CN112288157A - Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning - Google Patents
Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning Download PDFInfo
- Publication number
- CN112288157A CN112288157A CN202011164303.9A CN202011164303A CN112288157A CN 112288157 A CN112288157 A CN 112288157A CN 202011164303 A CN202011164303 A CN 202011164303A CN 112288157 A CN112288157 A CN 112288157A
- Authority
- CN
- China
- Prior art keywords
- data
- wind
- reinforcement learning
- similarity
- wind speed
- 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 44
- 230000002787 reinforcement Effects 0.000 title claims abstract description 43
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000013528 artificial neural network Methods 0.000 claims abstract description 14
- 238000013135 deep learning Methods 0.000 claims abstract description 14
- 239000000284 extract Substances 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 17
- 238000011217 control strategy Methods 0.000 claims description 10
- 238000007405 data analysis Methods 0.000 claims description 9
- 238000013473 artificial intelligence Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 125000004122 cyclic group Chemical group 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 5
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 claims description 2
- 238000005259 measurement Methods 0.000 claims 1
- 230000004304 visual acuity Effects 0.000 claims 1
- 238000012067 mathematical method Methods 0.000 abstract description 3
- 238000010248 power generation Methods 0.000 abstract description 3
- 230000000875 corresponding effect Effects 0.000 description 12
- 230000002596 correlated effect Effects 0.000 description 4
- 230000001965 increasing effect Effects 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 230000005684 electric field Effects 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/02—Computing arrangements based on specific mathematical models using fuzzy logic
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Biophysics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Economics (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Fuzzy Systems (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Development Economics (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Automation & Control Theory (AREA)
- Game Theory and Decision Science (AREA)
- Algebra (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
Abstract
The invention belongs to the field of new energy power generation, in particular to a wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning, aiming at the problem that the existing prediction is difficult to accurately predict, the following scheme is proposed, and the method comprises the following steps: s1, analyzing the correlation between temperature, humidity, wind power level and wind speed by a mathematical method; s2, researching and analyzing a large amount of measured data by using a fuzzy clustering algorithm, classifying historical samples with complex and various data according to the similarity of the samples, and selecting data with larger similarity with wind speed as training data of a prediction model; s3, constructing a deep reinforcement learning neural network, wherein deep learning mainly analyzes input historical sample information and further extracts corresponding characteristic information from the historical sample information.
Description
Technical Field
The invention relates to the technical field of new energy power generation, in particular to a wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning.
Background
With the increasing exhaustion of the traditional energy sources, new energy and renewable energy technologies are gradually developed, wherein the solar energy and wind energy resources are rich, but the instability of grid-connected power generation of the solar energy and wind energy is caused by the irregular change of the solar energy and the wind energy along with seasons, weather, environment and the like, and the instability is a bottleneck factor restricting the safe and stable operation of the whole power system, the power prediction is rapidly and accurately carried out, and the safe, ordered and stable operation of a power grid can be realized.
Wind energy has strong randomness and volatility, and the two characteristics can cause the wind power accessed into a power grid to have severe fluctuation characteristics, so that a series of problems are brought to a system, and the economic, safe, stable and reliable operation state of the power system can be destroyed in serious cases, so that accurate and effective wind power plant wind speed and power prediction can enable operators to make a scheduling plan in advance, arrange unit output, system standby and other related measures to reduce the influence caused by wind power grid connection, but due to numerous factors influencing the wind speed and the variability of wind power, the current prediction is difficult to accurately predict, only can be continuously improved in the aspects of algorithm research, model optimization, useful information extraction and the like, and along with abnormal meteorological event prediction, supply and demand pressure period and balance analysis, energy transaction and price prediction, medium and long-term operation and maintenance plan optimization, The increasing demands of electric field cash flow prediction, financial management and the like, and medium-and long-term prediction also get more and more attention and attention due to the quantitative decision support capability of the electric field cash flow prediction and financial management and the like.
Disclosure of Invention
The invention aims to solve the problem that accurate prediction is difficult to achieve in the prior art, and provides a wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning.
In order to achieve the purpose, the invention adopts the following technical scheme:
a wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning comprises the following steps:
s1, analyzing the correlation between temperature, humidity, wind power level and wind speed by a mathematical method;
s2, researching and analyzing a large amount of measured data by using a fuzzy clustering algorithm, classifying various historical samples according to the similarity of the samples, and selecting data with larger similarity to wind speed as training data of a prediction model;
s3, constructing a deep reinforcement learning neural network, wherein deep learning mainly analyzes input historical sample information and further extracts corresponding characteristic information from the historical sample information, and reinforcement learning forms a new combined prediction model for wind power plant power prediction based on the characteristic information obtained by the deep learning further according to control method rules and a trained artificial intelligence system, so that twice optimization of the prediction model is realized.
Preferably, in S1, a concept of a correlation coefficient in probability statistics is applied, a two-dimensional random vector is formed by a minimum wind speed of a maximum wind speed, an average wind speed, a maximum temperature, a minimum temperature, and an average wind power level of the maximum wind speed and the average wind speed, and a correlation coefficient of the two-dimensional random vector is calculated to reflect a correlation degree between the wind speed and a main meteorological factor, where an index for measuring the correlation degree between variables is the correlation coefficient, the correlation coefficient is generally represented by ρ and is a dimensionless numerical value, and a value range of ρ is [ -1,1], | ρ | is larger, indicating that a linear correlation degree between the two variables is higher; conversely, the lower the linear correlation degree between the two variables, wherein the lowest temperature is slightly positively correlated with the wind speed, and the wind speed may increase due to the rise of the temperature; the highest temperature is in negative correlation with the wind speed, and the wind speed is reduced when the highest temperature rises; the average wind speed is positively correlated, the wind speed is increased by enhancing the wind power, and the correlation with the average wind speed is maximum; in winter, the correlation between the wind speed and the temperature is the smallest in spring, and in summer and autumn, the correlation between the wind speed and the wind power level is larger. Preferably, in S2, the specific processing procedure of fuzzy clustering is as follows:
selecting representative features with stronger resolution as statistical indexes of data classification, and standardizing the data:
is the average value of the data, δ is the standard deviation of the data, assuming that the set of things to be classified is X ═ X1,x2,...,xnAssigning values to the elements in the set according to actual conditions and standards, assigning the values to a number between 0 and 1 as a similarity coefficient, wherein the similarity coefficient represents the similarity degree between the elements, and the process of determining the similarity coefficient is called calibration;
set X ═ X1,x2,...,xnIn which xi={ui1,ui2,...,uim},ui1,ui2,…,uimIs a set of characteristic factor data which can be used to characterize xi。xiAnd xjThe coefficient of similarity between is rij(0≤rij1) when r isij=0,xiAnd xjThe similarity of (A) is 0, and the two are completely different; when r isijWhen 1, xiAnd xjThe similarity of (A) is 1, the two are the same, rijThe value of (d) can be chosen by the number product method:
because dimensions and magnitude of each characteristic index are different, wind speed and temperature data need to be normalized;
after the required data is normalized, a fuzzy similarity relation is established for the historical data of the previous day, the Euclidean distance between samples is calculated, the category with the shortest Euclidean distance is taken as the category to be predicted, the influence of weather related factors can be fully considered, and the prediction precision is effectively improved.
Preferably, in S3, the preprocessed training data is input into the deep reinforcement learning convolutional neural network, the feature information corresponding to the input data information is extracted by using the powerful data analysis and learning capability of the artificial intelligence, and the target of power prediction control is achieved according to the control method rule and the trained system.
Preferably, in S3, the accurate power prediction of the wind power system is taken as a target, a priori training sample data is used as an input of the deep reinforcement learning and control network, and a convolution cyclic neural network controller is obtained on the basis of a large number of sample training by using functions of data analysis and intelligent learning, the control strategy evaluator generates a sample and an estimated return function according to a reward function of the wind power system at the acquired k time, the weight update rate is based on the obtained return function to obtain a corresponding deep learning network weight, and the predicted refined wind power plant power is output through continuous learning and iteration by the proposed deep reinforcement learning cooperative control method.
Compared with the prior art, the invention has the beneficial effects that:
the method comprises the steps of processing conventional historical data by using a fuzzy clustering technology, selecting wind speed data with a high similarity value with prediction for several days, inputting the selected data serving as training data into a deep reinforcement learning network, extracting characteristic information corresponding to input data information by using artificial intelligent powerful data analysis and learning capacity, realizing the target of power prediction control according to a control method rule and a trained system, optimizing the data and the network, and combining the advantages of the two to obtain a more accurate power prediction technology.
Drawings
FIG. 1 is a power prediction control schematic diagram of a wind farm power prediction method based on fuzzy clustering and deep reinforcement learning according to the present invention;
FIG. 2 is a deep reinforcement learning control method and a working schematic diagram of a wind farm power prediction method based on fuzzy clustering and deep reinforcement learning provided by the invention;
FIG. 3 is a circular neural network structure diagram of a wind farm power prediction method based on fuzzy clustering and deep reinforcement learning according to the present invention;
FIG. 4 is a flowchart of a wind farm power prediction method based on fuzzy clustering and deep reinforcement learning according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
Referring to fig. 1-4, a wind farm power prediction method based on fuzzy clustering and deep reinforcement learning includes the following steps:
s1, analyzing the correlation between temperature, humidity, wind power level and wind speed by a mathematical method;
s2, researching and analyzing a large amount of measured data by using a fuzzy clustering algorithm, classifying various historical samples according to the similarity of the samples, and selecting data with larger similarity to wind speed as training data of a prediction model;
s3, constructing a deep reinforcement learning neural network, wherein deep learning mainly analyzes input historical sample information and further extracts corresponding characteristic information from the historical sample information, and reinforcement learning forms a new combined prediction model for wind power plant power prediction based on the characteristic information obtained by the deep learning further according to control method rules and a trained artificial intelligence system, so that twice optimization of the prediction model is realized.
In the invention, in S1, a concept of a correlation coefficient in probability statistics is applied, a two-dimensional random vector is formed by the minimum wind speed of the maximum wind speed, the average wind speed, the highest temperature, the lowest temperature and the average wind power level which are the same as the minimum wind speed and the average wind speed, the correlation coefficient of the two-dimensional random vector is calculated to reflect the correlation degree of the wind speed and the main meteorological factors, the index for measuring the correlation degree between the variables is the correlation coefficient, the correlation coefficient is generally represented by rho and is a dimensionless numerical value, the greater the value range of the correlation coefficient is [ -1,1], | rho | value is, the higher the linear correlation degree between the two variables is; conversely, the lower the linear correlation degree between the two variables, wherein the lowest temperature is slightly positively correlated with the wind speed, and the wind speed may increase due to the rise of the temperature; the highest temperature is in negative correlation with the wind speed, and the wind speed is reduced when the highest temperature rises; the average wind speed is positively correlated, the wind speed is increased by enhancing the wind power, and the correlation with the average wind speed is maximum; in winter, the correlation between the wind speed and the temperature is the smallest in spring, and in summer and autumn, the correlation between the wind speed and the wind power level is larger.
In the present invention, in S2, the specific processing procedure of fuzzy clustering is as follows:
selecting representative features with stronger resolution as statistical indexes of data classification, and standardizing the data:
is the average value of the data, δ is the standard deviation of the data, assuming that the set of things to be classified is X ═ X1,x2,…,xnAssigning values to the elements in the set according to actual conditions and standards, assigning the values to a number between 0 and 1 as a similarity coefficient, wherein the similarity coefficient represents the similarity degree between the elements, and the process of determining the similarity coefficient is called calibration;
set X ═ X1,x2,…,xnIn which xi={ui1,ui2,...,uim},ui1,ui2,...,uimIs a set of characteristic factor data which can be used to characterize xi。xiAnd xjThe coefficient of similarity between is rij(0≤rij1) when r isij=0,xiAnd xjThe similarity of (A) is 0, and the two are completely different; when r isijWhen 1, xiAnd xjThe similarity of (A) is 1, the two are the same, rijThe value of (d) can be chosen by the number product method:
because dimensions and magnitude of each characteristic index are different, wind speed and temperature data need to be normalized;
after the required data is normalized, a fuzzy similarity relation is established for the historical data of the previous day, the Euclidean distance between samples is calculated, the category with the shortest Euclidean distance is taken as the category to be predicted, the influence of weather related factors can be fully considered, and the prediction precision is effectively improved.
In the invention, in S3, the preprocessed training data is input into a deep reinforcement learning convolutional neural network, the corresponding characteristic information of the input data information is extracted by utilizing the powerful data analysis and learning capacity of artificial intelligence, and the target of power prediction control is realized according to the control method rule and the trained system.
In the invention, in S3, accurate power prediction of a wind power system is taken as a target, priori training sample data is taken as input of a deep reinforcement learning and control network, functions of data analysis and intelligent learning are utilized, a convolutional recurrent neural network controller is obtained on the basis of training of a large number of samples, a control strategy evaluator generates a sample and an estimated return function according to a reward function of the wind power system at the acquired k moment, the weight updating rate is based on the obtained return function to obtain corresponding weight of the deep learning network, and the predicted refined wind power plant power is output through continuous learning and iteration of the proposed deep reinforcement learning cooperative control method.
Because the dimension and the magnitude of the characteristic index are different, normalization processing is required, firstly, the wind speed is normalized through a formula (3), and the temperature data is normalized according to a formula (4)
Let the cut-in wind speed v of the fanCutting intoAt a cut-out wind speed v of 3m/sCutting outIs 25 ofm/s, rated wind speed vmAt 14m/s, the wind speed of each level is normalized according to equation (5):
the 12 grades are normalized given by equations (4) (5): grades 0, 1, 2, 10, 11, 12 are normalized to 0, grades 7, 8, 9 are normalized to 1, grade 3 is normalized to 0.0036-0.218, grade 4 is normalized to 0.227-0.445, grade 5 is normalized to 0.456-0.7, and grade 6 is normalized to 0.709-0.982.
After the data normalization processing is formed, a fuzzy clustering similar relation needs to be established for the historical data of the previous 30 days, and the Euclidean distance between samples is calculated by using a formula (6):
wherein λkAs the index weight, since the similarity of the samples is inversely proportional to the euclidean distance, the worse the similarity of the samples, the larger the euclidean distance, the more the euclidean distance and the fuzzy similarity form a mapping relationship, that is:
obtaining a fuzzy similarity matrix formed by the formula (7) and including fuzzy similarity between all samples, and finding the minimum mu in the matrix RijIt can be determined that i, j among all samples is the sample with the smallest similarity. Setting the samples as clustering centers, calculating distance matrixes between all the samples and the clusters according to a formula (3), and calculating distance matrixes respectively belonging to U according to a formula (6)iClass, UjClass membership matrix M, the membership matrix can be divided into UiClass W being a cluster centeriAnd with UjW as cluster centerjAnd (4) class. The category with the shortest Euclidean distance is taken as the category to be predicted, so that the weather can be fully consideredAnd the prediction precision is effectively improved under the influence of relevant factors.
The invention is to analyze the relation between wind speed and temperature, wind power grade and humidity, and the main point is to perform correlation analysis on the function of wind speed and each meteorological factor.
The processed data are input into a deep reinforcement learning network, the deep reinforcement learning network is constructed by utilizing the powerful data analysis and learning capacity of artificial intelligence, and a relatively accurate prediction result is finally obtained, so that the aim of predicting the accurate power of the wind power plant is fulfilled. The control method and the working principle of the invention are shown in figure 1.
In fig. 1, the wind power prediction system is divided into two layers, namely a fuzzy clustering control layer and a deep reinforcement learning network control layer. The fuzzy clustering control layer is mainly used for classifying historical samples with various data according to the similarity of the samples, and selecting the data with larger similarity with a required prediction base sample as the input delta r of the deep reinforcement learning network control layer1,Δr2,…,Δrn(control variable Δ r)iWhich may be wind speed, air temperature, power, etc.). The specific process is as follows: firstly, aiming at accurate power prediction of a wind power system, training sample data and prediction information of renewable energy sources/loads are acquired in a priori mode; secondly, the sample data information after fuzzy clustering is used as the input of a deep learning and control network; and then, obtaining the deep learning network controller on the basis of training of a large number of samples by utilizing functions of data analysis and intelligent learning. Finally, the reference correcting variable delta r of each input sample of the wind power system is given through the processing of the controller1,Δr2,…,ΔrnTherefore, under the constraint condition of safe operation of the wind power system, the power prediction information of the fixed-point wind turbine is predicted.
Sample generation and training
The control strategy evaluation and the training weight update rate are divided into three parts, namely sample generation, return estimation and strategy update. Unlike supervised learning, reinforcement learning does not require manual collection and labeling of samples, but rather generates samples through algorithm interactions with the environment. The specific sample generation process is as follows: at time t-k, the status information y (k) is recorded as s0(ii) a Theta (k-1) is used as a weight coefficient of RNN, and s is input0Substituting RNN to obtain the corresponding reference input delta r provided by the controller1,Δr2,…,ΔrnIs marked as a0(ii) a Under the action of the controller of the bottom layer equipment, at the moment of k +1, the running state information of the bottom layer equipment is s1At this time, the RNN weight coefficient is made constant (still is theta (k-1)), and substituting into RNN obtains the corresponding reference input a provided by the controller1By analogy, a track τ is obtained with time t-k as the origin and under the action of policy RNN (θ (k-1)),(s)0,a0,s1,a1,…,sT-1,aT-1,sTThe trajectory is a sample at the moment when the reinforcement learning t is k; calculating rewards from the obtained samples to obtain an estimated return of the control strategyAnd finally, according to the estimated return of the control strategy, combining with a random-lifting Huohu-rising algorithm to obtain a weight updating value at each moment.
The deep reinforcement learning control method and the working principle of the wind power system proposed by the invention are shown in FIG. 2.
In fig. 2, the main role of the deep reinforcement learning cooperative control network is to provide corresponding reference input Δ r for each underlying device controller1,Δr2,…,Δrn(control variable Δ r)iWhich may be wind speed, air temperature, power, etc.). The specific process is as follows: at the moment t ═ k, the deep learning neural network inputs state information y (k) which is sample data, the weight parameter is theta (k-1), and the neural network outputs predicted power data; meanwhile, the control strategy evaluator generates a sample and an estimated return function according to the acquired reward function D (k) of the wind power system at the moment k, and the weight update rateAnd obtaining corresponding deep learning network weight theta (k) based on the obtained return function, and continuously learning and iterating through the proposed deep reinforcement learning cooperative control method to realize a control target.
And (3) control strategy:
in the scheme, a cyclic neural network is adopted as a control strategy, the structure of the cyclic neural network is shown in fig. 3, wherein input data signals such as the highest air temperature, the lowest air temperature, the average temperature, the maximum wind power level, the minimum wind power level and the average wind power are input into the cyclic neural network, the input of the cyclic neural network is sample data state information, and the corresponding reference input delta r is provided for the control strategy from an output layer through a hidden layer1,Δr2,…,ΔrnBecause the hidden layer of the RNN has internal circulation, dynamic information can be better processed, and the dynamic performance of the cooperative controller can be improved.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.
Claims (5)
1. A wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning is characterized by comprising the following steps:
s1, adopting a mathematical analysis method to discuss the correlation of temperature, humidity, wind power level and wind speed;
s2, researching a large amount of actual measurement data by using a fuzzy clustering algorithm, classifying historical samples of complex data according to sample similarity, and selecting data with greater similarity to wind speed as training data of a prediction model;
s3, constructing a deep reinforcement learning neural network, wherein deep learning mainly analyzes input historical sample information and further extracts corresponding characteristic information from the historical sample information, and reinforcement learning forms a new combined prediction model for wind power plant power prediction based on the characteristic information obtained by the deep learning further according to control method rules and a trained artificial intelligence system, so that twice optimization of the prediction model is realized.
2. The wind farm power prediction method based on fuzzy clustering and deep reinforcement learning of claim 1, wherein in S1, a concept of correlation coefficients in probability statistics is adopted, a two-dimensional random vector is formed by a maximum wind speed, a minimum wind speed, an average wind speed and corresponding maximum temperature, minimum temperature and average wind level, and the correlation coefficients are calculated to reflect the correlation degree of the wind speed and main meteorological factors.
3. The wind farm power prediction method based on fuzzy clustering and deep reinforcement learning as claimed in claim 1, wherein in S2, the specific processing procedure of fuzzy clustering is as follows:
standardizing the data, and selecting the characteristics with stronger resolving power as statistical indexes of data classification:
is the average of the data, δ is the standard deviation of the data, assuming the set to be classified is X ═ X1,x2,...,xnAssigning values to the elements in the set according to a standard, wherein the value is a number between 0 and 1 as a similarity coefficient, and the similarity coefficient represents the similarity degree between the elements, so that the process of determining the similarity coefficient is called calibration;
set X ═ X1,x2,...,xnIn which xi={ui1,ui2,...,uim},ui1,ui2,...,uimIs a set of characteristic factor data which can be used to characterize xi。xiAnd xjThe coefficient of similarity between is rij(0≤rij1) when r isij=0,xiAnd xjThe similarity of (A) is 0, and the two are completely different; when r isijWhen 1, xiAnd xjThe similarity of (A) is 1, the two are the same, rijThe value of (d) can be chosen by the number product method:
because the dimension and the magnitude of each characteristic index are different, the wind speed and temperature data need to be normalized;
after the required data is normalized, establishing a fuzzy similarity relation for the historical data of the previous day, calculating Euclidean distances among samples, and taking the class with the shortest Euclidean distance as the class to be predicted.
4. The wind farm power prediction method based on fuzzy clustering and deep reinforcement learning of claim 1, characterized in that in S3, the preprocessed training data are input into a deep reinforcement learning convolutional neural network, feature information corresponding to the input data information is extracted by using powerful data analysis and learning ability of artificial intelligence, and the goal of power prediction control is achieved according to control method rules and a trained system.
5. The wind farm power prediction method based on fuzzy clustering and deep reinforcement learning of claim 1, characterized in that in S3, aiming at accurate power prediction of a wind power system, a priori training sample data is used as input of a deep reinforcement learning and control network, and a convolution cyclic neural network controller is obtained based on training of a large number of samples by using functions of data analysis and intelligent learning, a control strategy evaluator generates a sample and an estimated return function according to a collected reward function of the wind power system at a time k, a corresponding deep learning network weight is obtained based on the obtained return function for a weight update rate, and the predicted refined wind farm power is continuously learned and iterated through the proposed deep reinforcement learning cooperative control method to output.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011164303.9A CN112288157A (en) | 2020-10-27 | 2020-10-27 | Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011164303.9A CN112288157A (en) | 2020-10-27 | 2020-10-27 | Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112288157A true CN112288157A (en) | 2021-01-29 |
Family
ID=74372522
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011164303.9A Pending CN112288157A (en) | 2020-10-27 | 2020-10-27 | Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112288157A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113449928A (en) * | 2021-07-14 | 2021-09-28 | 清华四川能源互联网研究院 | Wind field power prediction method, system, server and computer readable storage medium |
CN115238583A (en) * | 2022-07-27 | 2022-10-25 | 山东理工大学 | Business process remaining time prediction method and system supporting incremental logs |
CN115511227A (en) * | 2022-11-21 | 2022-12-23 | 国能日新科技股份有限公司 | Wind power generation power prediction method and device based on stable learning |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103324980A (en) * | 2013-04-25 | 2013-09-25 | 华北电力大学(保定) | Wind power station wind speed prediction method |
KR20140018497A (en) * | 2012-08-01 | 2014-02-13 | 한국전력공사 | Prediction method of short-term wind speed and wind power and power supply line voltage prediction method therefore |
CN103955755A (en) * | 2014-04-18 | 2014-07-30 | 国家电网公司 | Wind electricity power short-term prediction method by adopting composite data source based on self-learning polynomial kernel function support vector machine |
CN108932671A (en) * | 2018-06-06 | 2018-12-04 | 上海电力学院 | A kind of LSTM wind-powered electricity generation load forecasting method joined using depth Q neural network tune |
CN109325607A (en) * | 2017-08-01 | 2019-02-12 | 中国电力科学研究院 | A kind of short-term wind power forecast method and system |
CN111262275A (en) * | 2020-03-26 | 2020-06-09 | 西安电子科技大学 | Global cooperative control method for simulated space solar power station |
KR102143757B1 (en) * | 2019-05-07 | 2020-08-12 | 오토시맨틱스 주식회사 | Wind Power Generator incorporating therein AI Technology, including Deep Learning |
-
2020
- 2020-10-27 CN CN202011164303.9A patent/CN112288157A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20140018497A (en) * | 2012-08-01 | 2014-02-13 | 한국전력공사 | Prediction method of short-term wind speed and wind power and power supply line voltage prediction method therefore |
CN103324980A (en) * | 2013-04-25 | 2013-09-25 | 华北电力大学(保定) | Wind power station wind speed prediction method |
CN103955755A (en) * | 2014-04-18 | 2014-07-30 | 国家电网公司 | Wind electricity power short-term prediction method by adopting composite data source based on self-learning polynomial kernel function support vector machine |
CN109325607A (en) * | 2017-08-01 | 2019-02-12 | 中国电力科学研究院 | A kind of short-term wind power forecast method and system |
CN108932671A (en) * | 2018-06-06 | 2018-12-04 | 上海电力学院 | A kind of LSTM wind-powered electricity generation load forecasting method joined using depth Q neural network tune |
KR102143757B1 (en) * | 2019-05-07 | 2020-08-12 | 오토시맨틱스 주식회사 | Wind Power Generator incorporating therein AI Technology, including Deep Learning |
CN111262275A (en) * | 2020-03-26 | 2020-06-09 | 西安电子科技大学 | Global cooperative control method for simulated space solar power station |
Non-Patent Citations (4)
Title |
---|
宋世德: "《简明模糊数学基础》", vol. 2016, 30 April 2016, 西北农林科技大学出版社, pages: 168 - 172 * |
张传辉: "基于模糊C均值聚类和案例推理的风电功率预测研究", 中国优秀硕士学位论文全文数据库, no. 2018, 15 January 2018 (2018-01-15), pages 042 - 539 * |
梁耀光: "风电短期发电功率预测优化算法研究", 中国优秀硕士学位论文全文数据库, no. 2017, 15 June 2017 (2017-06-15), pages 042 - 364 * |
邱锡鹏: "《神经网络与深度学习》", 31 July 2020, 机械工业出版社, pages: 360 - 365 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113449928A (en) * | 2021-07-14 | 2021-09-28 | 清华四川能源互联网研究院 | Wind field power prediction method, system, server and computer readable storage medium |
CN115238583A (en) * | 2022-07-27 | 2022-10-25 | 山东理工大学 | Business process remaining time prediction method and system supporting incremental logs |
CN115238583B (en) * | 2022-07-27 | 2024-02-13 | 山东理工大学 | Business process remaining time prediction method and system supporting incremental log |
CN115511227A (en) * | 2022-11-21 | 2022-12-23 | 国能日新科技股份有限公司 | Wind power generation power prediction method and device based on stable learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110414788B (en) | Electric energy quality prediction method based on similar days and improved LSTM | |
CN107766990B (en) | Method for predicting power generation power of photovoltaic power station | |
CN112288157A (en) | Wind power plant power prediction method based on fuzzy clustering and deep reinforcement learning | |
CN105701572B (en) | Photovoltaic short-term output prediction method based on improved Gaussian process regression | |
CN116937579B (en) | Wind power interval prediction considering space-time correlation and interpretable method thereof | |
CN111461921B (en) | Load modeling typical user database updating method based on machine learning | |
CN111525587B (en) | Reactive load situation-based power grid reactive voltage control method and system | |
CN110717610A (en) | Wind power prediction method based on data mining | |
CN109272258B (en) | Regional wind and solar power generation resource evaluation method based on K-means clustering | |
CN111582548A (en) | Power load prediction method based on multivariate user behavior portrait | |
CN114021483A (en) | Ultra-short-term wind power prediction method based on time domain characteristics and XGboost | |
CN114819369A (en) | Short-term wind power prediction method based on two-stage feature selection and random forest improvement model | |
CN116526473A (en) | Particle swarm optimization LSTM-based electrothermal load prediction method | |
CN113379116A (en) | Cluster and convolutional neural network-based line loss prediction method for transformer area | |
CN111815039A (en) | Weekly scale wind power probability prediction method and system based on weather classification | |
CN115995810A (en) | Wind power prediction method and system considering weather fluctuation self-adaptive matching | |
CN115641231A (en) | Typical scene extraction method and system for power system, electronic equipment and medium | |
CN113762591A (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM counterstudy | |
CN117113086A (en) | Energy storage unit load prediction method, system, electronic equipment and medium | |
CN117096860A (en) | Overhead transmission line current-carrying capacity interval prediction method and equipment based on LSSVM model | |
CN110991723A (en) | Application method of artificial intelligence in seasonal load prediction | |
CN110163437A (en) | Day-ahead photovoltaic power generation power prediction method based on DPK-means | |
CN115689001A (en) | Short-term load prediction method based on pattern matching | |
Lv et al. | Research on short-term load forecasting approach for smart grid | |
CN115481788A (en) | Load prediction method and system for phase change energy storage system |
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 |