CN110070228B - BP neural network wind speed prediction method for neuron branch evolution - Google Patents
BP neural network wind speed prediction method for neuron branch evolution Download PDFInfo
- Publication number
- CN110070228B CN110070228B CN201910336963.1A CN201910336963A CN110070228B CN 110070228 B CN110070228 B CN 110070228B CN 201910336963 A CN201910336963 A CN 201910336963A CN 110070228 B CN110070228 B CN 110070228B
- Authority
- CN
- China
- Prior art keywords
- neural network
- wind speed
- evolution
- firefly
- population
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
Abstract
The invention discloses a BP neural network wind speed prediction method for neuron branch evolution, which comprises the steps of obtaining historical meteorological data of a wind power plant; constructing a wind speed influence characteristic factor; preprocessing the historical meteorological data, including correlation analysis, input variable selection and normalization processing; establishing a BP neural network model, training the BP neural network model by using the preprocessed historical meteorological data, adopting a grouped evolution idea in the training process, and evolving a parameter population in the neural network model by combining differential operation and a firefly algorithm; and acquiring real meteorological data of the wind power plant, inputting the real meteorological data into a neural network, and calculating a predicted wind speed value. The invention realizes grouping optimization by taking hidden layer neurons as branch references, and is supplemented by global optimization after judging grouping effects, so that the invention has relatively lower computational complexity and higher prediction accuracy compared with the traditional method.
Description
Technical Field
The invention belongs to the field of wind speed prediction of wind power plants, and particularly relates to a BP neural network wind speed prediction method for neuron branch evolution.
Background
Wind power generation, a renewable energy source, is widely spotlighted by scientists and engineers around the world. The wind power generation technology is the most mature one of renewable energy conversion technologies, particularly in China, wind power resources are very rich, and wind energy has the characteristics of cleanness, high efficiency and cyclic utilization, and wind power generation fields are established at each place at times, however, in the process of actual utilization of wind power, accurate prediction of wind power generation is found to be very important, on one hand, accurate prediction of the power of wind power generation is provided, and a very strong guiding significance is provided for scientific planning of the establishment of a power grid, so that the power grid and relay equipment thereof have better stability and adaptability, on the other hand, accurate prediction of the power of wind power generation can be achieved, electric power can be planned reasonably, excessive waste caused by energy can be prevented, and power supply accidents caused by insufficient energy can also be prevented.
Based on the above, many engineers make attempts at wind speed prediction, and many methods are proposed, for example, wandelmin et al propose a short-term wind speed prediction model based on a genetic BP neural network, which improves the neural network prediction model and is applicable to wind power prediction of a wind power plant, so that the accuracy of wind power prediction is improved to a certain extent, but due to intermittency, periodicity and volatility of wind power generation and complex environmental factors involved in wind power prediction, the problem of long calculation time of an algorithm model caused by more input variables of the neural network model is solved, and excessive input variables do not necessarily bring useful information, and meanwhile, training of the neural network is easy to fall into local optimization, the convergence speed is slow, and the problems of easy oscillation in the learning process are solved.
Disclosure of Invention
In view of this, the present invention aims to provide a neural network wind speed prediction method for neuron branch evolution with high prediction accuracy and low time complexity.
The object of the invention is achieved by the following steps:
step 1, acquiring historical meteorological data of a wind power plant;
step 2, constructing a wind speed influence characteristic factor, and calculating the numerical value of the characteristic factor;
step 3, preprocessing the data of the wind speed influence factor, including normalization processing, correlation analysis and input variable selection;
step 4, establishing a BP neural network model, and training the BP neural network model by using the preprocessed data of the characteristic factors;
step 5, calculating a characteristic factor value of the real meteorological data of the wind power plant;
step 6, inputting the characteristic factor values of the real meteorological data into a neural network, performing inverse normalization on the output values, and calculating a predicted wind speed value;
the learning training of the BP neural network described in step 4 includes:
step 401, sorting the preprocessed data of the characteristic factors according to a training set format, wherein the training set is used for training a BP neural network, each record in the training set comprises a plurality of input values and 1 output value, the input values are the preprocessed data of the characteristic factors, and the 1 output value is a wind speed value at a certain time after a historical time;
step 402, taking neurons of a hidden layer in the BP neural network as branch references, and classifying parameters to be optimized related to the neurons of each hidden layer into one branch;
step 403, initializing the nth parameter to be optimized corresponding to the nth hidden layer neuron into a set of random values, randomly generating an initial population according to the parameter to be optimized, setting other parameters to be optimized as fixed values, and setting n as a neuron iteration variable;
step 404, updating the population by using a firefly algorithm;
step 405, performing a difference operator operation on the population to update the nth sub-population, wherein the difference operator includes difference variation, difference intersection and difference selection;
step 406, detecting whether a preset maximum iteration number is reached or whether a specified precision is reached, and if not, continuing to step 404; if yes, go to step 407;
step 407, recording the values of the n-th parameter to be optimized after learning evolution, and then performing learning evolution of the (n +1) -th parameter to be optimized;
step 408, after calculating m times, obtaining all parameter values related to all m neurons in the hidden layer;
step 409, in the (m +1) th calculation, the threshold of the output layer neuron is obtained.
Further, in step 404, the updating the population using the firefly algorithm includes:
40401, mapping the population to a firefly population, wherein each individual in the population corresponds to a firefly in the firefly population, and initializing relevant parameters in a firefly algorithm;
40402, calculating the illumination intensity of each firefly;
40403, updating the positions of the fireflies according to the relationship between the illumination intensity of each firefly and the neighboring fireflies;
40404, calculating the updated firefly position and illumination intensity;
step 40405, obtaining the optimal positions of the globally optimal firefly and the individual firefly.
Further, in step 408, after m times of calculation, if the number of times of branch evolution for more than half of m times reaches the maximum iteration number and the evolution stops, global evolution is required, the process of global evolution is to select the optimal 1/m individuals in the previous m populations to form a new population, and perform joint evolution by using a firefly method and a differential evolution method, and when the preset maximum iteration number or the specified accuracy is reached, the evolution stops, and the optimal individuals in the population are the optimal results after the evolution.
The historical meteorological data in the step 1 comprise data of 5 factors of wind speed, wind direction, temperature, humidity and air pressure; the characteristic factors in the step 2 comprise wind speed, wind direction, temperature, humidity, air pressure, wind speed range, wind speed standard deviation, wind direction range, wind direction standard deviation, temperature range, temperature standard deviation, humidity range, humidity standard deviation, air pressure range and air pressure standard deviation.
The correlation analysis in the step 3 is used for analyzing the degree of closeness of the characteristic factors and the future wind speed so as to determine the optimal input variable combination; the correlation analysis adopts a covariance matrix analysis method or a Pearson correlation analysis method; and (3) selecting the input variables, wherein the selected input variables are a wind speed mean value, a wind speed range, a wind speed standard deviation, a wind direction mean value, a temperature mean value, a humidity mean value and a humidity range.
The parameters to be optimized related to each neuron in step 402 include the weights of the hidden layer neuron and all input layer neurons, the threshold of the hidden layer neuron, and the weights of the hidden layer neuron and all output layer neurons.
The method takes historical meteorological data related to wind speed as a research object, processes original data, predicts by using an improved BP neural network model, and fully utilizes the optimization ideas in a firefly algorithm and a differential evolution algorithm in the process of learning the historical data by the BP neural network, so that the BP neural network can find a global optimum value more easily in the parameter learning process, the prediction accuracy is improved, and meanwhile, the idea of neuron branch evolution is provided, so that the neural network has lower time complexity and better learning effect in the learning process.
Drawings
FIG. 1 is a schematic overall flow chart of an embodiment of the present invention;
FIG. 2 is a schematic flow chart of the parameters of the Branch evolution neural network of the present invention;
Detailed Description
The invention is further described with reference to the accompanying drawings, but the invention is not limited in any way, and any alterations or substitutions based on the teaching of the invention are within the scope of the invention.
Wind speed audience of a wind power plant is influenced by multiple factors. The general research shows that the prediction of the wind speed of the wind farm is related to at least five influencing factors of historical wind speed, wind direction, temperature, humidity and air pressure, in order to more comprehensively depict the influence of the five factors on the wind speed to be predicted, the invention expands the factors to obtain ten influence factors of wind speed extreme difference, wind speed standard deviation, wind direction extreme difference, wind direction standard deviation, temperature extreme difference, temperature standard deviation, humidity extreme difference, humidity standard deviation, air pressure extreme difference and air pressure standard deviation, constructs a characteristic factor system for influencing the wind speed, therefore, the data required to be collected for realizing the invention is the numerical value of five factors of wind speed, wind direction, temperature, humidity and air pressure in the historical period, the data processed by the invention is time series data of the above fifteen wind speed influence factors, and fifteen factor values corresponding to each historical period are called as a group of input.
The BP neural network is a multi-layer feedforward neural network trained according to an error back propagation algorithm, and is one of the most widely applied neural networks at present. The basic BP algorithm includes two processes of forward propagation of data signals and back propagation of errors. That is, the error output is calculated in the direction from the input to the output, and the weight and the threshold are adjusted in the direction from the output to the input. During forward propagation, an input signal acts on an output node through a hidden layer, an output signal is generated through nonlinear transformation, and if actual output does not accord with expected output, the process of backward propagation of errors is carried out. The error back transmission is to back transmit the output error to the input layer by layer through the hidden layer, and to distribute the error to all units of each layer, and to use the error signal obtained from each layer as the basis for adjusting the weight of each unit. Generally speaking, the adjustment of the weight and the threshold can adopt a maximum gradient descent method, a quasi-newton method, and the like, through repeated feasible domain search, the relevant parameters can be continuously optimized, and when the optimization is reached to a certain degree, the training is stopped. The BP neural network is widely applied to different practical scenes, and has strong learning and fitting capabilities for classification and prediction problems, but the traditional training method has the prominent defects that the learning speed is possibly slow, oscillation is possibly caused, and the optimization process is easy to fall into local optimization.
As shown in fig. 1, a BP neural network wind speed prediction method for neuron branch evolution with high prediction accuracy and low computational complexity in an embodiment of the present invention includes:
step 1, acquiring historical meteorological data of a wind power plant;
step 2, constructing a wind speed influence characteristic factor, and calculating a numerical value of the characteristic factor;
step 3, preprocessing the data of the wind speed influence factor, including normalization processing, correlation analysis and input variable selection;
step 4, establishing a BP neural network model, and training the BP neural network model by using the preprocessed data of the characteristic factors;
step 5, calculating a characteristic factor value of the real meteorological data of the wind power plant;
and 6, inputting the characteristic factor values of the real meteorological data into a neural network, performing inverse normalization on the output values, and calculating the predicted wind speed value.
The historical meteorological data in the step 1 comprise wind speed, wind direction, temperature, humidity and air pressure; the characteristic factors in the step 2 comprise wind speed, wind direction, temperature, humidity, air pressure, wind speed range, wind speed standard deviation, wind direction range, wind direction standard deviation, temperature range, temperature standard deviation, humidity range, humidity standard deviation, air pressure range and air pressure standard deviation.
The normalization process in step 3 is to convert the data of the characteristic factors into between [0,1], as shown in the following formula
XiRepresenting the data before normalization in the ith characteristic factor, XimaxRepresents the maximum value of data in the ith characteristic factor, XiminIs the minimum value of the data in the ith characteristic factor, X'iNormalizing the data of the ith characteristic factor; the correlation analysis is used for analyzing the degree of closeness of the characteristic factors and the future wind speed to determine the optimal input variable combination, and the correlation analysis adopts covariance and covariance matrix componentsAn analytical method or a Pearson correlation analytical method,
the training of the BP neural network model in the step 4 comprises the following steps:
step 401, the preprocessed feature factor data are arranged according to a training set format, the training set is used for training a BP neural network, each record in the training set includes a plurality of input values and 1 output value, the input values are the preprocessed feature factor data, and the 1 output value is a wind speed value at a certain time after a historical time.
Step 402, taking the neurons of the hidden layer in the BP neural network as branch references, and classifying the parameters to be optimized related to each hidden layer neuron into one branch.
The BP neural network comprises a three-layer structure, namely an input layer, a hidden layer and an output layer, wherein the hidden layer can be a plurality of layers theoretically, but when the total number of layers is more than three, the possibility that the training of the BP neural network falls into local minimum points is increased, and the BP neural network is easy to over-learn. Layers in the BP neural network are all connected, and neurons in each layer are not connected. Therefore, the invention takes the neurons of the hidden layers as the branch reference, branches the parameters to be optimized, and the parameters to be optimized related to the neurons of each hidden layer are classified into one branch. By the branch method, the association between branches is reduced to the minimum, the association of the parameters to be optimized in each branch is reduced to the minimum, and the branch evolution of the parameters is realized. For example, the parameters related to the nth neuron in the hidden layer include the weight of the nth neuron connected to all neurons in the input layer, the threshold of the nth neuron, and the weight of the nth neuron connected to all neurons in the output layer.
Step 403, initializing the nth parameter to be optimized corresponding to the nth hidden layer neuron into a set of random values, randomly generating an initial population according to the parameter to be optimized, setting other parameters to be optimized as fixed values, and setting n as a neuron iteration variable.
Step 404, updating the population with a firefly algorithm.
In step 404, said updating population group with firefly algorithm comprises,
40401, mapping the population to a firefly population, wherein each individual in the population corresponds to a firefly in the firefly population, and initializing relevant parameters in a firefly algorithm;
40402, calculating the illumination intensity of each firefly;
40403, updating the positions of the fireflies according to the relationship between the illumination intensity of each firefly and the neighboring fireflies;
40404, calculating the updated firefly position and illumination intensity;
40405, acquiring the optimal positions of the globally optimal fireflies and the individual fireflies;
and 405, performing a difference operator operation on the population to update the nth sub-population, wherein the difference operator comprises difference variation, difference intersection and difference selection. In step 405, the difference operator operates only on the relevant parameters to be optimized for the nth sub-population, without any operation on the previously determined parameters of (n-1) neurons.
Step 406, detecting whether a preset maximum iteration number is reached or whether a specified precision is reached, and if not, continuing to step 404; if yes, go to step 407;
step 407, recording the values of the n-th parameter to be optimized after learning evolution, and then performing learning evolution of the (n +1) -th parameter to be optimized.
Step 408, after m times of calculation, obtaining all parameter values related to all m neurons in the hidden layer, wherein m is the number of neurons in the hidden layer;
and step 409, obtaining the threshold value of the neuron of the output layer in the (m +1) th calculation, and ending the training of the BP neural network model. The flow of the branch evolution is shown in fig. 2.
In step 408, after m times of calculation, if the number of times of branch evolution for more than half of m times reaches the maximum iteration number and the evolution stops, global evolution is required, the global evolution process is that respectively selecting the optimal 1/m individuals in the previous m populations to form a new population, performing joint evolution by adopting a firefly method and a differential evolution method, and when the preset maximum iteration number or the specified precision is reached, stopping the evolution, and the optimal individuals in the population are the optimal results after the evolution. This step is to prevent that after branching the population, it may cause that evolution of part of the population cannot converge, which results in that a better solution cannot be obtained, and because of the superposition effect, the global solution performance is worse because more branching solutions cannot converge.
The number of input layer neurons and the number of output layer neurons needed by the neural network depends on the problem to be solved, the number of the problems is determined once the problem is determined, but the number of hidden layer neurons in the neural network can be changed, generally, the hidden layer neurons can be changed within a certain range according to the number of the input layer neurons and the output layer neurons, and the problems can be determined only by repeated experiments and optimization in engineering.
In order to better show the beneficial effects brought by the invention, an embodiment is provided for demonstrating the whole process flow of the method. In the embodiment, the data is originally acquired in a China meteorological network, 2000 groups of data with 200 days are acquired every hour, ten groups of data are acquired every day, and each group of data comprises data of five influencing factors including wind speed, wind direction, temperature, humidity and air pressure at the acquisition moment.
Through ten groups of data of each day, calculating the data of each day wind speed influence factor, namely the values of fifteen influence factors, namely a wind speed mean value, a wind direction mean value, a temperature mean value, a humidity mean value, an air pressure mean value, a wind speed range, a wind speed standard deviation, a wind direction range, a wind direction standard deviation, a temperature range, a temperature standard deviation, a humidity range, a humidity standard deviation, an air pressure range and an air pressure standard deviation, wherein the mean value is the mean value of data acquired at ten moments of a certain influence factor, the range is the difference between the maximum value and the minimum value in the data acquired at the ten moments of the certain influence factor, and the standard deviation is the square root of the arithmetic mean value of the squares of the data acquired at the ten moments of the certain influence factor and the mean value.
The method comprises the steps of converting 2000 groups of data into 200 groups of data of wind speed influence factors, preprocessing the data of the wind speed influence factors, carrying out correlation analysis on fifteen influence factors and wind speed to be predicted by adopting a Pearson correlation analysis method, obtaining 7 influence factors with the maximum correlation, wherein the 7 influence factors are respectively a wind speed mean value, a wind speed range, a wind speed standard deviation, a wind direction mean value, a temperature mean value, a humidity mean value and a humidity range, and therefore, selecting the data of the 7 influence factors as values of input variables and carrying out normalization processing.
And establishing a BP neural network model, using 190 groups of data in the 200 groups of data for training, using 7 parameters of the influence factors of the previous day, namely the wind speed mean value, the wind speed range, the wind speed standard deviation, the wind direction mean value, the temperature mean value and the humidity mean value as input parameters of the neural network, outputting the input parameters as wind speed values corresponding to the next day, and using the remaining 10 groups for testing. When the number of the neurons in the hidden layer of the model is determined to be 10, the neural network has the best prediction capability for the experimental data training set. In the experiment, the fixed value is set to 0.5. The results obtained by the experiment are shown in table 1. The actual value refers to a wind speed value to be predicted which is actually collected in a test set, the branch evolution value is a predicted value obtained by using the method, and the traditional evolution value is a predicted value obtained by only adopting a firefly algorithm and a difference operation to evolve a neural network parameter without branch evolution.
TABLE 1 comparison of actual values and predicted values for the two methods
| Serial number | Actual value | Classical evolutionary value | Branch evolution value |
| 1 | 4.9 | 5.37 | 4.75 |
| 2 | 2.5 | 2.73 | 2.27 |
| 3 | 2.7 | 1.29 | 1.91 |
| 4 | 6.5 | 5.69 | 6.17 |
| 5 | 2.7 | 4.16 | 3.59 |
| 6 | 1.6 | 1.01 | 2.34 |
| 7 | 7.4 | 7.48 | 6.58 |
| 8 | 8.3 | 8.95 | 8.66 |
| 9 | 1.7 | 1.32 | 1.91 |
| 10 | 1.5 | 1.06 | 1.88 |
As can be seen from the table, the error of the branch evolution is obviously better than that of the traditional evolution method. It can thus be seen that the branch evolution has a higher accuracy for the neural network model to proceed. In addition, the invention provides a normalization processing, correlation analysis and input variable selection method in the aspect of data preprocessing, thereby greatly reducing the training time complexity of the neural network and simultaneously still maintaining the prediction accuracy.
According to the invention and the embodiment, the BP neural network wind speed prediction method for neuron branch evolution comprises the steps of firstly collecting all climate data which can affect wind speed, constructing more comprehensive characteristic factors which affect the wind speed on the basis of original climate data, calculating the numerical values of the characteristic factors, preprocessing the data of the characteristic factors, facilitating data dimension reduction and guaranteeing the effectiveness of information, then providing a joint evolution algorithm adopting fireflies and a differential algorithm in the process of model training, decomposing a large-scale parameter training into a plurality of small-scale parameter groups when a specific parameter is optimized, and optimizing the parameter groups respectively, so that the calculation complexity is relatively lower and the prediction accuracy is higher compared with the traditional evolution method.
Claims (8)
1. A BP neural network wind speed prediction method for neuron branch evolution is characterized by comprising the following steps:
step 1, acquiring historical meteorological data of a wind power plant;
step 2, constructing a wind speed influence characteristic factor, and calculating the numerical value of the characteristic factor;
step 3, preprocessing the data of the characteristic factors, including normalization processing, correlation analysis and input variable selection;
step 4, establishing a BP neural network model, and training the BP neural network model by using the preprocessed data of the characteristic factors;
step 5, calculating a characteristic factor value of the real meteorological data of the wind power plant;
step 6, inputting the characteristic factor values of the real meteorological data into a neural network, performing inverse normalization on the output values, and calculating a predicted wind speed value;
the learning training of the BP neural network described in step 4 includes:
step 401, sorting the preprocessed data of the characteristic factors according to a training set format, wherein the training set is used for training a BP neural network, each record in the training set comprises a plurality of input values and 1 output value, the input values are the preprocessed data of the characteristic factors, and the 1 output value is a wind speed value at a certain time after a historical time;
step 402, taking neurons of a hidden layer in the BP neural network as branch references, and classifying parameters to be optimized related to the neurons of each hidden layer into one branch;
step 403, initializing the nth parameter to be optimized corresponding to the nth hidden layer neuron into a set of random values, randomly generating an initial population according to the parameter to be optimized, setting other parameters to be optimized as fixed values, and setting n as a neuron iteration variable;
step 404, updating the population by using a firefly algorithm;
step 405, performing a difference operator operation on the population to update the nth sub-population, wherein the difference operator includes difference variation, difference intersection and difference selection;
step 406, detecting whether a preset maximum iteration number is reached or whether a specified precision is reached, and if not, continuing to step 404; if yes, go to step 407;
step 407, recording the values of the n-th parameter to be optimized after learning evolution, and then performing learning evolution of the (n +1) -th parameter to be optimized;
step 408, after m times of calculation, obtaining all parameter values related to all m neurons in the hidden layer, wherein m is the number of neurons in the hidden layer;
step 409, in the (m +1) th calculation, the threshold of the output layer neuron is obtained.
2. The BP neural network wind speed prediction method of claim 1, wherein the updating the population using a firefly algorithm in step 404 comprises:
40401, mapping the population to a firefly population, wherein each individual in the population corresponds to a firefly in the firefly population, and initializing relevant parameters in a firefly algorithm;
40402, calculating the illumination intensity of each firefly;
40403, updating the positions of the fireflies according to the relationship between the illumination intensity of each firefly and the neighboring fireflies;
40404, calculating the updated firefly position and illumination intensity;
step 40405, obtaining the optimal positions of the globally optimal firefly and the individual firefly.
3. The method for predicting the wind speed of the BP neural network according to claim 2, wherein in step 408, after m times of computation, if the m times of branch evolution is half or more times of branch evolution and the evolution stops when the maximum iteration number is reached, then global evolution is required, wherein the global evolution process comprises the steps of respectively selecting the optimal 1/m individuals in the previous m populations to form a new population, performing joint evolution by using a firefly method and a differential evolution method, and stopping the evolution when the preset maximum iteration number is reached or a specified precision is reached, wherein the optimal individuals in the population are the optimal results after the evolution.
4. The BP neural network wind speed prediction method according to claim 2 or 3, wherein the historical meteorological data in step 1 comprises 5 factors of wind speed, wind direction, temperature, humidity and air pressure; the characteristic factors in the step 2 comprise wind speed, wind direction, temperature, humidity, air pressure, wind speed range, wind speed standard deviation, wind direction range, wind direction standard deviation, temperature range, temperature standard deviation, humidity range, humidity standard deviation, air pressure range and air pressure standard deviation.
5. The BP neural network wind speed prediction method according to claim 4, wherein the correlation analysis in step 3 is used to analyze how closely the characteristic factors are correlated with future wind speeds to determine the optimal input variable combination; the correlation analysis adopts a covariance matrix analysis method or a Pearson correlation analysis method; and (3) selecting the input variables, wherein the selected input variables are a wind speed mean value, a wind speed range, a wind speed standard deviation, a wind direction mean value, a temperature mean value, a humidity mean value and a humidity range.
6. The method of claim 5, wherein the parameters to be optimized associated with each of the neurons in step 402 comprise the weights of the hidden layer neuron and all input layer neurons, the threshold value of the hidden layer neuron, and the weights of the hidden layer neuron and all output layer neurons.
7. The BP neural network wind speed prediction method according to claim 6, wherein the number of neurons in an implicit layer in the neural network model is determined to be 10, and the parameters to be optimized of the neural network are classified into 10.
8. The BP neural network wind speed prediction method according to claim 7, wherein the fixed value in step 403 is 0.5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910336963.1A CN110070228B (en) | 2019-04-25 | 2019-04-25 | BP neural network wind speed prediction method for neuron branch evolution |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910336963.1A CN110070228B (en) | 2019-04-25 | 2019-04-25 | BP neural network wind speed prediction method for neuron branch evolution |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110070228A CN110070228A (en) | 2019-07-30 |
| CN110070228B true CN110070228B (en) | 2021-06-15 |
Family
ID=67368799
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910336963.1A Active CN110070228B (en) | 2019-04-25 | 2019-04-25 | BP neural network wind speed prediction method for neuron branch evolution |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110070228B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112348247B (en) * | 2020-11-03 | 2022-04-15 | 深圳大学 | Prediction method and device for light energy power generation, computer equipment and storage medium |
| CN112348700B (en) * | 2020-11-05 | 2022-11-29 | 国家电网公司华中分部 | Line capacity prediction method combining SOM clustering and IFOU equation |
| CN113191091B (en) * | 2021-06-03 | 2022-12-09 | 上海交通大学 | Wind speed prediction method, system and equipment based on hybrid deep learning mechanism |
| CN113719283B (en) * | 2021-09-07 | 2023-01-17 | 武汉理工大学 | Method and device for predicting working hours of mine rock drilling equipment |
| CN117195747B (en) * | 2023-11-02 | 2024-01-23 | 北京金鼎兴成磁性材料有限公司 | Uniform heat distribution optimization method for magnetic material drying |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0995156A2 (en) * | 1997-07-09 | 2000-04-26 | Siemens Aktiengesellschaft | Method and structure for the neural modelling of a dynamic system in a computer |
| CN101706335A (en) * | 2009-11-11 | 2010-05-12 | 华南理工大学 | Wind power forecasting method based on genetic algorithm optimization BP neural network |
| CN102663493A (en) * | 2012-03-23 | 2012-09-12 | 天津工业大学 | Delaying nerve network used for time sequence prediction |
| CN105574615A (en) * | 2015-12-16 | 2016-05-11 | 南京理工大学 | Spatial correlation and genetic algorithm (GA) based wind power forecast method for wavelet-BP neural network |
| JP2016136001A (en) * | 2015-01-23 | 2016-07-28 | 中国電力株式会社 | Prediction device |
-
2019
- 2019-04-25 CN CN201910336963.1A patent/CN110070228B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0995156A2 (en) * | 1997-07-09 | 2000-04-26 | Siemens Aktiengesellschaft | Method and structure for the neural modelling of a dynamic system in a computer |
| CN101706335A (en) * | 2009-11-11 | 2010-05-12 | 华南理工大学 | Wind power forecasting method based on genetic algorithm optimization BP neural network |
| CN102663493A (en) * | 2012-03-23 | 2012-09-12 | 天津工业大学 | Delaying nerve network used for time sequence prediction |
| JP2016136001A (en) * | 2015-01-23 | 2016-07-28 | 中国電力株式会社 | Prediction device |
| CN105574615A (en) * | 2015-12-16 | 2016-05-11 | 南京理工大学 | Spatial correlation and genetic algorithm (GA) based wind power forecast method for wavelet-BP neural network |
Non-Patent Citations (1)
| Title |
|---|
| 基于气象因子的 BP神经网络风电场风速预测;基于气象因子的 BP神经网络风电场风速预测;《可再生能源》;20091025;第86-89页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110070228A (en) | 2019-07-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110070228B (en) | BP neural network wind speed prediction method for neuron branch evolution | |
| CN110070226B (en) | Photovoltaic power prediction method and system based on convolutional neural network and meta-learning | |
| CN111160520A (en) | BP neural network wind speed prediction method based on genetic algorithm optimization | |
| CN105574615B (en) | wavelet-BP neural network wind power prediction method based on spatial correlation and GA | |
| CN111680820B (en) | Distributed photovoltaic power station fault diagnosis method and device | |
| CN109978283B (en) | Photovoltaic power generation power prediction method based on branch evolution neural network | |
| CN111929748A (en) | Meteorological element forecasting method and system | |
| CN113471982B (en) | Cloud edge cooperation and power grid privacy protection distributed power supply in-situ voltage control method | |
| CN113468817A (en) | Ultra-short-term wind power prediction method based on IGOA (optimized El-electric field model) | |
| CN112215428B (en) | Photovoltaic power generation power prediction method and system based on error correction and fuzzy logic | |
| CN109242136A (en) | A kind of micro-capacitance sensor wind power Chaos-Genetic-BP neural network prediction technique | |
| Chen et al. | Multiperiod-ahead wind speed forecasting using deep neural architecture and ensemble learning | |
| Chen et al. | Research on wind power prediction method based on convolutional neural network and genetic algorithm | |
| CN109408896B (en) | Multi-element intelligent real-time monitoring method for anaerobic sewage treatment gas production | |
| Buonanno et al. | Comprehensive method for modeling uncertainties of solar irradiance for PV power generation in smart grids | |
| CN113804997A (en) | Voltage sag source positioning method based on bidirectional WaveNet deep learning | |
| CN113422371A (en) | Distributed power supply local voltage control method based on graph convolution neural network | |
| Zhou et al. | Short-term wind power prediction based on EMD-LSTM | |
| CN112836876A (en) | Power distribution network line load prediction method based on deep learning | |
| CN111488974A (en) | Deep learning neural network-based ocean wind energy downscaling method | |
| CN116470486A (en) | Ultra-short-term photovoltaic power generation power prediction method, device and storage medium | |
| CN115660893A (en) | Transformer substation bus load prediction method based on load characteristics | |
| CN115115139A (en) | Real-time prediction method for wind power | |
| CN114881338A (en) | Power distribution network line loss prediction method based on maximum mutual information coefficient and deep learning | |
| CN113515890A (en) | Renewable energy day-ahead scene generation method based on federal learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |