CN115758901A - EHO-GRU network-based short-term distributed photovoltaic output power prediction method - Google Patents
EHO-GRU network-based short-term distributed photovoltaic output power prediction method Download PDFInfo
- Publication number
- CN115758901A CN115758901A CN202211496034.5A CN202211496034A CN115758901A CN 115758901 A CN115758901 A CN 115758901A CN 202211496034 A CN202211496034 A CN 202211496034A CN 115758901 A CN115758901 A CN 115758901A
- Authority
- CN
- China
- Prior art keywords
- nomadic
- gru network
- elephant
- output power
- algorithm
- 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
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Photovoltaic Devices (AREA)
Abstract
The invention relates to the field of photovoltaic power prediction, in particular to a short-term distributed photovoltaic output power prediction method based on an EHO-GRU network. The method comprises the following steps of 1, establishing a historical information data set for predicting short-term distributed photovoltaic output power; step 2, aiming at the historical information data set established in the step 1, establishing a GRU network for predicting short-term distributed photovoltaic output power; and 3, aiming at the GRU network established in the step 2, establishing an EHO algorithm model for optimizing GRU network parameters. According to the method, the potential time sequence incidence relation between the photovoltaic output power and the natural environment information is mined by utilizing an update gate in the GRU network, and non-universal information of the photovoltaic output power and the natural environment information is eliminated so as to improve the generalization capability of the model and improve the effectiveness and universality of the prediction method. The accuracy of short-term distributed photovoltaic output power prediction can be improved by optimizing parameters of a prediction network based on a cluster nomadic algorithm.
Description
Technical Field
The invention relates to the field of photovoltaic power prediction, in particular to a short-term distributed photovoltaic output power prediction method based on an EHO-GRU network.
Background
The construction of a novel power system mainly based on new energy is an important support for achieving the goals of carbon peak reaching and carbon neutralization. Distributed photovoltaic power stations are increasingly incorporated into power grids as important components of new energy power generation units, and gradually become main power generation units in the power grids.
However, due to the influence of random natural environment factors, the output power of the distributed photovoltaic power generation has a certain uncertainty, which brings a great challenge to the stable operation of the power grid. Therefore, it is necessary to design a photovoltaic output power prediction method to provide reliable decision basis for the power dispatching department, and improve the operation stability of the power system.
At present, the photovoltaic power prediction method mainly comprises a statistical method and a machine learning method. Due to the fact that the photovoltaic output power and the multi-dimensional natural environment information have a nonlinear correlation relationship, a large prediction deviation generally exists in a traditional statistical method. The method based on machine learning can learn the incidence relation between the photovoltaic output power and the natural environment information through a training nonlinear algorithm network, and is gradually becoming the mainstream photovoltaic power prediction method. However, there are many network parameters that need to be set in advance in the machine learning algorithm. The selection result of the parameters directly influences the photovoltaic output power prediction accuracy based on machine learning. Therefore, the traditional photovoltaic power prediction method based on the machine learning algorithm still has certain randomness.
Disclosure of Invention
The invention aims to provide a short-term distributed photovoltaic output power prediction method based on an EHO-GRU network aiming at the defects, which is characterized in that a short-term distributed photovoltaic output power intelligent prediction model considering multi-dimensional environment influence factors and output power time sequence relevance is established based on a recurrent neural network (GRU), parameters in the GRU network are optimized by using an e-shaped swarm algorithm (EHO), and the influence of parameter selection randomness in the GRU network on the short-term distributed photovoltaic output power prediction precision is reduced.
The invention is realized by adopting the following technical scheme:
a short-term distributed photovoltaic output power prediction method based on an EHO-GRU network comprises the following steps:
step 1, establishing a historical information data set for predicting short-term distributed photovoltaic output power;
step 2, aiming at the historical information data set established in the step 1, establishing a GRU network for predicting short-term distributed photovoltaic output power;
and 3, aiming at the GRU network established in the step 2, establishing an EHO algorithm model for optimizing GRU network parameters.
Further, the specific steps of step 1 include:
s1-1, collecting historical information data;
the historical information data set in the step 1 comprises a historical environment information data set and a historical power grid operation information data set in the last year; the collection interval of the historical information data is 1 minute;
the historical environmental information data set comprises irradiance, temperature, wind direction, wind speed, humidity and air pressure; the historical power grid operation information data set comprises historical power generation power data and historical load power data;
s1-2, selecting characteristics of the historical information data set.
Further, the step S1-2 includes:
s1-2-1, taking photovoltaic historical generated power data as an output vector, and taking irradiance, temperature, wind direction, wind speed, humidity, air pressure and historical load as input vectors;
s1-2-2, converting each column of characteristic values in an input vector into a histogram by utilizing a histogram function in a wrapping type characteristic selection algorithm (LGBFS), sequencing the importance of the input vector by utilizing a T _ Split function and a T _ Gain function in the LGBFS, and eliminating the characteristics with the importance of less than 5% to form the optimized input vector.
Further, the step 2 comprises:
s2-1, taking the optimized input vector obtained in the step 1 as an input variable x of the GRU network a ；
S2-2, input variable x in step S2-1 a Training through a plurality of hidden layers in a GRU network in sequence, wherein the hidden layers comprise a reset gate function r at t moment t And the update gate function u at time t t ；
Wherein, the update gate function is used for reserving key time sequence related information of the input variable, specifically, the update gate function is used for reserving the key time sequence related information of the input variable
In the formula, g is a Sigmoid activation function;
is the input vector at the time t;
the output quantity of the hidden layer at the time t-1; w 1 For the dimension of the matrix to be l × c 1 Where l is the dimension of the hidden layer output vector, c 1 And an input variable x a Are the same in dimension; v 1 For the dimension of the matrix to be l × c 2 A weight matrix of (2), wherein c 2 Outputting a vector dimension for a previous hidden layer; w 1 ，V 1 The elements in the matrix are real numbers with random initial values respectively.
The reset gate function is used for removing unnecessary time sequence related information of the input variable, and the generalization capability of the GRU network can be improved; the reset gate function may be specifically expressed as
In the formula, W 2 Is the dimension of the matrix, is l × c 2 A weight matrix of (a); v 2 Is the dimension of the matrix, is l × c 2 A weight matrix of (a); w 2 ，V 2 The elements in the matrix are real numbers with random initial values respectively.
The state quantity of the current hidden layer is output by the last hidden layer
After passing through reset gate and input vector
The splicing composition can be expressed as
In the formula, H t The state quantity of the current hidden layer; tan h is the conversion of input to [ -1,1]An activation function within a range; w 3 Is that dimension of matrix is l x (c) 1 +c 2 ) The elements in the matrix are real numbers with random initial values. h is t-1 The output quantity of the hidden layer at time t-1.
The state quantity of the current hidden layer is the output quantity of the current hidden layer after passing through the update gate
In particular to
It can be seen that the current hidden layer output quantity
Continuously sending to the next hidden layer as input, and circularly reciprocating to form output p of GRU network a Namely the predicted photovoltaic power generation power.
Further, said step 3 uses elephant group nomadic algorithm (EHO) to train matrix W of GRU network in step 2 1 ，V 1 ，W 2 ，V 2 And W 3 Optimizing the parameters;
the method comprises the following specific steps:
s3-1, initializing elephant group nomadic algorithm parameters; the elephant swarm nomadic algorithm parameters comprise the number n of elephants in the elephant swarm, the searching dimension of the elephant swarm is D, and the maximum iteration number is I max And an initial nomadic location (initial solution) x m ；
S3-2, calculating a nomadic position fitness function F of the nomadic position of the current turn elephant in the group nomadic algorithm, and evaluating the goodness and badness of the nomadic position; the nomadic location fitness function of the elephant herding and nomadic algorithm can be expressed as
In the formula, e m Is the predicted average percentage error; e.g. of the type r To predict the root mean square error; n is a radical of p Predicting the number of power points;
respectively representing the true value and the predicted value of the ith power point;
s3-3, searching a new nomadic position based on a neighborhood search algorithm; in particular to
v m,i ＝x m,i +α(x m,i -x gb )
In the formula, v m,i Obtaining a new nomadic position for the elephant i through a neighborhood search algorithm; x is the number of m,i Is the original nomadic position of elephant i; x is the number of gb The nomadic position with the highest fitness function in all the current nomadic positions is determined; alpha is [0,1]A random number in between;
s3-4, designing a probability selection formula to determine whether the elephant is nomadic to a position with a higher fitness function value of the nomadic position; the nomadic location selection probability for elephant i can be expressed as
In the formula (f) i Updating the probability for the nomadic location of elephant i; f i The function value of the suitability degree of the nomadic position of the elephant i is obtained;
s3-5, judging whether the search times reach the maximum iteration times I max (ii) a If yes, outputting the optimal solution of the current turn of the elephant herding algorithm, namely W 1 ，V 1 ，W 2 ，V 2 And W 3 The optimum parameter of (2); and if not, performing the nomadic position search of the next round, and returning to the step S3-2.
Compared with the prior art, the invention has the following beneficial effects:
1. the short-term distributed photovoltaic output power intelligent prediction method considering the time sequence relevance and the generalization of the multidimensional prediction influence factors is provided. Potential time sequence incidence relation between the photovoltaic output power and the natural environment information is mined by using an update gate in the GRU network, and non-universality information between the photovoltaic output power and the natural environment information is eliminated by using a reset gate in the GRU network so as to improve generalization capability of the model and improve effectiveness and universality of the prediction method.
2. A short-term distributed photovoltaic output power prediction network parameter optimization method based on a cluster nomadic algorithm is designed. The accuracy of short-term distributed photovoltaic output power prediction can be improved by optimizing parameters of a prediction network based on a cluster nomadic algorithm.
Drawings
FIG. 1 is a schematic flow chart of the steps of the method of the present invention;
fig. 2 is a schematic diagram of solving the optimal parameters of the GRU network according to the method of the present invention.
Detailed Description
The invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
Referring to fig. 1 and fig. 2, the short-term distributed photovoltaic output power prediction method based on the EHO-GRU network of the present invention includes the following steps:
step 1, establishing a historical information data set for predicting short-term distributed photovoltaic output power;
step 2, aiming at the historical information data set established in the step 1, establishing a GRU network for predicting short-term distributed photovoltaic output power;
and 3, aiming at the GRU network established in the step 2, establishing an EHO algorithm model for optimizing GRU network parameters.
The specific embodiment is as follows: and selecting the distributed photovoltaic data of the power grid in a certain area for case analysis.
The regional distributed photovoltaic comprises 1152 photovoltaic panels, and the rated output power of each photovoltaic panel is 253W.
The method is adopted to optimize the short-term distributed photovoltaic output power prediction network parameters, and specifically comprises the following steps:
step 1, establishing a historical information data set for predicting short-term distributed photovoltaic output power;
s1-1, collecting historical information data;
the obtained original data comprise photovoltaic output power, irradiance, temperature, wind direction, wind speed, humidity and air pressure which are sampled at 1-minute intervals within 1 month of 2019 to 2020.
S1-2, selecting characteristics of a historical information data set;
s1-2-1, taking photovoltaic historical generated power data as an output vector, and taking irradiance, temperature, wind direction, wind speed, humidity, air pressure and historical load as input vectors;
s1-2-2, converting each column of characteristic values in an input vector into a histogram by utilizing a histogram function in a wrapping type characteristic selection algorithm (LGBFS), sequencing the importance of the input vector by utilizing a T _ Split function and a T _ Gain function in the LGBFS, and eliminating the characteristics with the importance of less than 5% to form the optimized input vector.
Step 2, aiming at the historical information data set established in the step 1, establishing a GRU network for predicting short-term distributed photovoltaic output power;
s2-1, taking the optimized input vector obtained in the step 1 as an input variable x of the GRU network a ；
S2-2, training the input variable sequence in the step S2-1 through a plurality of hidden layers in a GRU network;
and 3, aiming at the GRU network established in the step 2, establishing an EHO algorithm model for optimizing GRU network parameters.
Referring to fig. 2, a matrix W to be trained of the GRU network in step 2 is mapped to a matrix W to be trained using an herring-grazing algorithm (EHO) 1 ，V 1 ，W 2 ，V 2 And W 3 Optimizing the parameters;
the method comprises the following specific steps:
s3-1, initializing elephant group nomadic algorithm parameters; the elephant trunk nomadic algorithm parameters comprise the number n of elephants in the elephant trunk, the searching dimension of the elephant trunk is D, and the maximum iteration number is I max And an initial nomadic location (initial solution) x m ；
S3-2, calculating a nomadic position fitness function F of the nomadic position of the current turn elephant in the group nomadic algorithm, and evaluating the goodness and badness of the nomadic position; the nomadic location fitness function of the elephant herding and nomadic algorithm can be expressed as
In the formula, e m Is the predicted average percentage error; e.g. of a cylinder r To predict the root mean square error; n is a radical of p Predicting the number of power points;
are respectively the ithThe real value and the predicted value of the power point;
s3-3, searching a new nomadic position based on a neighborhood search algorithm; in particular to
vm,i＝xm,i+α(xm,i-xgb)
In the formula, v m,i Obtaining a new nomadic position for the elephant i through a neighborhood search algorithm; x is the number of m,i Is the original nomadic position of elephant i; x is the number of gb The nomadic position with the highest fitness function in all the current nomadic positions is taken as the nomadic position; alpha is [0,1]A random number in between;
s3-4, designing a probability selection formula to determine whether the elephant is nomadic to a position with a higher fitness function value of the nomadic position; the nomadic location selection probability for elephant i can be expressed as
In the formula (f) i Updating the probability for the nomadic location of elephant i; f i The function value of the suitability degree of the nomadic position of the elephant i is obtained;
s3-5, judging whether the search times reach the maximum iteration times I max (ii) a If yes, outputting the optimal solution of the current turn of the elephant herding algorithm, namely W 1 ，V 1 ，W 2 ，V 2 And W 3 The optimum parameter of (2); and if not, performing the nomadic position search of the next round, and returning to the step S3-2.
Here, 4 photovoltaic power prediction methods are considered for comparison.
Method 1, the prediction method of the present invention;
a method 2, a particle swarm algorithm-based prediction method;
method 3, a prediction method based on a multilayer perceptron;
the method 4 comprises the following steps: a prediction method based on genetic algorithm.
The predicted mean percentage error and the predicted root mean square error for each method are shown in table 1, respectively.
TABLE 1 comparison of prediction errors for the methods
From the data in the table, it can be seen that the method of the present invention has the lowest predicted average percentage error and predicted root mean square error, because the method designed by the present invention optimizes the parameters in the intelligent prediction network by using the herring-grazing algorithm, so that the prediction performance of the GRU network can be fully exerted.
Claims (5)
1. A short-term distributed photovoltaic output power prediction method based on an EHO-GRU network is characterized by comprising the following steps:
step 1, establishing a historical information data set for predicting short-term distributed photovoltaic output power;
step 2, aiming at the historical information data set established in the step 1, establishing a GRU network for predicting short-term distributed photovoltaic output power;
and 3, aiming at the GRU network established in the step 2, establishing an EHO algorithm model for optimizing GRU network parameters.
2. The EHO-GRU network-based short-term distributed photovoltaic output power prediction method of claim 1, wherein the specific steps of step 1 comprise:
s1-1, collecting historical information data;
the historical information data set in the step 1 comprises a historical environment information data set and a historical power grid operation information data set in the last year; the collection interval of the historical information data is 1 minute;
the historical environmental information data set comprises irradiance, temperature, wind direction, wind speed, humidity and air pressure; the historical power grid operation information data set comprises historical power generation power data and historical load power data;
s1-2, selecting characteristics of the historical information data set.
3. The EHO-GRU network-based short-term distributed photovoltaic output power prediction method according to claim 2, wherein the step S1-2 comprises:
s1-2-1, taking photovoltaic historical generated power data as an output vector, and taking irradiance, temperature, wind direction, wind speed, humidity, air pressure and historical load as input vectors;
s1-2-2, converting each column of characteristic values in an input vector into a histogram by utilizing a histogram function in a wrapping type characteristic selection algorithm, sequencing the importance of the input vector by utilizing a T _ Split and a T _ Gain function in LGBFS, and eliminating the characteristics with the importance of less than 5% to form the optimized input vector.
4. The EHO-GRU network-based short-term distributed photovoltaic output power prediction method of claim 3, wherein the step 2 comprises:
s2-1, taking the optimized input vector obtained in the step 1 as an input variable x of the GRU network a ；
S2-2, input variable x in step S2-1 a Training through a plurality of hidden layers in a GRU network in sequence, wherein the hidden layers comprise a reset gate function r at t moment t And the update gate function u at time t t ；
Wherein, the update gate function is used for reserving key time sequence related information of the input variable, specifically
In the formula, g is a Sigmoid activation function;
is the input vector at the time t;
the output quantity of the hidden layer at the time t-1; w 1 For the dimension of the matrix to be l × c 1 Where l is the dimension of the hidden layer output vector, c 1 And an input variable x a OfThe number is the same; v 1 For the dimension of the matrix to be l × c 2 A weight matrix of (2), wherein c 2 Outputting a vector dimension for a previous hidden layer; w 1 ，V 1 Elements in the matrix are real numbers with random initial values respectively;
the reset gate function is used for removing unnecessary time sequence related information of the input variable, and the generalization capability of the GRU network can be improved; the reset gate function may be specifically expressed as
In the formula, W 2 Is the dimension of the matrix, is l × c 2 A weight matrix of (a); v 2 Is the dimension of the matrix, is l × c 2 A weight matrix of (a); w 2 ，V 2 Elements in the matrix are real numbers with random initial values respectively;
the state quantity of the current hidden layer is output by the last hidden layer
After passing through reset gate and input vector
The splicing composition can be expressed as
In the formula, H t The state quantity of the current hidden layer; tan h is the conversion of input to [ -1,1]An activation function within a range; w 3 Is that the dimension of the matrix is l x (c) 1 +c 2 ) The elements in the matrix are real numbers with random initial values. h is t-1 The output quantity of the hidden layer at the moment t-1 is shown;
the state quantity of the current hidden layer is the output quantity of the current hidden layer after passing through the update gate
In particular to
Output quantity of current hidden layer
Continuously sending to the next hidden layer as input, and circularly reciprocating to form output p of GRU network a Namely the predicted photovoltaic power generation power.
5. The method for predicting the short-term distributed photovoltaic output power based on the EHO-GRU network according to claim 4, wherein the step 3 is to apply an algorithm (EHO) like group herding to the W matrix to be trained of the GRU network in the step 2 1 ，V 1 ，W 2 ，V 2 And W 3 Optimizing the parameters;
the method comprises the following specific steps:
s3-1, initializing elephant group nomadic algorithm parameters; the elephant swarm nomadic algorithm parameters comprise the number n of elephants in the elephant swarm, the searching dimension of the elephant swarm is D, and the maximum iteration number is I max And the initial nomadic location, i.e. the initial solution x m ；
S3-2, calculating a nomadic position fitness function F of the nomadic position of the current turn elephant in the group nomadic algorithm, and evaluating the goodness and badness of the nomadic position; the nomadic location fitness function of the elephant herding and nomadic algorithm can be expressed as
In the formula, e m Is the predicted average percentage error; e.g. of the type r To predict the root mean square error; n is a radical of p Predicting the number of power points;
are respectively the firstThe real values and the predicted values of the i power points;
s3-3, searching a new nomadic position based on a neighborhood search algorithm; in particular to
v m,i ＝x m,i +α(x m,i -x gb )
In the formula, v m,i Obtaining a new nomadic position for the elephant i through a neighborhood search algorithm; x is the number of m,i Is the original nomadic position of elephant i; x is the number of gb The nomadic position with the highest fitness function in all the current nomadic positions is taken as the nomadic position; alpha is [0,1]A random number in between;
s3-4, designing a probability selection formula to determine whether the elephant is nomadic to a position with a higher fitness function value of the nomadic position; the nomadic location selection probability for elephant i can be expressed as
In the formula (f) i Updating the probability for the nomadic location of elephant i; f i The function value of the suitability degree of the nomadic position of the elephant i is obtained;
s3-5, judging whether the search times reach the maximum iteration times I max (ii) a If yes, outputting the optimal solution of the current turn of the elephant herding algorithm, namely W 1 ，V 1 ，W 2 ，V 2 And W 3 The optimum parameter of (2); and if not, performing the nomadic position search of the next round, and returning to the step S3-2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211496034.5A CN115758901A (en) | 2022-11-25 | 2022-11-25 | EHO-GRU network-based short-term distributed photovoltaic output power prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211496034.5A CN115758901A (en) | 2022-11-25 | 2022-11-25 | EHO-GRU network-based short-term distributed photovoltaic output power prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115758901A true CN115758901A (en) | 2023-03-07 |
Family
ID=85338703
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211496034.5A Pending CN115758901A (en) | 2022-11-25 | 2022-11-25 | EHO-GRU network-based short-term distributed photovoltaic output power prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115758901A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116822902A (en) * | 2023-07-17 | 2023-09-29 | 国网江苏省电力有限公司灌云县供电分公司 | Resource aggregate cluster division method for regional power grid industrial load and peripheral distributed new energy based on artificial double image swarm algorithm |
-
2022
- 2022-11-25 CN CN202211496034.5A patent/CN115758901A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116822902A (en) * | 2023-07-17 | 2023-09-29 | 国网江苏省电力有限公司灌云县供电分公司 | Resource aggregate cluster division method for regional power grid industrial load and peripheral distributed new energy based on artificial double image swarm algorithm |
| CN116822902B (en) * | 2023-07-17 | 2024-06-18 | 国网江苏省电力有限公司灌云县供电分公司 | Resource aggregate cluster division method for regional power grid industrial load and peripheral distributed new energy based on artificial double image swarm algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fan et al. | Development of prediction models for next-day building energy consumption and peak power demand using data mining techniques | |
| CN110942205B (en) | Short-term photovoltaic power generation power prediction method based on HIMVO-SVM | |
| CN112380765A (en) | Photovoltaic cell parameter identification method based on improved balance optimizer algorithm | |
| Kolhe et al. | GA-ANN for short-term wind energy prediction | |
| CN109636054A (en) | Solar energy power generating amount prediction technique based on classification and error combination prediction | |
| CN110866633A (en) | Micro-grid ultra-short term load prediction method based on SVR support vector regression | |
| CN115481788B (en) | Phase change energy storage system load prediction method and system | |
| CN112307672A (en) | BP neural network short-term wind power prediction method based on cuckoo algorithm optimization | |
| CN117114438A (en) | Building area energy system cold and hot load data driving prediction method with flexibility and interpretability | |
| CN115758901A (en) | EHO-GRU network-based short-term distributed photovoltaic output power prediction method | |
| CN114676923A (en) | Method and device for predicting generated power, computer equipment and storage medium | |
| CN114648147A (en) | IPSO-LSTM-based wind power prediction method | |
| CN116845875A (en) | WOA-BP-based short-term photovoltaic output prediction method and device | |
| CN116187540A (en) | Wind power station ultra-short-term power prediction method based on space-time deviation correction | |
| CN117454939A (en) | Wind power prediction method for optimizing BP neural network based on NSABO-Gold algorithm | |
| CN111401659A (en) | Ultra-short-term or short-term photovoltaic power generation power prediction method based on case reasoning | |
| Wu et al. | Short-term electric load forecasting model based on PSO-BP | |
| CN114611757A (en) | Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network | |
| CN113095534A (en) | Wind power prediction method combining ARIMA and improved Elman neural network | |
| Wang et al. | A hybrid model with combined feature selection based on optimized VMD and improved multi-objective coati optimization algorithm for short-term wind power prediction | |
| CN113762591A (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM counterstudy | |
| Kermia et al. | Photovoltaic power prediction using a recurrent neural network RNN | |
| CN116681154A (en) | Photovoltaic power calculation method based on EMD-AO-DELM | |
| CN113449466B (en) | Solar radiation prediction method and system for optimizing RELM based on PCA and chaos GWO | |
| Xiyun et al. | A combination method in photovoltaic power forecasting based on the correlation coefficient |
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 |