characterized in that the operation and numerical weather forecast data of the wind farm are collected in real time, the trained XGboost and random forest models are respectively used for prediction, a wind power preliminary prediction result and a wind power error prediction result are correspondingly obtained, and the final wind power prediction result can be obtained by adding the two model prediction results.
8. A short-term wind power integrated prediction system based on error correction is characterized by comprising the following components:
the data acquisition module is used for acquiring operation data and weather forecast data of the wind power plant;
the model prediction module is used for inputting the data into a trained wind power prediction model and a wind power error prediction model respectively for prediction;
the data output module is used for adding the prediction results output by the two models to obtain a final short-term wind power prediction result;
the training data set of the wind power prediction model is historical operation data and weather data in a preset time period of the wind power plant; inputting historical operating data and weather data in a preset time period of a wind power plant into a wind power prediction model, outputting a wind power prediction value, and obtaining a power error data set based on the wind power prediction value; and the power error data set, historical operating data in a preset time period of the wind power plant and weather data are used as a training data set of the wind power error prediction model.
9. A terminal device comprising a processor and a memory, the processor being arranged to implement instructions; the memory is used for storing a plurality of instructions, wherein the instructions are suitable for being loaded by the processor and executing the short-term wind power integration prediction method based on error correction according to any one of claims 1-8.
10. A computer-readable storage medium having stored thereon a plurality of instructions, wherein the instructions are adapted to be loaded by a processor of a terminal device and to perform the error correction based short-term wind power integration prediction method of any one of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110608220.2A CN113361761A (en) | 2021-06-01 | 2021-06-01 | Short-term wind power integration prediction method and system based on error correction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110608220.2A CN113361761A (en) | 2021-06-01 | 2021-06-01 | Short-term wind power integration prediction method and system based on error correction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113361761A true CN113361761A (en) | 2021-09-07 |
Family
ID=77530814
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110608220.2A Pending CN113361761A (en) | 2021-06-01 | 2021-06-01 | Short-term wind power integration prediction method and system based on error correction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113361761A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114077929A (en) * | 2021-11-24 | 2022-02-22 | 山东大学 | Wind power prediction method and system based on IS-ARIMA-LSTM prediction model |
| CN115660898A (en) * | 2022-12-06 | 2023-01-31 | 中国华能集团清洁能源技术研究院有限公司 | SVR-based classification type wind power short-term power prediction precision improving method and device |
| CN115936924A (en) * | 2022-12-14 | 2023-04-07 | 广西电网有限责任公司 | Wind power plant wind energy prediction method and system |
| CN117039886A (en) * | 2023-08-28 | 2023-11-10 | 南京图德科技有限公司 | Improved random algorithm for new energy power generation prediction error |
| WO2024087876A1 (en) * | 2022-10-28 | 2024-05-02 | 国网湖南省电力有限公司电力科学研究院 | Wind power forecasting error determination method, and processor and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102496927A (en) * | 2011-12-16 | 2012-06-13 | 中国电力科学研究院 | Wind power station power projection method based on error statistics modification |
| CN109067586A (en) * | 2018-08-16 | 2018-12-21 | 海南大学 | Ddos attack detection method and device |
| CN110601250A (en) * | 2018-06-12 | 2019-12-20 | 南京工程学院 | Method for evaluating stability of static voltage of access point based on new energy power station |
| CN112633421A (en) * | 2021-03-09 | 2021-04-09 | 国网江西综合能源服务有限公司 | Method and device for detecting abnormal electricity utilization behavior of user |
-
2021
- 2021-06-01 CN CN202110608220.2A patent/CN113361761A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102496927A (en) * | 2011-12-16 | 2012-06-13 | 中国电力科学研究院 | Wind power station power projection method based on error statistics modification |
| CN110601250A (en) * | 2018-06-12 | 2019-12-20 | 南京工程学院 | Method for evaluating stability of static voltage of access point based on new energy power station |
| CN109067586A (en) * | 2018-08-16 | 2018-12-21 | 海南大学 | Ddos attack detection method and device |
| CN112633421A (en) * | 2021-03-09 | 2021-04-09 | 国网江西综合能源服务有限公司 | Method and device for detecting abnormal electricity utilization behavior of user |
Non-Patent Citations (1)
| Title |
|---|
| HUAN ZHENG等: ""A XGBoost Model with Weather Similarity Analysis and Feature Engineering for Short-Term Wind"", 《APPLIED SCIENCES》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114077929A (en) * | 2021-11-24 | 2022-02-22 | 山东大学 | Wind power prediction method and system based on IS-ARIMA-LSTM prediction model |
| WO2024087876A1 (en) * | 2022-10-28 | 2024-05-02 | 国网湖南省电力有限公司电力科学研究院 | Wind power forecasting error determination method, and processor and storage medium |
| CN115660898A (en) * | 2022-12-06 | 2023-01-31 | 中国华能集团清洁能源技术研究院有限公司 | SVR-based classification type wind power short-term power prediction precision improving method and device |
| CN115936924A (en) * | 2022-12-14 | 2023-04-07 | 广西电网有限责任公司 | Wind power plant wind energy prediction method and system |
| CN115936924B (en) * | 2022-12-14 | 2023-08-25 | 广西电网有限责任公司 | Wind energy prediction method and system for wind power plant |
| CN117039886A (en) * | 2023-08-28 | 2023-11-10 | 南京图德科技有限公司 | Improved random algorithm for new energy power generation prediction error |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113361761A (en) | Short-term wind power integration prediction method and system based on error correction | |
| CN113282122B (en) | Commercial building energy consumption prediction optimization method and system | |
| CN109886465B (en) | Power distribution network load prediction method based on intelligent electric meter user cluster analysis | |
| CN107346459B (en) | Multi-mode pollutant integrated forecasting method based on genetic algorithm improvement | |
| CN111119282A (en) | Pressure monitoring point optimal arrangement method for water supply pipe network | |
| CN111639838B (en) | Water quality monitoring point layout optimization method suitable for water supply pipe network | |
| CN108399470B (en) | Indoor PM2.5 prediction method based on multi-example genetic neural network | |
| CN113177673B (en) | Air conditioner cold load prediction optimization method, system and equipment | |
| CN109242200B (en) | Wind power interval prediction method of Bayesian network prediction model | |
| CN115986728A (en) | Power grid situation prediction method considering uncertainty factors and terminal | |
| CN112307672A (en) | BP neural network short-term wind power prediction method based on cuckoo algorithm optimization | |
| CN116842337A (en) | Transformer fault diagnosis method based on LightGBM (gallium nitride based) optimal characteristics and COA-CNN (chip on board) model | |
| CN113705922A (en) | Improved ultra-short-term wind power prediction algorithm and model establishment method | |
| Shang et al. | Research on intelligent pest prediction of based on improved artificial neural network | |
| CN116702937A (en) | Photovoltaic output day-ahead prediction method based on K-means mean value clustering and BP neural network optimization | |
| Alrashidi et al. | Short-term PV output forecasts with support vector regression optimized by cuckoo search and differential evolution algorithms | |
| CN108734349B (en) | Improved genetic algorithm-based distributed power supply location and volume optimization method and system | |
| CN114611757A (en) | Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network | |
| CN114897204A (en) | Method and device for predicting short-term wind speed of offshore wind farm | |
| CN108537581B (en) | Energy consumption time series prediction method and device based on GMDH selective combination | |
| CN113762591A (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM counterstudy | |
| CN113962819A (en) | Method for predicting dissolved oxygen in industrial aquaculture based on extreme learning machine | |
| CN114065646B (en) | Energy consumption prediction method based on hybrid optimization algorithm, cloud computing platform and system | |
| CN115622056B (en) | Energy storage optimal configuration method and system based on linear weighting and selection method | |
| Wu et al. | Combined IXGBoost-KELM short-term photovoltaic power prediction model based on multidimensional similar day clustering and dual decomposition |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210907 |
|
| RJ01 | Rejection of invention patent application after publication |