CN111242355A - Photovoltaic probability prediction method and system based on Bayesian neural network - Google Patents
Photovoltaic probability prediction method and system based on Bayesian neural network Download PDFInfo
- Publication number
- CN111242355A CN111242355A CN202010008652.5A CN202010008652A CN111242355A CN 111242355 A CN111242355 A CN 111242355A CN 202010008652 A CN202010008652 A CN 202010008652A CN 111242355 A CN111242355 A CN 111242355A
- Authority
- CN
- China
- Prior art keywords
- data
- weather forecast
- point
- photovoltaic
- dimensional space
- 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 67
- 238000013531 bayesian neural network Methods 0.000 title claims abstract description 65
- 238000009826 distribution Methods 0.000 claims abstract description 66
- 238000012545 processing Methods 0.000 claims abstract description 39
- 230000009467 reduction Effects 0.000 claims abstract description 37
- 238000013528 artificial neural network Methods 0.000 claims description 30
- 238000012549 training Methods 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 8
- 238000012614 Monte-Carlo sampling Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- 238000003062 neural network model Methods 0.000 claims description 4
- 238000010219 correlation analysis Methods 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000011946 reduction process Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 10
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000004590 computer program Methods 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 5
- 238000011176 pooling Methods 0.000 description 5
- 230000007613 environmental effect Effects 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 3
- 238000010248 power generation Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005204 segregation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Data Mining & Analysis (AREA)
- Human Resources & Organizations (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Primary Health Care (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a photovoltaic probability prediction method and a system based on a Bayesian neural network, which comprise the following steps: acquiring weather forecast data of a point to be predicted and historical output data of photovoltaic equipment; performing dimension reduction processing on the weather forecast data, and obtaining characteristic data based on the weather forecast data subjected to the dimension reduction processing and historical output data of the photovoltaic equipment; and substituting the characteristic data into a pre-constructed improved Bayesian neural network model to obtain the photovoltaic output distribution of the points to be predicted. The photovoltaic output distribution of the points to be predicted is obtained, and compared with a deterministic prediction mode, the photovoltaic probability prediction method provided by the invention has smaller average interval width when the same prediction accuracy is achieved, the prediction precision is improved, and the method has important significance for improving the safety and stability of a power grid.
Description
Technical Field
The invention relates to the field of new energy power prediction, in particular to a photovoltaic probability prediction method and system based on a Bayesian neural network.
Background
In recent years, with the arrival of energy crisis, the utilization rate of new energy is increased year by year, taking the china of 2018 as an example, the installed capacity of distributed photovoltaic is increased by 2096 ten thousand kilowatts more than 2017, and the installed capacity of distributed photovoltaic is increased by up to 71% on a year-by-year basis. The distributed power supply output is highly dependent on the external weather condition, so that the overall randomness is high, and the time-space correlation between the power supply and the environment is also high. Compared with the traditional power supply, the distributed power supply is more easily influenced by factors inside and outside the system, and the change characteristics are difficult to be directly reflected simply through a plurality of characteristics. The direct influence factors of the photovoltaic power generation power include the intensity of the solar radiation outside the ground, the air quality, the thickness and the height of cloud layers, the installation angle of the photoelectric panel, the photoelectric conversion efficiency of the photoelectric panel and the like, and the indirect factors further include various environmental factors such as wind power, temperature, humidity, rainfall and the like. It is urgently needed to extract more refined characteristics according to the self characteristics of the photovoltaic distributed power supply and improve the prediction efficiency and precision.
At present, numerical weather forecast and historical photovoltaic power generation data are mostly used as supports for power prediction of distributed photovoltaic power supplies, and a deterministic power value is used as a prediction result. The prediction mode can provide certain reference for the scheduling system to calculate the day-ahead scheduling scheme, but the scheduling system still cannot judge the possible fluctuation condition of the photovoltaic power supply in a short time scale. When the load of the power grid is large and the weather is suddenly changed, scheduling according to a deterministic prediction result may cause large difference between power output and expected power output, which causes problems of frequency out-of-limit, voltage out-of-limit and the like in the power distribution network. And the photovoltaic probability prediction can provide uncertainty information of a prediction point while predicting the output value of the photovoltaic power supply, and provide reference for possible fluctuation conditions of the photovoltaic power supply, so that the scheduling system is guided to deal with possible photovoltaic output fluctuation in advance. Therefore, how to research the photovoltaic output probability prediction method according to the characteristics of the distributed photovoltaic power supply in the power distribution network becomes an important premise for ensuring real-time power balance of the power grid and fully absorbing new energy.
Disclosure of Invention
Aiming at the defects of the existing photovoltaic power prediction technology, the invention aims to provide a photovoltaic probability prediction method based on a Bayesian neural network, and aims to consider the characteristics of different environmental factors and perform adaptive improvement on the structure of the Bayesian neural network according to the data types of the environmental factors, so that a model can finally accurately predict the interval of photovoltaic power and guide a scheduling system to deal with possible photovoltaic output fluctuation in advance.
The invention provides a photovoltaic probability prediction method based on a Bayesian neural network, which comprises the following steps:
acquiring weather forecast data of a point to be predicted and historical output data of photovoltaic equipment;
performing dimension reduction processing on the weather forecast data, and obtaining characteristic data based on the weather forecast data subjected to the dimension reduction processing and historical output data of the photovoltaic equipment;
and substituting the characteristic data into a pre-constructed improved Bayesian neural network model to obtain the photovoltaic output distribution of the points to be predicted.
Preferably, the constructing of the bayesian neural network model comprises:
collecting weather forecast data and historical output data of photovoltaic equipment at the same time according to a time sequence;
performing correlation analysis on the weather forecast data and the historical output data of the photovoltaic equipment to obtain input characteristics with low correlation with photovoltaic output;
carrying out dimension reduction processing on the input features with low photovoltaic output correlation;
obtaining characteristic data based on the weather forecast data after the dimension reduction processing and the historical output data of the photovoltaic equipment;
designing the input end of the Bayesian neural network into a full-connection neural network module, a one-dimensional convolution neural network module and a direct input module;
inputting the data type of the characteristic data into the input end of the Bayes neural network, training the Bayes neural network by adopting a variational inference method, and obtaining a Bayes neural network model capable of outputting photovoltaic output probability by learning the random distribution characteristics of photovoltaic output under different inputs.
Preferably, the inputting the data type according to the feature data into the input end of the bayesian neural network includes:
inputting the weather data subjected to dimensionality reduction into a full-connection neural network module;
inputting the output data closest to the prediction point in the historical output data into a direct input module;
and inputting the rest historical output data into a one-dimensional convolution neural network module.
Preferably, the training of the bayesian neural network by using the variational inference method, and the learning of the random distribution characteristics of the photovoltaic output under different inputs, to obtain the bayesian neural network model capable of outputting the photovoltaic output probability, include:
taking the characteristic data as an input characteristic of a Bayesian neural network, and taking an actual photovoltaic output value of the predicted point as a label;
training a Bayes neural network by taking the lower evidence bound as a loss function to obtain a relation between input characteristics and photovoltaic output;
and constructing a Bayesian neural network model based on the relation between the input features and the photovoltaic output.
Preferably, the evidence is as follows:
L(X,Y,q)＝Eq(W)[logp(Y|X,W)+logp(W)-logq(W)]
in the formula: l (X, Y, q) is the lower bound of evidence, X is the input data of the probability layer, Y is the output data of the probability layer, W is the weight, p (W) is the prior distribution of weights, q (W) is the introduced distribution, Eq(W)For the mathematical expectation of q (W), p (Y | X, W) is the weight of the output data in the determined probability layer and the input data stripThe probability distribution is represented by logit posterior probability logit (Y | X, W) and logit prior probability logit (p) (W).
Preferably, the bringing the feature data into a pre-constructed improved bayesian neural network model to obtain the photovoltaic output distribution of the point to be predicted includes:
repeatedly bringing the characteristic data into a pre-constructed improved Bayesian neural network model to obtain a plurality of groups of photovoltaic probability prediction results;
and processing the plurality of groups of prediction results by adopting a Monte Carlo sampling method to obtain the photovoltaic output distribution of the points to be predicted.
Preferably, the dimension reduction process includes:
respectively acquiring conditional probability between two points of weather forecast data in a high-dimensional space and conditional probability between two points of weather forecast data in a low-dimensional space;
obtaining the total difference of the whole weather forecast data set in the high-dimensional space and the low-dimensional space based on the conditional probability between the two weather forecast data in the high-dimensional space and the conditional probability between the two weather forecast data in the low-dimensional space;
the total difference of the whole weather forecast data set in the high-dimensional space and the low-dimensional space is minimized by adopting a gradient descent method.
Preferably, the conditional probability between two weather forecast data points in the high-dimensional space is calculated according to the following formula:
in the formula: p is a radical ofijIs the mean value of the conditional probability between the i-point weather forecast data and the j-point weather forecast data in the high-dimensional space, pi|jIs the conditional probability, p, between the i-point weather forecast data and the j-point weather forecast data in the high dimensional spacej|iThe conditional probability between j-point weather forecast data and i-point weather forecast data in a high-dimensional space is obtained;
wherein the conditional probability p between the i point and the j point in the high dimensional spacei|jCalculated as follows:
in the formula: x is the number ofiIs the ith weather forecast data point, x, in a high dimensional spacejIs the jth weather forecast data point, sigma, in a high dimensional spaceiIs the standard deviation, x, of the weather forecast data set with point i as the center pointkThe kth weather forecast data point in high dimensional space.
Preferably, the conditional probability between two weather forecast data points in the low-dimensional space is calculated according to the following formula:
in the formula: q. q.sijIs the conditional probability, y, between the weather forecast data at point i and the weather forecast data at point j in the low dimensional spaceiIs the ith weather forecast data point, y, in a low dimensional spacejIs the jth weather forecast data point, y, in a low dimensional spacekThe kth weather forecast data point in the low dimensional space.
Preferably, the total difference between the whole weather forecast data set in the high-dimensional space and the low-dimensional space is calculated according to the following formula:
in the formula: c is the total difference between the high-dimensional space and the low-dimensional space of the whole weather forecast data set, DKLIs the distribution difference of the i-point weather forecast data in a high-dimensional space and a low-dimensional space, PiDistribution of i-point weather forecast data in a low-dimensional space, QiDistribution of i-point weather forecast data in a low-dimensional space, pijIs the conditional probability between the weather forecast data of point i and the weather forecast data of point j in the high dimensional space, qijThe conditional probability between the weather forecast data of the point i and the weather forecast data of the point j in the low-dimensional space is shown.
Preferably, the weather forecast data includes: the temperature, the humidity, the atmospheric pressure, the wind speed, the sky cloud volume ratio, the haze index, the visibility and the weather type of the area where the photovoltaic equipment is located.
Based on the same inventive concept, the invention also provides a photovoltaic probability prediction system based on the Bayesian neural network, which comprises the following steps:
the acquiring module is used for acquiring weather forecast data of a point to be predicted and historical output data of the photovoltaic equipment;
the dimension reduction processing module is used for carrying out dimension reduction processing on the weather forecast data and obtaining characteristic data based on the weather forecast data subjected to the dimension reduction processing and historical output data of the photovoltaic equipment;
and the prediction module is used for substituting the characteristic data into a pre-constructed improved Bayesian neural network model to obtain the photovoltaic output distribution of the point to be predicted.
Preferably, the prediction module includes:
the prediction unit is used for substituting the characteristic data into a pre-constructed improved Bayesian neural network model for multiple times to obtain multiple groups of photovoltaic probability prediction results;
and the sampling processing unit is used for processing the multiple groups of prediction results by adopting a Monte Carlo sampling method to obtain the photovoltaic output distribution of the points to be predicted.
Compared with the closest prior art, the technical scheme provided by the invention has the following beneficial effects:
(1) according to the technical scheme provided by the invention, weather forecast data of a point to be predicted and historical output data of photovoltaic equipment are obtained; performing dimension reduction processing on the weather forecast data, and obtaining characteristic data based on the weather forecast data subjected to the dimension reduction processing and historical output data of the photovoltaic equipment; and substituting the characteristic data into a pre-constructed improved Bayesian neural network model to obtain the photovoltaic output distribution of the points to be predicted. The photovoltaic probability prediction method provided by the invention has the advantages that the photovoltaic probability prediction is realized through the technical scheme provided by the invention, the photovoltaic output distribution of the point to be predicted is obtained, compared with the existing deterministic prediction mode, the photovoltaic probability prediction method provided by the invention has smaller average interval width when the same prediction accuracy is reached, the prediction precision is improved, and the method has important significance for improving the safety and stability of a power grid.
(2) According to the technical scheme provided by the invention, the input features with low correlation are processed by adopting a dimension reduction method, the data density is effectively improved, and the overfitting of the model is reduced.
(3) According to the technical scheme provided by the invention, the input end of the Bayesian neural network is designed into a full-connection neural network module, a one-dimensional convolution neural network module and a direct input module according to the data type of the input characteristics, so that the information extraction capability of the network on different input characteristics is improved.
Drawings
FIG. 1 is a flow chart of a photovoltaic probability prediction method based on a Bayesian neural network in the invention;
fig. 2 is a schematic diagram of a specific photovoltaic probability prediction process based on a bayesian neural network in this embodiment;
FIG. 3 is a schematic diagram of a Bayesian neural network with improved input terminals;
FIG. 4 is a comparison graph of the prediction results when the method of the present invention and the conventional point prediction algorithm are used to predict photovoltaic output under sunny conditions in the example;
FIG. 5 is a comparison graph of the prediction results when the method of the present invention and the conventional point prediction algorithm are used to predict photovoltaic output under cloudy conditions in the examples;
FIG. 6 is a comparison graph of the prediction results when the method of the present invention and the conventional point prediction algorithm are used to predict photovoltaic output under rainy conditions in the example;
FIG. 7 is a graph of photovoltaic output interval results obtained when photovoltaic output fluctuates by the method of the present invention in the examples.
Detailed Description
For a better understanding of the present invention, reference is made to the following description taken in conjunction with the accompanying drawings and examples.
Example 1
As shown in fig. 1, the photovoltaic probability prediction method based on the bayesian neural network provided by the present invention includes:
s1, acquiring weather forecast data of a point to be predicted and historical output data of the photovoltaic equipment;
s2, performing dimensionality reduction on the weather forecast data, and obtaining characteristic data based on the weather forecast data subjected to dimensionality reduction and historical output data of the photovoltaic equipment;
and S3, substituting the characteristic data into a pre-constructed improved Bayes neural network model to obtain the photovoltaic output distribution of the points to be predicted.
The photovoltaic probability prediction method provided by the invention needs to construct an improved Bayesian neural network model, the model can be constructed and trained in advance for the same photovoltaic equipment, the trained model is directly used for prediction after the training is finished, and if the improved Bayesian neural network model is obtained, the following specific introduction is that:
step 1, collecting historical data of a photovoltaic system, collecting weather forecast data of a weather station in the same period, and performing dimension reduction processing on the data;
step 2, constructing a Bayesian neural network optimized for the input feature types;
step 3, training a Bayesian neural network: taking the predicted point time, historical data and dimension-reduced weather data as input features, taking the actual photovoltaic output value of the predicted point as a label, and taking the lower bound of the evidence as a loss function to train the Bayesian neural network;
further, the step 1 specifically includes:
firstly, collecting historical data and environmental data of a photovoltaic power station: the historical data is 30 output points before the predicted point, and the numerical weather forecast is from a weather station of a city where the photovoltaic equipment is located.
Next, the collected data is normalized by the formula:
in the formula: u is the data set to be normalized, μ is the mean of the data set U, and σ is the standard deviation of the data set U.
Thirdly, the correlation between the input data and the photovoltaic output is quantitatively analyzed by adopting a Pearson correlation coefficient, the input data set is A, and the element is aiThe photovoltaic output data set is B, wherein the element is BiThen, the correlation r between the two data sets is calculated as:
in the formula:
and
mean values for data sets a and B, respectively.
Based on the correlation in the previous step, a plurality of input features with low correlation are subjected to dimensionality reduction by using a t-distributed stored neighboring Embedding (t-SNE) algorithm, and the t-SNE algorithm assumes that corresponding data points in a high-dimensional space and a low-dimensional space respectively obey Gaussian distribution and t distribution. Let xiFor the ith data point in the high-dimensional space, the conditional probability between the i point and the j point in the high-dimensional space is:
in the formula: sigmaiIs the standard deviation of the data set with the i point as the center point.
Due to pi|j≠pj|iThe conditional probability p between the i point and the j point is usually described by the mean of two conditional probabilitiesijTo reduce the calculation error:
let yiFor the ith data point in the low dimensional space, the conditional probability q between the i point and the j point in the low dimensional spaceijComprises the following steps:
introducing KL divergence to represent the distribution difference of the same point in the high-dimensional space and the low-dimensional space, and if the distribution of the i point in the data set in the high-dimensional space is PiDistribution in a low dimensional space is QiThen the total difference C of the entire dataset in the high-dimensional space and the low-dimensional space can be expressed as:
the t-SNE minimizes the above formula by using a gradient descent method, i.e., minimizes KL divergence among all data points, so that points closer in the high-dimensional space are still closer in the low-dimensional space and points farther away in the high-dimensional space are still farther in the low-dimensional space. The invention adopts the algorithm to carry out dimension reduction on the input data with lower correlation, thereby obviously improving the data value of numerical weather forecast, reducing the dimension of the input data, increasing the data density and accelerating the training speed of the model.
And finally, directly converting the season, date and other time-related information of the predicted point into the maximum output power of the photovoltaic equipment of the predicted point by adopting a neural network.
Further, the step 2 specifically includes:
firstly, numerical weather forecast in input characteristics comprises information such as temperature and humidity, more hidden variables are contained in the information, and the nonlinear relation between the information and photovoltaic output is strong. Therefore, the invention embeds the independent full-connection neural network at the input end of the Bayesian neural network as a processing module of the data. The hidden layer unit of the fully-connected neural network has a nonlinear activation function F for describing the nonlinear relation between data, and when the modified linear unit is adopted, the following steps are carried out:
F(x)＝max(0,x) (7)
at this time, the output Y of the i-th layeriComprises the following steps:
Yi＝F(Wi·Xi+Bi) (8)
in the formula: represents the dot product; wiIs the weight matrix of the ith layer; b isiA bias matrix for the ith layer; xiIs the input value of the ith layer.
Secondly, the photovoltaic historical output data and the weather data in the input features have obvious difference in data characteristics, and the historical output data shows stronger chronology. Therefore, the invention embeds an independent one-dimensional convolution neural network as a processing module at the input end of the Bayesian neural network.
The ith convolutional layer of the one-dimensional convolutional neural network is inputted with time series data
Then, after convolution by convolution kernel, the data Y is outputiThe t-th element of (1)
Is calculated by
In the formula: represents the dot product; wiConvolution kernel weight for the ith layer; b isiA cell bias matrix for the ith layer; f (x) is the same as formula (7).
Generally, the length of convolution kernel is 2k +1, where k is any positive integer, and
corresponding to
Given by:
in order to obtain effective time sequence information distributed in data, so that time sequence data is converted into variables with remarkable characteristics, the calculation time of a neural network is reduced, and the result after convolution needs to be pooled. The invention adopts an average pooling method, if the core of the ith pooling layer is N, the output value of the pooling unit in the pooling layer can be expressed as
In the formula: x is the number ofi,jFor inputting data X in the pooling unitiThe internal jth element.
And finally, because the correlation between the photovoltaic output data closest to the prediction point in the input features and the prediction point is strongest, the photovoltaic output data is directly input into a probability layer of the Bayesian neural network.
Further, the step 3 specifically includes:
first, because the bayesian neural network has a probability layer, its training mode is different from the traditional deep neural network. Let W be the weight set of the probability layer, X represent the input data of the probability layer, Y represent the output data of the probability layer, and X and Y are both known quantities during training. According to bayes' theorem, the posterior probability distribution of weights can be determined by:
in the formula, P (wix, Y) is a posterior distribution of probability layer weights, P (W) is a prior distribution of weights, P (Y | X, W) is a probability distribution of output data under the condition of determining probability layer weights and input data, and P (Y | X) is a probability distribution of input data under the condition of determining input data.
Secondly, since the weight W of the probability layer is a high-dimensional random variable, the calculation of p (Y | X, W) is very difficult, and thus it is difficult to calculate the posterior distribution of the probability layer weights directly by the bayesian theorem. A variation reasoning method is adopted to approximate the posterior distribution of the network weight, a simple distribution q (W) is introduced to approximate the distribution p (W | X, Y), the difference between the two distributions is measured by KL divergence (Kullback-Leibler divergence), and when q (W) is known, the weight distribution can be optimized by minimizing the KL divergence. KL divergence is difficult to calculate directly, and in the variation and segregation estimation, an evidenceCelower bound (ELBO) L is introduced, and is defined as:
L(X,Y,q)＝logp(Y|X)-DKL[q(W)||p(W)](13)
in the formula: dKL[q(W)||p(W)]Representing the KL divergence between distribution q (W) and distribution p (W).
The expression of KL divergence is expanded according to the definition, ELBO can be further simplified:
L(X,Y,q)＝Eq(W)[logp(Y|X,W)+logp(W)-logq(W)](14)
in the formula (14), the first term of the expected value to be obtained is the posterior log probability, and the likelihood can be obtained by approximating the conditional likelihood to the small batch sampling in the total sample; the second term is the prior probability of p (w), which can be obtained from the initialized parameters, and is generally assumed to be a standard gaussian distribution; the third term is obtained according to the specific distribution of the assumed q (W), by using mean-field method, i.e. assuming that all q (W) follow a Gaussian distribution with mean value 0 and are independent of each other, the distribution q (W) isi) Has a variance of σiIt can be expected by taking the average after taking the logarithm of each distribution. Therefore, the lower bound of evidence is converted into a form convenient for calculation, and can be optimized by a method for calculating a gradient. Because q (W) is assumed to obey Gaussian distribution with mean value of 0 and being independent of each other, the variance of the Gaussian distribution is only needed to be updated in the training process of the probability layer in the Bayesian neural network, and the variance sigma of the ith probability weight unit isiThe update formula is as follows:
σi′＝σi-f[L(xi,yi,q),σi](15)
in the formula: f is an iterative factor calculation formula and needs to be given by a specific optimization algorithm.
And when the iteration is carried out for enough times, the parameters in the Bayesian neural network tend to be stable, namely the network training is completed.
According to the technical scheme provided by the invention, the weight of the neural network weight is assumed to be probability distribution, so that the capability of the neural network for learning the random characteristics of data is improved, and the photovoltaic output probability prediction is effectively realized.
How to predict by using the trained model comprises the following steps:
and (3) acquiring input data of a point to be predicted, wherein the input data comprises photovoltaic equipment output data before the point to be predicted and weather forecast data of the point to be predicted, and the input data is preprocessed in the step 1 and then used as input characteristics of the Bayesian neural network.
Due to the existence of the probability layer in the Bayesian neural network, when the same group of input features are input into the model, the output photovoltaic output prediction result has randomness.
According to the method, a Monte Carlo sampling method is adopted to predict the predicted point for multiple times, so that the photovoltaic output distribution range of the predicted point is obtained, namely the output result of photovoltaic probability prediction.
Compared with the traditional deterministic prediction method, the photovoltaic probability prediction method based on the Bayesian neural network aims at the following points: (1) the correlation between the input characteristic data and the photovoltaic output is comprehensively analyzed, and the dimension reduction is performed on the high-dimensional input data by adopting various methods, so that the data density is improved, the training difficulty is reduced, and the prediction precision is optimized; (2) the structure of the Bayesian neural network is improved, so that the Bayesian neural network can adapt to input characteristics of different data types, and the generalization of a photovoltaic probability prediction method is improved; (3) the problem that photovoltaic output fluctuation is difficult to describe in a traditional deterministic prediction mode is solved, and a photovoltaic output probability prediction result is completely given by describing possible output and distribution of photovoltaic equipment at a prediction point.
The present embodiment further describes the present invention in detail with reference to the accompanying drawings, as shown in fig. 2, the present invention provides a photovoltaic probability prediction method based on a bayesian neural network on the basis of a photovoltaic prediction method, and specifically includes the following steps:
step 1: historical data of the photovoltaic equipment and weather data of a weather station nearby the photovoltaic equipment are collected and subjected to dimension reduction.
The photovoltaic power plant data of gathering includes: the active power output by distributed photovoltaic power generation equipment in a certain park. The collected weather data includes: the temperature, humidity, atmospheric pressure, wind speed, sky cloud content ratio, haze index, visibility and weather type of the area where the equipment is located.
The method is characterized in that the historical output data of the photovoltaic equipment has high positive correlation, dimension reduction is not needed, the correlation between weather data such as temperature, humidity and the like and the photovoltaic output data is not obvious, hidden information in the data needs to be further extracted through dimension reduction, dimension reduction is carried out on the data by adopting a t-distribution neighborhood embedding method, a photovoltaic output coefficient is introduced to visually display a dimension reduction result, and the photovoltaic output coefficient η is calculated according to the following formula:
η＝P/Pmax(1)
in the formula: p is the actual output of the photovoltaic equipment; pmaxThe theoretical maximum output of the photovoltaic device at the moment is shown.
After the weather data is reduced to two dimensions, different data points show obvious aggregative property according to the photovoltaic output coefficient.
Step 2: the input data includes data of different characteristics such as weather data, time data, historical data and the like, and the input end of the bayesian neural network is divided into a fully connected neural network part, a one-dimensional convolutional neural network part and a direct input part according to the characteristics of the data, as shown in fig. 3.
And step 3: after the processing, a plurality of effective data for training the Bayesian neural network can be obtained, and the effective data is trained by adopting a variational reasoning method until the network parameters are converged in a stable range.
And 4, step 4: and giving input data of the predicted point, and realizing multi-time sampling of the predicted point. Under the sunny condition with smooth photovoltaic output, a comparison graph of the prediction result of the method and the traditional point prediction is shown in fig. 4; the results of the cloud and rainy days with severe photovoltaic output fluctuation are shown in fig. 5 and 6, respectively. A comparison graph of the prediction result interval of the method and the point prediction result adopting the deep neural network is shown in FIG. 7. The method has a good covering effect on the actual output value, can provide a possible range of fluctuation of the photovoltaic output, and can not respond to the short-time fluctuation of the photovoltaic output by deterministic prediction.
Compared with the classical normal distribution method, the prediction result obtained by the method provided by the invention has smaller average interval width when the same prediction accuracy is achieved, the prediction precision is improved, and the method has important significance for improving the safety and stability of the power grid.
Example 2
Based on the same inventive concept, the invention also provides a photovoltaic probability prediction system based on the Bayesian neural network, which comprises the following steps:
the acquiring module is used for acquiring weather forecast data of a point to be predicted and historical output data of the photovoltaic equipment;
the dimension reduction processing module is used for carrying out dimension reduction processing on the weather forecast data and obtaining characteristic data based on the weather forecast data subjected to the dimension reduction processing and historical output data of the photovoltaic equipment;
and the prediction module is used for substituting the characteristic data into a pre-constructed improved Bayesian neural network model to obtain the photovoltaic output distribution of the point to be predicted.
In an embodiment, the prediction module comprises:
the prediction unit is used for substituting the characteristic data into a pre-constructed improved Bayesian neural network model for multiple times to obtain multiple groups of photovoltaic probability prediction results;
and the sampling processing unit is used for processing the multiple groups of prediction results by adopting a Monte Carlo sampling method to obtain the photovoltaic output distribution of the points to be predicted.
In an embodiment, the system further comprises a construction module for constructing the improved bayesian neural network model.
The building module is specifically configured to:
collecting weather forecast data and historical output data of photovoltaic equipment at the same time according to a time sequence;
performing correlation analysis on the weather forecast data and the historical output data of the photovoltaic equipment to obtain input characteristics with low correlation with photovoltaic output;
carrying out dimension reduction processing on the input features with low photovoltaic output correlation;
obtaining characteristic data based on the weather forecast data after the dimension reduction processing and the historical output data of the photovoltaic equipment;
designing the input end of the Bayesian neural network into a full-connection neural network module, a one-dimensional convolution neural network module and a direct input module;
inputting the data type of the characteristic data into the input end of the Bayes neural network, training the Bayes neural network by adopting a variational inference method, and obtaining a Bayes neural network model capable of outputting photovoltaic output probability by learning the random distribution characteristics of photovoltaic output under different inputs.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The present invention is not limited to the above embodiments, and any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention are included in the scope of the claims of the present invention which are filed as the application.
Claims (13)
1. A photovoltaic probability prediction method based on a Bayesian neural network is characterized by comprising the following steps:
acquiring weather forecast data of a point to be predicted and historical output data of photovoltaic equipment;
performing dimension reduction processing on the weather forecast data, and obtaining characteristic data based on the weather forecast data subjected to the dimension reduction processing and historical output data of the photovoltaic equipment;
and substituting the characteristic data into a pre-constructed improved Bayesian neural network model to obtain the photovoltaic output distribution of the points to be predicted.
2. The method of claim 1, wherein the bayesian neural network model is constructed by:
collecting weather forecast data and historical output data of photovoltaic equipment at the same time according to a time sequence;
performing correlation analysis on the weather forecast data and the historical output data of the photovoltaic equipment to obtain input characteristics with low correlation with photovoltaic output;
carrying out dimension reduction processing on the input features with low photovoltaic output correlation;
obtaining characteristic data based on the weather forecast data after the dimension reduction processing and the historical output data of the photovoltaic equipment;
designing the input end of the Bayesian neural network into a full-connection neural network module, a one-dimensional convolution neural network module and a direct input module;
inputting the data type of the characteristic data into the input end of the Bayes neural network, training the Bayes neural network by adopting a variational inference method, and obtaining a Bayes neural network model capable of outputting photovoltaic output probability by learning the random distribution characteristics of photovoltaic output under different inputs.
3. The method of claim 2, wherein inputting the data type per the feature data into an input of the bayesian neural network comprises:
inputting the weather data subjected to dimensionality reduction into a full-connection neural network module;
inputting the output data closest to the prediction point in the historical output data into a direct input module;
and inputting the rest historical output data into a one-dimensional convolution neural network module.
4. The method of claim 3, wherein the training of the Bayesian neural network by the variational inference method to obtain the Bayesian neural network model capable of outputting the photovoltaic output probability by learning the random distribution characteristics of the photovoltaic output at different inputs comprises:
taking the characteristic data as an input characteristic of a Bayesian neural network, and taking an actual photovoltaic output value of the predicted point as a label;
training a Bayes neural network by taking the lower evidence bound as a loss function to obtain a relation between input characteristics and photovoltaic output;
and constructing a Bayesian neural network model based on the relation between the input features and the photovoltaic output.
5. The method of claim 4, wherein the lower bound of evidence is represented by the following formula:
L(X,Y,q(W))＝Eq(W)[logp(Y|X,W)+logp(W)-logq(W)]
in the formula: l (X, Y, q) is the lower bound of evidence, X is the input data of the probability layer, Y is the output data of the probability layer, W is the weight, p (W) is the prior distribution of weights, q (W) is the introduced distribution, Eq(W)For the mathematical expectation of q (W), p (Y | X, W) is the probability distribution of the output data under the condition of determining the probability layer weights and the input data, logp (Y | X, W) is the posterior log probability, and logp (W) is the prior log probability of p (W).
6. The method of claim 1, wherein said introducing said feature data into a pre-constructed modified bayesian neural network model to obtain a photovoltaic output distribution of points to be predicted comprises:
repeatedly bringing the characteristic data into a pre-constructed improved Bayesian neural network model to obtain a plurality of groups of photovoltaic probability prediction results;
and processing the plurality of groups of prediction results by adopting a Monte Carlo sampling method to obtain the photovoltaic output distribution of the points to be predicted.
7. The method of any of claims 1 or 2, wherein the dimension reduction process comprises:
respectively acquiring conditional probability between two points of weather forecast data in a high-dimensional space and conditional probability between two points of weather forecast data in a low-dimensional space;
obtaining the total difference of the whole weather forecast data set in the high-dimensional space and the low-dimensional space based on the conditional probability between the two weather forecast data in the high-dimensional space and the conditional probability between the two weather forecast data in the low-dimensional space;
the total difference of the whole weather forecast data set in the high-dimensional space and the low-dimensional space is minimized by adopting a gradient descent method.
8. The method of claim 7, wherein the conditional probability between two points of weather forecast data in the high dimensional space is calculated as follows:
in the formula: p is a radical ofijIs the mean value of the conditional probability between the i-point weather forecast data and the j-point weather forecast data in the high-dimensional space, pi|jIs the conditional probability, p, between the i-point weather forecast data and the j-point weather forecast data in the high dimensional spacej|iThe conditional probability between j-point weather forecast data and i-point weather forecast data in a high-dimensional space is obtained;
wherein the conditional probability p between the i point and the j point in the high dimensional spacei|jCalculated as follows:
in the formula: x is the number ofiIs the ith weather forecast data point, x, in a high dimensional spacejIs the jth weather forecast data point, sigma, in a high dimensional spaceiIs the standard deviation, x, of the weather forecast data set with point i as the center pointkThe kth weather forecast data point in high dimensional space.
9. The method of claim 7, wherein the conditional probability between two points of weather forecast data in the low dimensional space is calculated as follows:
in the formula: q. q.sijIs the conditional probability, y, between the weather forecast data at point i and the weather forecast data at point j in the low dimensional spaceiIs the ith weather forecast data point, y, in a low dimensional spacejIs the jth weather forecast data point, y, in a low dimensional spacekThe kth weather forecast data point in the low dimensional space.
10. The method of claim 7, wherein the total difference between the entire set of weather forecast data in the high-dimensional space and the low-dimensional space is calculated as follows:
in the formula: c is the total difference between the high-dimensional space and the low-dimensional space of the whole weather forecast data set, DKLIs the distribution difference of the i-point weather forecast data in a high-dimensional space and a low-dimensional space, PiDistribution of i-point weather forecast data in a low-dimensional space, QiDistribution of i-point weather forecast data in a low-dimensional space, pijIs the conditional probability between the weather forecast data of point i and the weather forecast data of point j in the high dimensional space, qijThe conditional probability between the weather forecast data of the point i and the weather forecast data of the point j in the low-dimensional space is shown.
11. The method of claim 1, wherein the weather forecast data comprises: the temperature, the humidity, the atmospheric pressure, the wind speed, the sky cloud volume ratio, the haze index, the visibility and the weather type of the area where the photovoltaic equipment is located.
12. A photovoltaic probability prediction system based on a Bayesian neural network is characterized by comprising:
the acquiring module is used for acquiring weather forecast data of a point to be predicted and historical output data of the photovoltaic equipment;
the dimension reduction processing module is used for carrying out dimension reduction processing on the weather forecast data and obtaining characteristic data based on the weather forecast data subjected to the dimension reduction processing and historical output data of the photovoltaic equipment;
and the prediction module is used for substituting the characteristic data into a pre-constructed improved Bayesian neural network model to obtain the photovoltaic output distribution of the point to be predicted.
13. The system of claim 12, wherein the prediction module comprises:
the prediction unit is used for substituting the characteristic data into a pre-constructed improved Bayesian neural network model for multiple times to obtain multiple groups of photovoltaic probability prediction results;
and the sampling processing unit is used for processing the multiple groups of prediction results by adopting a Monte Carlo sampling method to obtain the photovoltaic output distribution of the points to be predicted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010008652.5A CN111242355A (en) | 2020-01-06 | 2020-01-06 | Photovoltaic probability prediction method and system based on Bayesian neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010008652.5A CN111242355A (en) | 2020-01-06 | 2020-01-06 | Photovoltaic probability prediction method and system based on Bayesian neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111242355A true CN111242355A (en) | 2020-06-05 |
Family
ID=70874324
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010008652.5A Pending CN111242355A (en) | 2020-01-06 | 2020-01-06 | Photovoltaic probability prediction method and system based on Bayesian neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111242355A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112052996A (en) * | 2020-09-03 | 2020-12-08 | 哈尔滨工业大学 | Small hydropower station cluster power prediction method based on convolutional neural network technology |
| CN113326658A (en) * | 2021-06-03 | 2021-08-31 | 中国南方电网有限责任公司 | Photovoltaic energy storage grid-connected control method based on neural network |
| CN113722375A (en) * | 2021-08-10 | 2021-11-30 | 国网浙江省电力有限公司电力科学研究院 | Double-layer hybrid photovoltaic prediction method based on multi-algorithm optimization and computer equipment |
| CN114358449A (en) * | 2022-03-21 | 2022-04-15 | 国网天津市电力公司电力科学研究院 | Electric vehicle charging load space-time distribution prediction method based on graph neural network |
| US11650351B2 (en) * | 2020-02-12 | 2023-05-16 | Nec Corporation | Semi-supervised deep model for turbulence forecasting |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104268638A (en) * | 2014-09-11 | 2015-01-07 | 广州市香港科大霍英东研究院 | Photovoltaic power generation system power predicting method of elman-based neural network |
| CN106779175A (en) * | 2016-11-28 | 2017-05-31 | 西安交通大学 | It is a kind of to consider that non-linear photovoltaic at times is exerted oneself Forecasting Methodology |
| CN108038518A (en) * | 2018-01-02 | 2018-05-15 | 华北电力大学（保定） | A kind of photovoltaic generation power based on meteorological data determines method and system |
| CN109902801A (en) * | 2019-01-22 | 2019-06-18 | 华中科技大学 | A kind of flood DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method based on variation reasoning Bayesian neural network |
-
2020
- 2020-01-06 CN CN202010008652.5A patent/CN111242355A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104268638A (en) * | 2014-09-11 | 2015-01-07 | 广州市香港科大霍英东研究院 | Photovoltaic power generation system power predicting method of elman-based neural network |
| CN106779175A (en) * | 2016-11-28 | 2017-05-31 | 西安交通大学 | It is a kind of to consider that non-linear photovoltaic at times is exerted oneself Forecasting Methodology |
| CN108038518A (en) * | 2018-01-02 | 2018-05-15 | 华北电力大学（保定） | A kind of photovoltaic generation power based on meteorological data determines method and system |
| CN109902801A (en) * | 2019-01-22 | 2019-06-18 | 华中科技大学 | A kind of flood DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM method based on variation reasoning Bayesian neural network |
Non-Patent Citations (1)
| Title |
|---|
| 赵康宁 等: "基于改进贝叶斯神经网络的光伏出力概率预测", 《电网技术》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11650351B2 (en) * | 2020-02-12 | 2023-05-16 | Nec Corporation | Semi-supervised deep model for turbulence forecasting |
| CN112052996A (en) * | 2020-09-03 | 2020-12-08 | 哈尔滨工业大学 | Small hydropower station cluster power prediction method based on convolutional neural network technology |
| CN112052996B (en) * | 2020-09-03 | 2022-02-25 | 哈尔滨工业大学 | Small hydropower station cluster power prediction method based on convolutional neural network technology |
| CN113326658A (en) * | 2021-06-03 | 2021-08-31 | 中国南方电网有限责任公司 | Photovoltaic energy storage grid-connected control method based on neural network |
| CN113326658B (en) * | 2021-06-03 | 2024-03-12 | 中国南方电网有限责任公司 | Photovoltaic energy storage grid-connected control method based on neural network |
| CN113722375A (en) * | 2021-08-10 | 2021-11-30 | 国网浙江省电力有限公司电力科学研究院 | Double-layer hybrid photovoltaic prediction method based on multi-algorithm optimization and computer equipment |
| CN113722375B (en) * | 2021-08-10 | 2023-10-17 | 国网浙江省电力有限公司电力科学研究院 | Double-layer hybrid photovoltaic prediction method based on multi-algorithm optimization and computer equipment |
| CN114358449A (en) * | 2022-03-21 | 2022-04-15 | 国网天津市电力公司电力科学研究院 | Electric vehicle charging load space-time distribution prediction method based on graph neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111242355A (en) | Photovoltaic probability prediction method and system based on Bayesian neural network | |
| CN111260126B (en) | Short-term photovoltaic power generation prediction method considering correlation degree of weather and meteorological factors | |
| CN107194495B (en) | Photovoltaic power longitudinal prediction method based on historical data mining | |
| CN114792156B (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
| CN112418346B (en) | Numerical weather forecast total radiation system error classification calculation method | |
| CN115115125B (en) | Photovoltaic power interval probability prediction method based on deep learning fusion model | |
| CN115796004A (en) | Photovoltaic power station ultra-short term power intelligent prediction method based on SLSTM and MLSTNet models | |
| CN113052469A (en) | Method for calculating wind-solar-water-load complementary characteristic of small hydropower area lacking measurement runoff | |
| CN113516271A (en) | Wind power cluster power day-ahead prediction method based on space-time neural network | |
| CN113837434A (en) | Solar photovoltaic power generation prediction method and device, electronic equipment and storage medium | |
| CN116341613A (en) | Ultra-short-term photovoltaic power prediction method based on Informar encoder and LSTM | |
| CN113111592A (en) | Short-term wind power prediction method based on EMD-LSTM | |
| CN114139783A (en) | Wind power short-term power prediction method and device based on nonlinear weighted combination | |
| CN116845875A (en) | WOA-BP-based short-term photovoltaic output prediction method and device | |
| CN116014722A (en) | Sub-solar photovoltaic power generation prediction method and system based on seasonal decomposition and convolution network | |
| CN116739172A (en) | Method and device for ultra-short-term prediction of offshore wind power based on climbing identification | |
| CN118095891A (en) | Active power distribution network payload prediction method and system considering source load meteorological characteristic decoupling | |
| CN117076738A (en) | Medium-long term weather prediction method, system and medium based on variable grid model | |
| CN116502074A (en) | Model fusion-based photovoltaic power generation power prediction method and system | |
| CN117134315A (en) | Distribution transformer load prediction method and device based on BERT algorithm | |
| CN116822331A (en) | Multi-energy fusion wind-solar power generation prediction method | |
| CN115456286A (en) | Short-term photovoltaic power prediction method | |
| CN113449466A (en) | Solar radiation prediction method and system for optimizing RELM based on PCA and chaos GWO | |
| CN118469106B (en) | Photovoltaic power prediction method considering historical similarity days and combination weights | |
| CN118100137A (en) | Multi-dimensional data coupling distributed photovoltaic output high-resolution prediction method and 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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200605 |