CN116542882A - Photovoltaic power generation smoothing method, system and storage medium - Google Patents
Photovoltaic power generation smoothing method, system and storage medium Download PDFInfo
- Publication number
- CN116542882A CN116542882A CN202310821202.1A CN202310821202A CN116542882A CN 116542882 A CN116542882 A CN 116542882A CN 202310821202 A CN202310821202 A CN 202310821202A CN 116542882 A CN116542882 A CN 116542882A
- Authority
- CN
- China
- Prior art keywords
- power generation
- photovoltaic
- photovoltaic power
- weight
- neural network
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 114
- 238000010248 power generation Methods 0.000 title claims abstract description 75
- 238000009499 grossing Methods 0.000 title claims abstract description 38
- 238000003860 storage Methods 0.000 title claims abstract description 18
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 50
- 238000013528 artificial neural network Methods 0.000 claims abstract description 44
- 230000002068 genetic effect Effects 0.000 claims abstract description 38
- 238000005457 optimization Methods 0.000 claims abstract description 16
- 238000004146 energy storage Methods 0.000 claims abstract description 15
- 230000008878 coupling Effects 0.000 claims abstract description 8
- 238000010168 coupling process Methods 0.000 claims abstract description 8
- 238000005859 coupling reaction Methods 0.000 claims abstract description 8
- 230000008569 process Effects 0.000 claims description 35
- 238000000354 decomposition reaction Methods 0.000 claims description 23
- 230000006870 function Effects 0.000 claims description 17
- 230000035772 mutation Effects 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 7
- 230000007614 genetic variation Effects 0.000 claims description 7
- 210000002569 neuron Anatomy 0.000 claims description 6
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 230000006641 stabilisation Effects 0.000 abstract description 5
- 238000011105 stabilization Methods 0.000 abstract description 5
- 230000009466 transformation Effects 0.000 description 9
- 108090000623 proteins and genes Proteins 0.000 description 7
- 238000012545 processing Methods 0.000 description 6
- 239000013598 vector Substances 0.000 description 6
- 230000033228 biological regulation Effects 0.000 description 5
- 238000010606 normalization Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000005338 heat storage Methods 0.000 description 4
- 238000003672 processing method Methods 0.000 description 4
- 239000000126 substance Substances 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000005855 radiation Effects 0.000 description 3
- 238000010521 absorption reaction Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000029087 digestion Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005265 energy consumption Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000001737 promoting effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000000087 stabilizing effect Effects 0.000 description 2
- 108010076504 Protein Sorting Signals Proteins 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000002759 chromosomal effect Effects 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 239000003344 environmental pollutant Substances 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000010353 genetic engineering Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000000178 monomer Substances 0.000 description 1
- 231100000719 pollutant Toxicity 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- 230000002195 synergetic effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/004—Generation forecast, e.g. methods or systems for forecasting future energy generation
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
- H02J2300/20—The dispersed energy generation being of renewable origin
- H02J2300/22—The renewable source being solar energy
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/50—Photovoltaic [PV] energy
- Y02E10/56—Power conversion systems, e.g. maximum power point trackers
Abstract
The invention discloses a photovoltaic power generation smoothing method, a system and a storage medium, wherein the method is applied to a photovoltaic photo-thermal coupling power generation system and comprises the following steps: optimizing an Elman neural network based on a genetic algorithm to predict photovoltaic power generation power; meanwhile, the neural network adopts a self-adaptive learning rate adjustment method based on the magnitude of network errors to adjust the learning rate, and the weight and the threshold value of the neural network are adjusted by adopting an adjustment method with additional momentum; and simultaneously, the energy storage is combined to stabilize the fluctuation of the photovoltaic photo-thermal output power based on a smoothing algorithm, so that smooth and stable output power is obtained, and the optimal capacity power of the storage battery is determined. According to the invention, the high-precision prediction of photovoltaic power generation is realized by adopting the Elman neural network optimized by various optimization strategies, the prediction error of the photovoltaic power is not more than 10%, the fluctuation stabilization is realized by matching with the energy storage power supply, and the stable operation of the power grid is ensured.
Description
Technical Field
The invention belongs to the technical field of solar clean utilization, and relates to a photovoltaic power generation smoothing method, a photovoltaic power generation smoothing system and a storage medium.
Background
Solar photovoltaic power generation and wind energy have the characteristics of volatility and uncertainty, and weather conditions have great influence on the output power of the solar photovoltaic power generation and wind energy. The photovoltaic power generation is connected to the grid on a large scale, so that important influence can be generated on the operation characteristics of the power grid, and meanwhile, the power quality problems such as harmonic pollution, voltage fluctuation and the like can be brought to the power grid.
The solar thermal power generation technology can be coupled with large-scale low-cost heat storage (about 0.05 yuan/kWh), improves the stability of power output, can participate in flexible peak regulation, does not additionally cause the phenomenon of increasing pollutant emission or greatly increasing energy consumption during peak regulation, and is a power grid friendly renewable energy power generation technology with development potential.
Along with the continuous increase of the generation scale of the new wave energy, the additional cost paid by the power grid system for the new wave energy power is continuously increased, and the development of a low-cost renewable power supply peak regulation technology is needed to ensure the safe operation of the power grid system. Wherein, the two aspects of the prediction of the improved generated power and the smoothing of the output power can be considered: the prediction of the photovoltaic power generation power can quantify the prediction uncertainty, and data support is provided for the operation and control of a power grid containing high-permeability photovoltaic power generation, so that the power grid optimization problem under the condition of the photovoltaic uncertainty is solved better, and the photovoltaic digestion capacity of the power grid is improved; and a smooth control strategy of the photovoltaic power generation system is researched, so that the output power of the system is stabilized, and the operation efficiency of the system is improved.
Based on the method, the invention provides a photovoltaic photo-thermal new energy online smoothing system and a photovoltaic photo-thermal new energy online smoothing method, and an intelligent energy system with the synergistic output of multiple energy sources is constructed by taking the operation constraint of photovoltaic, energy storage and load, the system safety constraint and the coupling characteristic among multiple energy sources into consideration, so that the stability and the absorption of fluctuation renewable energy sources are satisfied. The power generation smoothing system is favorable for improving the power supporting capacity of new energy, promoting clean and stable production of electric energy, reducing wind and light abandoning, promoting the absorption of renewable energy and flexible interactive regulation of source network storage, is an important grip for improving the comprehensive efficiency of the electric energy system, and has important practical significance and profound strategic significance for constructing a clean, low-carbon, safe and high-efficiency modern electric energy system.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a photovoltaic power generation smoothing method, a photovoltaic power generation smoothing system and a storage medium. According to the scheme, high-precision prediction of photovoltaic power generation and photovoltaic photo-thermal power generation internet surfing stabilization can be realized, and smooth output power is obtained.
The technical scheme adopted by the invention is as follows:
in a first aspect, an embodiment of the present invention provides a photovoltaic power generation smoothing method, where the method is applied to a photovoltaic photo-thermal coupling power generation system, and includes the following steps:
optimizing an Elman neural network based on a genetic algorithm to predict photovoltaic power generation power; meanwhile, the neural network adopts a self-adaptive learning rate adjustment method based on the magnitude of network errors to adjust the learning rate, and the weight and the threshold value of the neural network are adjusted by adopting an adjustment method with additional momentum;
and simultaneously, the energy storage is combined to stabilize the fluctuation of the photovoltaic photo-thermal output power based on a smoothing algorithm, so that smooth and stable output power is obtained, and the optimal capacity power of the storage battery is determined.
In the above technical solution, further, the Elman neural network structure is divided into four layers: the learning algorithm adopted by the input layer, the hidden layer, the receiving layer and the output layer is a gradient descent method, if the output of the output layer is inconsistent with the expected output, the output error is transferred to a counter propagation stage, and is transmitted to the input layer through the hidden layer and distributed to the neurons of each layer to obtain error signals of each layer, so that the weight and the threshold value of the neurons of each layer are updated to achieve ideal error precision.
Further, the self-adaptive learning rate adjustment method based on the network error size specifically comprises the following steps:
in the method, in the process of the invention,for the sum of squares of errors of the actual output and the expected output of the network, k represents the kth iteration, k+1 represents the kth+1 iteration,/o->Is the learning rate.
Further, the adjusting method with additional momentum is as follows: when the weight and the threshold value are updated each time, a momentum factor proportional to the previous adjustment quantity is additionally added to the adjustment quantity of the existing weight and the threshold value, the essence of the added momentum is the linear superposition of the previous weight and the threshold value adjustment quantity, and the added momentum is correspondingly increased along with the increase of the weight and the threshold value adjustment quantity, so that the stability and the operation speed of the network are improved, and the adjustment method of the weight and the threshold value of the Elman neural network with the added momentum comprises the following steps:
in the method, in the process of the invention,and->The adjustment amounts of the weight and the threshold value are respectively; />And->The adjustment quantity with additional momentum respectively representing the k+1st iteration weight and the threshold value; i. j is the node position of the output layer and the hidden layer respectively; />As a momentum factor, k represents the kth iteration, and k+1 represents the kth+1 iteration.
Further, the genetic algorithm-based optimization Elman neural network predicts photovoltaic power generation power; comprises the following steps:
setting population number and optimization targets based on a genetic algorithm, and coding an initialization weight and a threshold of the Elman neural network;
determining an fitness function, and calculating fitness of individuals in various groups;
determining a genetic strategy, and performing selection, crossing and mutation operations on the population to obtain offspring;
determining whether the population reaches an optimization target or reaches the maximum iteration number, if so, ending, and if not, returning to continue the fitness calculation until the population reaches the optimization target or reaches the maximum iteration number, so as to obtain the optimal weight and the threshold value of the network;
and performing network training and predicting results based on the optimal weight and the threshold.
Further, the mutation operation is performed in two steps:
the first step: according to the preset variation probabilityCalculating individual individuals in a populationThe probability of variation is:
wherein L represents the coding length, n represents the population individual number, and one [0,1] is given]Random variable x in between, whenWhen the variable is mutated, whenWhen the method is used, mutation operation is not needed;
and a second step of: calculating the probability of genetic variation in individuals, wherein the expected number of times of genetic variation is calculated under normal conditionsHowever, under the new variant condition, the variant probability is set asThe total expected number of genetic variations at this time wasThe method comprises the steps of carrying out a first treatment on the surface of the The two are equal to each other, i.e. a new mutation probability can be obtainedThe following is shown:
。
further, the smoothing algorithm is wavelet transformation, moving average, variational modal decomposition VMD or empirical modal decomposition EMD.
In a second aspect, an embodiment of the present invention provides a photovoltaic power generation smoothing system, where the system is applied to a photovoltaic photo-thermal coupling power generation system, to implement any one of the above photovoltaic power generation smoothing methods.
In a third aspect, an embodiment of the present invention provides a computer apparatus, including:
one or more processors;
a memory for storing one or more programs;
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement any of the photovoltaic power generation smoothing methods described above.
In a fourth aspect, embodiments of the present invention provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements any of the above photovoltaic power smoothing methods.
The beneficial effects of the invention are as follows:
according to the invention, the intelligent dispatching comprehensive control system is constructed by researching the output characteristics of the photovoltaic power generation system, the high-precision prediction of photovoltaic power generation is realized by adopting the Elman neural network optimized by various optimization strategies, the photovoltaic power prediction error is not more than 10%, the fluctuation stabilization is realized by matching with the energy storage battery, and the stable operation of the power grid is ensured. The all-weather real-time dynamic monitoring of the power grid is realized through the comprehensive control system, and accurate, real-time and stable data are provided; by automatically controlling the energy storage battery, the electric energy is output and consumed in time to stabilize fluctuation, and the dynamic stability of the photovoltaic power generation system is enhanced. FIG. 1 is a graph of actual power versus predicted power.
Drawings
FIG. 1 is a graph of actual power versus predicted power using the method of the present invention;
FIG. 2 is a training flow chart of an Elman neural network;
FIG. 3 is a flowchart of the operation of a genetic algorithm;
FIG. 4 is a flowchart of the operation of a genetic algorithm optimized Elman neural network;
FIG. 5 is a schematic diagram of the wavelet transform principle;
FIG. 6 is a schematic diagram of a moving average configuration battery capacity power scheme;
FIG. 7 is a schematic diagram of a VMD configured battery capacity power scheme;
fig. 8 is a schematic diagram of an EMD configured battery capacity power scheme.
Detailed Description
The technical scheme of the invention is further described in detail below with reference to the attached drawings and specific embodiments.
The technical scheme of the invention is divided into two parts for developing photovoltaic prediction and smoothing strategy research, wherein the first part optimizes the prediction of the photovoltaic system and improves the prediction precision of the photovoltaic system, and the second part develops the research on the output power smoothing of the photovoltaic system; the output power of the system is stabilized, the operation efficiency of the system is improved, the two parts are supported by each other, and the process is advanced layer by layer:
the photo-thermal system in the invention can adopt a solar Brayton cycle power generation system. The dynamic operation simulation platform of the gas turbine is established through a modularization method, the dynamic models such as the gas turbine dynamic simulation platform, the high-temperature chemical heat storage system and the like are coupled, the operation results of all working conditions are tested, the brayton power generation system model of the high-temperature chemical heat storage is established, the changes of the output stability, the load response speed and the maximum peak regulation capacity of the coupling system under different operation schemes are studied, and the design work of the air brayton cycle process system of the high-temperature chemical heat storage is completed.
Aiming at a photovoltaic photo-thermal coupling system formed by a Qingshan lake photo-thermal power generation system of Zhejiang university and a Qingshan lake photovoltaic power generation system of Zhejiang university as an example, a complete power station index system is established based on solar resource indexes, electric quantity indexes, energy consumption indexes, photovoltaic system performance indexes, equipment operation level indexes and power station operation indexes, and equipment modeling is carried out; the method comprises the steps of collecting local solar meteorological data in real time, calculating the utilization condition of solar resources, solar radiation quantity and average air temperature through an analysis model of the optical resources, deducing theoretical power generation condition, analyzing conversion efficiency of the optical resources, comparing actual electric quantity with theoretical electric quantity, calculating electric quantity loss condition through comparison of multistage electric quantity (direct current, alternating current and boosting), and analyzing performance index, equipment operation level index and electric quantity loss index of a photovoltaic system; and establishing a photovoltaic data model, predicting photovoltaic power generation power by establishing a short-term load prediction and photovoltaic power generation power prediction model, providing a strategy for stabilizing photovoltaic output power fluctuation by combining energy storage by combining algorithms such as wavelet decomposition and the like, smoothing unstable small fluctuation of photovoltaic output, and obtaining smoother photovoltaic output. Specific:
photovoltaic power generation power prediction algorithm
1. Photovoltaic power generation data selecting and processing method
The prediction of the photovoltaic power generation power can quantify the prediction uncertainty, and data support is provided for the operation and control of the power grid containing high-permeability photovoltaic power generation, so that the power grid optimization problem under the condition of the photovoltaic uncertainty is solved better, and the photovoltaic digestion capacity of the power grid is improved. Factors influencing photovoltaic power generation power are mainly two types: the first is an internal factor and the second is an external factor. The internal factors comprise the inclination angle of the photovoltaic panel, the geographical position of each photovoltaic power station, the photoelectric conversion efficiency and the like, and the internal factors are the same because the data of the same photovoltaic power station are used when the photovoltaic power generation power is predicted, and the internal factors do not need to be considered when the power is predicted; external factors including weather type, solar irradiance, temperature, humidity, wind speed, etc., if all of the external factors are taken into account, will necessarily affect the running speed of the predictive model. Therefore, before the prediction is performed, the main factors influencing the photovoltaic power generation power are firstly screened, and the reduction of the running speed of the prediction model caused by the excessively high data dimension is reduced on the premise of ensuring the prediction accuracy.
In the photovoltaic power generation power prediction, the day on which power is to be predicted is generally called a prediction day, a few days before the day is called a history day, and the most similar of the history day and the predicted solar and air condition factors is called a similar day. The principle of the similarity day theory is to construct a similarity vector, select a characteristic quantity to compare with and match with predicted day data, and then select the historical day data with highest similarity as a similarity day sample. When predicting photovoltaic power generation power, training of a prediction model is a key factor in determining final prediction accuracy. If the historical day data training with high similarity with the prediction day is adopted, the prediction effect of the model can be improved, the training time is shortened, the prediction accuracy can be improved, and the photovoltaic power generation power curves of the similar day and the prediction day are highly similar. The similarity day theory can efficiently perform preliminary screening on a large amount of data, can customize vector groups, has strong adaptability, and can automatically update data according to prediction requirements. The magnitude of the photovoltaic power generation power is related to weather types and meteorological factors, so that the similar weather types and meteorological factors are selected and combined with the photovoltaic power generation power sequence to serve as input data of a photovoltaic power prediction model, and the photovoltaic power generation power of a prediction day can be well obtained by the method, so that the calculation speed and the prediction accuracy are improved.
For the built photovoltaic power station, internal factors are determined, and the photovoltaic output power is mainly influenced by meteorological factors such as irradiance, temperature, wind speed, humidity and the like. Therefore, in the case where the seasonal factor at which the predicted day is located is determined, the feature vector of the similar day is selected as follows:
in the method, in the process of the invention,is of weather type;maximum irradiance, minimum irradiance, average irradiance on day i, respectively;respectively isMaximum temperature, minimum temperature, average temperature on day i;the maximum wind speed, the minimum wind speed and the average wind speed of the ith day are respectively;maximum humidity, minimum humidity, average humidity on day i, respectively.
In general, data collected by a photovoltaic power station are composed of different dimensions and magnitude orders, and if unbalanced scaling is performed in each dimension, optimal solutions of a model may be inconsistent with initial solutions, so that prediction accuracy of the model is reduced. In order to prevent interference of the data on operation, the invention adopts a normalization processing method to map the characteristic values of each dimension to the same interval of [0,1] respectively, so that the dimension of each characteristic value is the same, thereby improving the convergence speed and the prediction precision of the prediction model. Normalization process equation 6-2:
（6-2）
in the method, in the process of the invention,is data subjected to normalization processing; />Is input data; />And->Respectively, the maximum value and the minimum value of the input data.
After the feature vectors are defined, in order to unify the dimensions of the respective vectors, normalization formula 6-3 is performed on the respective components by using a polar difference method:
（6-3）
in the method, in the process of the invention,is->First part of the vectors>A component; />And->Respectively +.>Maximum and minimum of the individual components.
After normalization treatment, the prediction day and the thThe eigenvector of day is equation 6-4:
（6-4）
and->In->The correlation coefficient of each factor is equation 6-5:
（6-5）
in the method, in the process of the invention,； />for resolution factor->。
Summarizing the association coefficients of the components and definingAnd->The similarity between them is equation 6-6:
（6-6）
the similarity is calculated to select appropriate historical data for prediction, so that prediction accuracy is improved. Equation (6-6) uses a continuous multiplication to calculate the similarity, which has the advantage of maximally solving the weight problem of each component.
2. Elman neural network based on genetic algorithm optimization
(1) Elman neural network
The Elman neural network is a dynamic recurrent neural network, which is added with a receiving layer on the basis of a general feedforward network. The structure of Elman neural networks is generally divided into four layers: an input layer, an hidden layer receiving layer and an output layer.
Input layer: the main function is to input data;
hidden layer: the method is used for connecting feedback of the output layer and the receiving layer, and using updating of weights to influence input data, and a Sigmoid function is usually selected by an activation function;
bearing layer: the state layer is also called, receives a liability signal from the hidden layer and is used for memorizing the output of the hidden layer at one moment and finishing the delay input of data, so that the sensitivity to historical data is obtained, and the capability of processing dynamic information by the network is enhanced;
output layer: the output of the hidden layer is linearly weighted and then the result is output, typically the activate function selects the pureline function.
The Elman neural network is characterized in that delayed input and storage of data by the receiving layer are utilized to carry out self-connection on the output and input of the hidden layer. The mathematical model of the network is:
（6-7）
（6-8）
（6-9）
in the method, in the process of the invention,、/>、/>the output of the hidden layer, the bearing layer and the output layer are respectively; />An activation function for an hidden layer; />An activation function for the output layer; />The connection weight between the receiving layer and the hidden layer is used; />Connecting weights between the input layer and the hidden layer; />The connection weight between the hidden layer and the output layer is obtained; />A threshold value that is an implicit layer; />A threshold value for the output layer; />For feeding back gain factor->。
The learning algorithm adopted by the Elman neural network is a gradient descent method, if the output of the output layer is inconsistent with the expected output, the output error is transferred to a counter propagation stage, and is transmitted to the input layer through the hidden layer, and is distributed to the neurons of each layer to obtain error signals of each layer, so that the weight and the threshold value of the neurons of each layer are updated, and the ideal error precision is achieved. Error functionDefined as formulas 6-10:
（6-10）
in the method, in the process of the invention,is the desired output of the network; />Is the actual output of the network.
Learning training of Elman neural networks consists of two processes: forward propagation and error back propagation. In the forward propagation process, input information is processed layer by layer, after the input information is transmitted to an output layer, if the error is not 0, the error is propagated reversely, and the error gradually approaches to the target precision by calculating and adjusting weights among layers. Thus, each time forward and backward propagation is completed, one iteration is completed, and one weight is adjusted until the error reaches the requirement, the learning process can not be finished. Updating the weight values by formulas 6-11 and 6-12:
（6-11）
（6-12）
in the method, in the process of the invention,is an error function; />Is the learning rate.
By gradient descent, the error function can be usedConnecting weight to each layer>、/>、/>Obtaining a bias derivative so as to obtain a learning algorithm of the Elmman neural network:
（6-13）
（6-14）
（6-15）
in the method, in the process of the invention,、/>、/>respectively->、/>、/>A correction amount of (2); />、/>、/>The node positions of the output layer, the hidden layer and the input layer are respectively; />、/>、/>Several are +.>、/>、/>Is a learning rate of (a).
At a fixed learning rate, the overall training process is difficult to achieve efficiency. If the learning rate is too small, the convergence complexity of the network is greatly increased, so that the training process becomes very slow; if the learning rate is too high, the network cannot converge and wander around the optimal value. In order to improve the prediction accuracy of the model, an adaptive learning rate adjustment formula based on the magnitude of network errors is provided as 6-16;
（6-16）
in the method, in the process of the invention,the sum of squares of the errors of the actual output and the desired output of the network.
If the additional momentum is not applied, the Elman neural network is easy to sink into local minima, and the training of the network is difficult to achieve global optimum, so that a momentum factor proportional to the previous adjustment quantity is additionally added on the adjustment quantity of the existing weight and the threshold value, and the local minima are scratched by utilizing the action of the additional momentum. The essence of the additional momentum is linear superposition of the previous weight and the threshold adjustment quantity, and the additional momentum is correspondingly increased along with the increase of the weight and the threshold adjustment quantity, so that the stability and the operation speed of the network are improved. The adjustment method of the weight and the threshold value of the Elman neural network with the additional momentum is as shown in formulas 6-17 and 6-18:
（6-17）
（6-18）
in the method, in the process of the invention,and->The adjustment amounts of the weight and the threshold value are respectively; />Is a momentum factor.
The training procedure of Elman neural network is shown in fig. 2.
(2) Genetic algorithm
The main operation steps of the genetic algorithm are as follows: coding, fitness value calculation, selection, crossover, mutation and the like.
(1) Encoding: in the global optimizing process, the genetic algorithm does not directly optimize the parameters, but needs to convert the data of the parameters into character strings identified by the genetic algorithm, and the process is called coding. The coding process must ensure that the process is complete and non-redundant, i.e., that all parameters to be optimized are coded, that each chromosomal gene after coding and the last solution can be mapped fully, and that the gene strings and the actual solutions that are solved can also be mapped fully;
(2) fitness function: the suitability is the only evaluation standard in the population optimization process by the genetic algorithm, and the individual winner and winner elimination operation is carried out according to the only evaluation standard. In the evolution process, the fitness value is better than the external condition of nature in natural selection, and if the fitness is met by individuals in the population, the individuals are judged to be excellent, and the individuals are reserved; if the result is not satisfied, the individuals judged to be inferior are eliminated.
(3) Selection operation: a selection operation for selecting the optimal individual having the highest fitness based on the fitness value, then taking the optimal individual as the previous generation, and transferring the "gene" to the next generation. The most common choice for optimal individuals is the betting board method, which is mainly chosen with reference to the proportion of fitness of the individual. The fitness value of the individual and the average fitness value of the population are calculated respectively, and the fitness value of the individual can be selected and kept only when the fitness value reaches the proportion set by the average fitness of the population.
Setting up a populationIndicating that the number of all individuals in the population isEach individualHas a fitness value ofThe probability of not being eliminated is shown in equation (6-19):
(4) crossover operation: the method is an indispensable step in the crossover operation genetic algorithm, can not only keep the excellent genes of the previous generation, but also improve the bad genes, so that the genetic operation has higher fault tolerance and openness. The crossing step can be divided into three steps:
the first step: namely, selecting an individual meeting the fitness requirement through calculation of the individual fitness and the group fitness.
And a second step of: the position of the crossover is selected in the code string. If the coding length of the previous generation is determined to be L, one or more positions are randomly selected from the range of [ L, L-1] to perform the cross operation.
And a third step of: and performing crossover. Setting the probability of crossing, setting the probability value between 0 and 1, and carrying out gene crossing on the previous generation individuals selected in the first step and the second step and the crossing position selected according to the probability value to obtain new population individuals.
(5) Mutation operation: mutation operation is also an extremely important step in genetic mutation, and in terms of biological theory, a population is subjected to a huge disaster which is difficult for centuries suddenly, individuals which can adapt to the disaster are reserved, most of the individuals which cannot adapt to the disaster are eliminated, but a small part of the individuals exist for survival, and the genes slowly change and adapt to the environment and survive.
In normal genetic operation, the variation of individuals in a population is obtained by reversing binary symbols of individual genetic bit strings according to set probability, but the method can lead to waste of resources in the actual genetic operation process, so the variation operation is divided into two steps:
the first step: variation probability set before learning according to algorithmThe probability of variation of individual individuals in the population is calculated as shown in the formula (6-20):
where L represents the code length and n represents the population number, if given a [0,1]]Random variable x in between, whenWhen the variable is mutated, whenIn this case, the mutation operation is not required.
And a second step of: calculating the probability of genetic variation in individuals, wherein the expected number of times of genetic variation is calculated under normal conditionsHowever, under the new variant condition, the variant probability is set asAnd (3) representing. The total expected number of genetic variants at this time was. The two are equal to each other, i.e. a new mutation probability can be obtainedAs shown in the formula (6-21):
greater thanWhen L increases infinitely and tends to infinity, the two are equal.
The specific flow of the genetic algorithm is shown in fig. 3:
(3) Genetic algorithm optimized Elman neural network
In general, the learning process of the Elman neural network adjusts the weight and the threshold value through the back propagation of error values, but the initial weight and the threshold value of the Elman neural network are random numbers between [0,1], and the weight and the threshold value of the Elman neural network can be optimized by using the genetic algorithm, so that the Elman reaches the global optimal solution, and the prediction precision can be greatly improved. The parameters to be optimized by the genetic algorithm are the weights and thresholds of the Elman neural network.
The genetic algorithm optimizes the Elman neural network as follows:
(1) and (5) selecting a coding method. Because the Elman network predicts the photovoltaic power generation power, the weight and the threshold of the neural network are changed to enable the prediction error to reach a preset value, namely, the genetic algorithm is optimized to be the weight and the threshold of the Elman neural network. Firstly, determining the structure of a neural network according to the characteristics of photovoltaic power generation power data, determining the number of weights and thresholds of the network by determining the structure of an Elman network, and coding each weight and threshold of the network, wherein each weight and threshold parameter represents one individual of a population, and all the individuals form a population. If the number of the groups is too large, the iteration is too long, the calculation amount is too huge, and the method is unfavorable for advocating economy and practicability in practical engineering.
(2) And (5) determining a fitness function. Firstly, initializing an Elman neural network for photovoltaic power generation, initializing a weight and a threshold, performing preliminary training to obtain the photovoltaic power generation to obtain an actual predicted value, then calculating the photovoltaic power generation to obtain a difference value between an actual output value and a true value, and setting the square sum of the difference values as an initial fitness value of a genetic algorithm. The fitness function expression is shown in the formula (6-22):
in the method, in the process of the invention,the predicted value is represented by a value of the prediction,the desired value is indicated to be the desired value,indicating the number of predicted samples.
(3) Determination of genetic manipulation. According to the general principle of the genetic algorithm, the population scale of the genetic algorithm is mostly set between 20 and 200 for determination, and the genetic algorithm obtains an optimal solution through multiple iterations.
(4) The expression of the selection probability is shown in the formula (6-23):
in the above-mentioned method, the step of,representing the inverse of the fitness value of the individual,indicating the size of the population. Determination of the crossover method: the algorithm adopts a single-point crossing method to select two father parents，Then select a hybrid according toIs hybridized and a new pair of individuals is obtained，. Determination of the mutation method: i.e. determining the probability of variationDue to the real number coding, select the firstThe first of the individual stringsIndividual genesThen, the mutation operation is carried out, and the method is as follows:
in the formula (6-25),representation genesIs defined by the upper bound of (c),representation genesIs defined by the lower boundary of the (c),，representing the number of current iterations and,indicating the number of iterations that are set,is thatA random number therebetween; and when the iteration times of the genetic algorithm reach the set times or the fitness of all individuals in the population meets the fitness value, ending the iteration of the genetic algorithm.
(5) The initial population is randomly generated.
(6) And (5) calculating the fitness. And calculating the fitness value of the individual according to a formula of the fitness function.
(7) And (3) selecting, crossing and mutating the population according to the genetic strategy in the step (3) to obtain offspring.
(8) And (3) determining whether the population reaches an optimization target or the maximum iteration number, if so, ending, and if not, returning to the step (6) until the optimization target is reached.
A flowchart for optimizing Elman neural network using genetic algorithm is shown in fig. 4:
photovoltaic power generation smoothing algorithm
The approximate components in the photovoltaic photo-thermal output power can be extracted by adopting any one method of wavelet transformation, moving average, variation modal decomposition (Variation Mode Decomposition, VMD) and empirical modal decomposition (Empirical Mode Decomposition, EMD), so that the photovoltaic output power meeting the photovoltaic grid-connected condition is obtained, and finally the capacity power configuration of the storage battery is obtained.
1) Wavelet transformation
The wavelet transformation can realize the mutual conversion of the time domain and the frequency domain, has locality and multi-resolution characteristics, is suitable for non-stationary signals, and is mainly used in the fields of signal processing, image processing, voice analysis and other nonlinear science in engineering. Wavelet transformation is generally classified into continuous wavelet transformation and discrete wavelet transformation, and the continuous wavelet transformation is large in calculation amount and high in information redundancy, and is generally discretized and processed by a processing method using discrete wavelet.
Wavelet transformation is also called wavelet decomposition, which is to decompose a signal continuously. The original signal is first decomposed into two parts, a low frequency and a high frequency. Then the low-frequency signal is taken as the original signal to be continuously decomposed into a low-frequency part and a high-frequency part, and the steps are repeated. The specific flow is shown in fig. 5.
The two most important parameters of wavelet decomposition are the wavelet base and the number of layers of decomposition, the wavelet base has an important effect on the data characteristics of signal decomposition, and the number of layers of wavelet decomposition directly affects the frequency characteristics after decomposition.
2) Sliding average
A running average is an algorithm that divides over a time series and averages values over a period of time. After the signal is processed by the sliding average algorithm, the jitter of the signal is reduced and the signal becomes smoother. The photovoltaic output power changes along with the change of solar radiation, and the influence of solar radiation fluctuation on the photovoltaic output power can be effectively eliminated by using a moving average algorithm.
The process of the moving average is as follows, firstly selecting a window with a fixed size, and recording as L; an arithmetic average over a period of length L is then calculated. Then, the window is moved in the direction of increasing the time sequence, and the above steps are repeated until all the average values are obtained. The formula of the sliding average algorithm is shown in (6-26).
（6-26）
In the method, in the process of the invention,-an arithmetic mean of the nth set of data;-the nth initial output power value.
According to the principle of a sliding average algorithm, the larger the sliding window L is, the smoother the curve is after stabilization, so that when the capacity power of the storage battery is configured, the first L meeting the fluctuation requirement of the photovoltaic grid-connected power is the optimal solution. The specific algorithm flow is shown in fig. 6.
3）VMD
VMD is an adaptive, completely non-recursive mode variation and time-frequency signal processing method. The method can effectively process nonlinear and non-stationary signals. The adaptivity of the method is represented by the fact that the number of modal decomposition of a given sequence can be determined according to actual conditions, the subsequent searching and solving process can be adaptive to match the optimal center frequency and the priority bandwidth of each modal, effective separation of intrinsic modal components (Intrinsic Mode Functions, IMF) and frequency domain division of signals can be achieved, further effective decomposition components of original signals are obtained, and finally the optimal solution of the variation problem is obtained.
And the plurality of IMFs are decomposed by the VMD algorithm mode, wherein the low-frequency part shows the trend of the original signal, the signal is smoother, and the more the number of the set mode decomposition is, the smoother the IMF low-frequency part curve is. According to the characteristic, setting the modal decomposition number as n, and iteratively finding out the minimum modal decomposition number meeting the photovoltaic grid-connected power fluctuation from small to large; then taking the obtained IMF low-frequency part as the power of surfing the Internet after stabilization; and finally, calculating the capacity power configuration of the storage battery. The specific flow is shown in fig. 7.
4）EMD
EMD is a time-frequency domain signal processing scheme. The method can perform signal decomposition according to the time scale characteristics of the data, and any basis function is not required to be preset. The EMD has obvious advantages in processing non-stationary and non-linear data, is suitable for analyzing a non-linear non-stationary signal sequence, and has higher signal-to-noise ratio.
The principle of EMD is to decompose a complex signal into a limited number of IMFs, each IMF component contains local feature information of different time scales of the original signal, and the remaining remainder can approximately represent the features of the original signal.
The residual after the EMD decomposition can be approximated to the original signal, and the more smooth the residual is as the number of IMFs is decomposed. Solving the optimal capacity power of the storage battery in the mode shown in fig. 8, initializing the number of IMF decomposition, and judging whether the decomposed remainder can meet the photovoltaic grid-connected fluctuation requirement in an iterative mode. And after the requirements are met, taking the remainder as the photovoltaic grid-connected power, and calculating the capacity and the power of the storage battery.
The energy storage battery with the optimal capacity, which is matched with the installed capacity of the photovoltaic power station, is mainly used for stabilizing the fluctuation part of the photovoltaic exceeding the grid-connected requirement, and the minimum capacity of the required energy storage can be obtained after the smoothing treatment, so that the energy storage investment cost is reduced. The energy storage capacity required for completely filling the photovoltaic vacancy is larger without smoothing treatment. The smoothing is aimed at reducing investment costs. And matching a gas turbine with 30% of installed capacity of the photovoltaic power station and a high-temperature chemical Chu Reguang thermal power generation system according to the basic monomer test characteristics and the working range of high-temperature energy storage. The system can smooth the photovoltaic power generation system through the photo-thermal power generation system and the energy storage system, so that the photovoltaic power generation system is stably connected with the grid, and impact on a power grid is reduced.
In addition, an embodiment of the present invention further provides a computer device, where the computer device includes:
one or more processors;
a memory for storing one or more programs;
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement any of the photovoltaic power generation smoothing methods described above.
In addition, the embodiment of the present invention may further provide a computer readable storage medium having stored thereon a computer program, which when executed by a processor, implements any of the above photovoltaic power generation smoothing methods.
Claims (10)
1. The photovoltaic power generation smoothing method is characterized by being applied to a photovoltaic photo-thermal coupling power generation system and comprising the following steps of:
optimizing an Elman neural network based on a genetic algorithm to predict photovoltaic power generation power; meanwhile, the neural network adopts a self-adaptive learning rate adjustment method based on the magnitude of network errors to adjust the learning rate, and the weight and the threshold value of the neural network are adjusted by adopting an adjustment method with additional momentum;
when the power change exceeding the expected amplitude of the photovoltaic occurs in a future period, the photo-thermal power generation and the intervention of a battery energy storage system are scheduled in advance, the photovoltaic photo-thermal output power fluctuation is stabilized based on a smoothing algorithm by combining energy storage, the smooth and stable output power is obtained, and the optimal capacity power of the storage battery is determined.
2. The method of claim 1, wherein the Elman neural network is structured as four layers: the learning algorithm adopted by the input layer, the hidden layer, the receiving layer and the output layer is a gradient descent method, if the output of the output layer is inconsistent with the expected output, the output error is transferred to a counter propagation stage, and is transmitted to the input layer through the hidden layer and distributed to the neurons of each layer to obtain error signals of each layer, so that the weight and the threshold value of the neurons of each layer are updated to achieve ideal error precision.
3. The photovoltaic power generation smoothing method according to claim 2, wherein the adaptive learning rate adjustment method based on the magnitude of the network error specifically comprises:
in the method, in the process of the invention,for the sum of squares of errors of the actual output and the expected output of the network, k represents the kth iteration, k+1 represents the kth+1 iteration,/o->Is the learning rate.
4. The method of claim 2, wherein the adjusting method with additional momentum is: when the weight and the threshold value are updated each time, a momentum factor proportional to the previous adjustment quantity is additionally added to the adjustment quantity of the existing weight and the threshold value, the essence of the added momentum is the linear superposition of the previous weight and the threshold value adjustment quantity, and the added momentum is correspondingly increased along with the increase of the weight and the threshold value adjustment quantity, so that the stability and the operation speed of the network are improved, and the adjustment method of the weight and the threshold value of the Elman neural network with the added momentum comprises the following steps:
in the method, in the process of the invention,and->The adjustment amounts of the weight and the threshold value are respectively; />And->The adjustment quantity with additional momentum respectively representing the k+1st iteration weight and the threshold value; i. j is the node position of the output layer and the hidden layer respectively; />As a momentum factor, k represents the kth iteration, and k+1 represents the kth+1 iteration.
5. The photovoltaic power generation smoothing method according to claim 1, wherein the genetic algorithm-based optimization Elman neural network predicts photovoltaic power generation; comprises the following steps:
setting population number and optimization targets based on a genetic algorithm, and coding an initialization weight and a threshold of the Elman neural network;
determining an fitness function, and calculating fitness of individuals in various groups;
determining a genetic strategy, and performing selection, crossing and mutation operations on the population to obtain offspring;
determining whether the population reaches an optimization target or reaches the maximum iteration number, if so, ending, and if not, returning to continue the fitness calculation until the population reaches the optimization target or reaches the maximum iteration number, so as to obtain the optimal weight and the threshold value of the network;
and performing network training and predicting results based on the optimal weight and the threshold.
6. The smoothing method for photovoltaic power generation according to claim 5, wherein the mutation operation is performed in two steps:
the first step: according to the preset variation probabilityCalculating individual individuals in the population ∈ ->The probability of variation is:
；
wherein L represents the code length, n represents the population individual number, for one [0,1]]Random variable x in between, whenIn the case of a variant, this variable is subjected to a variation operation, when +.>When the method is used, mutation operation is not needed;
and a second step of: calculating the probability of genetic variation in individuals, wherein the expected number of times of genetic variation is calculated under normal conditionsHowever, under the new variant conditions, the variant probability is set to +.>The total expected number of genetic variants at this time is +.>The method comprises the steps of carrying out a first treatment on the surface of the The two are equal to each other, so that new mutation probability +.>The following is shown:
。
7. the method of claim 1, wherein the smoothing algorithm is wavelet transform, sliding average, variational modal decomposition VMD, or empirical modal decomposition EMD.
8. A photovoltaic power generation smoothing system, wherein the system is applied to a photovoltaic photo-thermal coupling power generation system to realize the photovoltaic power generation smoothing method according to any one of claims 1-7.
9. A computer device, the computer device comprising:
one or more processors;
a memory for storing one or more programs;
when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the photovoltaic power smoothing method of any of claims 1-7.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the photovoltaic power smoothing method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310821202.1A CN116542882B (en) | 2023-07-06 | 2023-07-06 | Photovoltaic power generation smoothing method, system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310821202.1A CN116542882B (en) | 2023-07-06 | 2023-07-06 | Photovoltaic power generation smoothing method, system and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116542882A true CN116542882A (en) | 2023-08-04 |
| CN116542882B CN116542882B (en) | 2023-09-19 |
Family
ID=87445710
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310821202.1A Active CN116542882B (en) | 2023-07-06 | 2023-07-06 | Photovoltaic power generation smoothing method, system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116542882B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105631558A (en) * | 2016-03-22 | 2016-06-01 | 国家电网公司 | BP neural network photovoltaic power generation system power prediction method based on similar day |
| US20170091615A1 (en) * | 2015-09-28 | 2017-03-30 | Siemens Aktiengesellschaft | System and method for predicting power plant operational parameters utilizing artificial neural network deep learning methodologies |
| CN108364084A (en) * | 2017-11-08 | 2018-08-03 | 长沙理工大学 | A kind of BP neural network photovoltaic power prediction technique based on Theory of Information Fusion |
| CN108446799A (en) * | 2018-03-12 | 2018-08-24 | 浙江大学 | Waste pressure turbine device generated power forecasting method based on Elman neural networks |
| CN108665095A (en) * | 2018-04-27 | 2018-10-16 | 东华大学 | Short term power prediction technique based on genetic algorithm optimization Elman neural networks |
| CN109858673A (en) * | 2018-12-27 | 2019-06-07 | 南京工程学院 | A kind of photovoltaic generating system power forecasting method |
| US20200191117A1 (en) * | 2016-04-28 | 2020-06-18 | State Grid Corporation Of China | Adaptive dynamic planning control method and system for energy storage station, and storage medium |
| CN114597954A (en) * | 2022-05-10 | 2022-06-07 | 青岛科技大学 | Light storage grid-connected control system coping with photovoltaic power fluctuation |
| CN115756077A (en) * | 2022-11-25 | 2023-03-07 | 长园飞轮物联网技术(杭州)有限公司 | Maximum power tracking method, device and system of photovoltaic optimizer |
| CN115983095A (en) * | 2022-12-01 | 2023-04-18 | 国网湖北省电力有限公司电力科学研究院 | Photovoltaic power generation prediction method based on clustering algorithm, neural network and genetic algorithm |
| CN116307652A (en) * | 2023-05-25 | 2023-06-23 | 华北电力大学 | Artificial intelligent resource allocation method for intelligent power grid |
-
2023
- 2023-07-06 CN CN202310821202.1A patent/CN116542882B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170091615A1 (en) * | 2015-09-28 | 2017-03-30 | Siemens Aktiengesellschaft | System and method for predicting power plant operational parameters utilizing artificial neural network deep learning methodologies |
| CN105631558A (en) * | 2016-03-22 | 2016-06-01 | 国家电网公司 | BP neural network photovoltaic power generation system power prediction method based on similar day |
| US20200191117A1 (en) * | 2016-04-28 | 2020-06-18 | State Grid Corporation Of China | Adaptive dynamic planning control method and system for energy storage station, and storage medium |
| CN108364084A (en) * | 2017-11-08 | 2018-08-03 | 长沙理工大学 | A kind of BP neural network photovoltaic power prediction technique based on Theory of Information Fusion |
| CN108446799A (en) * | 2018-03-12 | 2018-08-24 | 浙江大学 | Waste pressure turbine device generated power forecasting method based on Elman neural networks |
| CN108665095A (en) * | 2018-04-27 | 2018-10-16 | 东华大学 | Short term power prediction technique based on genetic algorithm optimization Elman neural networks |
| CN109858673A (en) * | 2018-12-27 | 2019-06-07 | 南京工程学院 | A kind of photovoltaic generating system power forecasting method |
| CN114597954A (en) * | 2022-05-10 | 2022-06-07 | 青岛科技大学 | Light storage grid-connected control system coping with photovoltaic power fluctuation |
| CN115756077A (en) * | 2022-11-25 | 2023-03-07 | 长园飞轮物联网技术(杭州)有限公司 | Maximum power tracking method, device and system of photovoltaic optimizer |
| CN115983095A (en) * | 2022-12-01 | 2023-04-18 | 国网湖北省电力有限公司电力科学研究院 | Photovoltaic power generation prediction method based on clustering algorithm, neural network and genetic algorithm |
| CN116307652A (en) * | 2023-05-25 | 2023-06-23 | 华北电力大学 | Artificial intelligent resource allocation method for intelligent power grid |
Non-Patent Citations (3)
| Title |
|---|
| ZIHAN SUN SCHOOL OF ELECTRICAL AND ELECTRONIC ENGINEERING, SHANDONG UNIVERSITY OF TECHNOLOGY, LINYI, CHINA: "Wind power prediction based on Elman neural network model optimized by improved genetic algorithm", 《2021 IEEE 2ND INTERNATIONAL CONFERENCE ON BIG DATA, ARTIFICIAL INTELLIGENCE AND INTERNET OF THINGS ENGINEERING (ICBAIE)》, pages 413 - 416 * |
| 丁明;王磊;毕锐;: "基于改进BP神经网络的光伏发电系统输出功率短期预测模型", 电力系统保护与控制, no. 11, pages 99 - 105 * |
| 胡克用: "改进神经网络模型在光伏发电预测中的应用", 《计算机系统应用》, vol. 28, no. 12, pages 37 - 46 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116542882B (en) | 2023-09-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mellit et al. | Application of neural networks and genetic algorithms for sizing of photovoltaic systems | |
| CN109886473B (en) | Watershed wind-solar water system multi-objective optimization scheduling method considering downstream ecology | |
| CN108053061B (en) | Solar irradiance prediction method based on improved convolutional neural network | |
| CN109002948B (en) | CDA-BP-based microgrid short-term photovoltaic power generation power prediction method | |
| CN111367349A (en) | Photovoltaic MPPT control method and system based on prediction model | |
| CN112947672B (en) | Maximum power point tracking method and device for photovoltaic cell | |
| CN114784823A (en) | Micro-grid frequency control method and system based on depth certainty strategy gradient | |
| CN112330487B (en) | Photovoltaic power generation short-term power prediction method | |
| CN111709109A (en) | Photovoltaic absorption capacity calculation method and device considering source-load time sequence correlation | |
| CN105719101A (en) | Photovoltaic system power generation power prediction method | |
| CN112149883A (en) | Photovoltaic power prediction method based on FWA-BP neural network | |
| CN115986728A (en) | Power grid situation prediction method considering uncertainty factors and terminal | |
| Goh et al. | Hybrid SDS and WPT-IBBO-DNM Based Model for Ultra-short Term Photovoltaic Prediction | |
| CN116542882B (en) | Photovoltaic power generation smoothing method, system and storage medium | |
| Zhang et al. | Short-term load forecasting based on GAD-BP neural network | |
| CN116862050A (en) | Time sequence network-based daily prediction method, system, storage medium and equipment for carbon emission factors | |
| CN111476437A (en) | Method and device for predicting ultra-short term interval of photovoltaic power generation power | |
| CN113779861B (en) | Photovoltaic Power Prediction Method and Terminal Equipment | |
| CN113610665B (en) | Wind power generation power prediction method based on multi-delay output echo state network | |
| CN115545256A (en) | Small-scale photovoltaic power prediction method based on multi-dimensional data feature learning | |
| CN109165792B (en) | Photovoltaic short-term output power prediction method based on SOA-WNN | |
| CN111859780A (en) | Micro-grid operation optimization method and system | |
| Liu et al. | Location and Capacity Determination of Energy Storage System Based on Improved Whale Optimization Algorithm | |
| CN111079995A (en) | Power load nonlinear harmonic comprehensive prediction method, device and storage medium | |
| Mellit et al. | Application of neural networks and genetic algorithms for predicting the optimal sizing coefficient of photovoltaic supply (PVS) systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |