CN116307240A - Photovoltaic power station short-term power generation power prediction method based on WOA-VMD-OLS - Google Patents

Photovoltaic power station short-term power generation power prediction method based on WOA-VMD-OLS Download PDF

Info

Publication number
CN116307240A
CN116307240A CN202310353336.5A CN202310353336A CN116307240A CN 116307240 A CN116307240 A CN 116307240A CN 202310353336 A CN202310353336 A CN 202310353336A CN 116307240 A CN116307240 A CN 116307240A
Authority
CN
China
Prior art keywords
power
photovoltaic power
vmd
power station
data
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
Application number
CN202310353336.5A
Other languages
Chinese (zh)
Inventor
潘珍
林信
于明
包忠强
黄丽娟
黄飞鹏
覃晖
周恒旺
罗启登
郭华
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangxi Power Grid Co Ltd
Original Assignee
Guangxi Power Grid Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guangxi Power Grid Co Ltd filed Critical Guangxi Power Grid Co Ltd
Priority to CN202310353336.5A priority Critical patent/CN116307240A/en
Publication of CN116307240A publication Critical patent/CN116307240A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • 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
    • 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

Abstract

The invention discloses a WOA-VMD-OLS-based photovoltaic power station short-term power generation power prediction method. According to the invention, historical power generation power data of the power station are obtained and preprocessed to construct a data set for predicting the power generation power of the photovoltaic power station, parameters of the VMD are optimized through WOA to obtain an optimal decomposition modal number K and a punishment factor alpha, a power generation power sub-sequence of the photovoltaic power station is decomposed through the VMD, the data are screened out and are input as a prediction model, an OLS prediction model is established, and therefore a short-term power generation power prediction value of the photovoltaic power station is obtained. The WOA-VMD-OLS prediction model solves the problems that the prior art cannot adapt to photovoltaic historical data, meteorological data fluctuation is large, prediction accuracy is low, the prediction model is not stable enough and accurate and the like, can achieve good prediction effect under the condition of large power fluctuation, has good adaptability, and provides technical reference for coping with impact brought to a power grid after photovoltaic grid connection.

Description

Photovoltaic power station short-term power generation power prediction method based on WOA-VMD-OLS
[ field of technology ]
The invention relates to the technical field of photovoltaic power generation and grid connection, in particular to a short-term power generation power prediction method of a photovoltaic power station based on WOA-VMD-OLS.
[ background Art ]
The output of the power generated by the photovoltaic power station is influenced by environmental factors such as weather, so that the output power of the photovoltaic power station has obvious intermittent fluctuation characteristics. At present, more and more photovoltaic power generation starts grid-connected operation, a large-scale photovoltaic power generation access brings a certain impact to safe and stable operation of a power grid, and an effective and accurate photovoltaic power generation power prediction method can enable power supply grid dispatcher to obtain photovoltaic power generation output power in advance, so that a reasonable dispatching strategy is formulated to reduce the impact of the photovoltaic grid-connected operation on the power grid.
The conventional photovoltaic power station power generation power prediction method comprises a physical method, a statistical method, an artificial intelligence technology method and the like, such as a time sequence prediction method, an autoregressive moving average model method, a gray prediction method, a wavelet analysis method, a neural network method of a similar day selection algorithm and an artificial intelligence technology, a support vector machine method and the like. In recent years, improved algorithms or combination models based on data decomposition, intelligent optimization algorithms, neural networks, machine learning algorithms, and various methods. Wang S, wei L F, etc. process historical photovoltaic data using a variational modal decomposition algorithm (VMD), but the determination of VMD decomposition parameters is set according to human experience, which may result in undesirable decomposition effects and affect prediction accuracy. Wang Rui, high strength and the like effectively eliminate the problem of modal aliasing by using self-adaptive white noise complete integrated empirical mode decomposition (CEEMDAN), but the related parameters of the model are set according to manual experience, so that the prediction precision is not guaranteed.
The existing photovoltaic power station generation power prediction method has the following defects and problems: 1) The physical method takes weather forecast information data as input, researches the characteristics of the photovoltaic power generation equipment, establishes a corresponding mathematical model of the power generation power of the photovoltaic power generation station and the weather forecast information data, and further predicts the power generation power of the photovoltaic power generation station. 2) The statistical method uses the statistical relation among the historical measurement data to find out the internal law of the historical data and predict the generated power, and the method has higher requirements on the processing of the original data such as eliminating the stability of the pathological data points and the time sequence, and is difficult to reflect the influence of nonlinear factors. 3) In recent years, machine learning and deep learning methods based on artificial intelligence technology require a large number of data samples for model training. 4) The existing method has poor prediction accuracy on the data with large fluctuation changes of the power generation power, weather and the like of the historical photovoltaic power station. 5) For the existing method adopting data decomposition, the decomposition parameters are set by manual experience, so that excessive decomposition loss is easily caused, and the final prediction result is influenced.
Therefore, how to provide a photovoltaic power station short-term generation power prediction method and device for solving the above technical problems is a problem that needs to be solved by those skilled in the art.
[ invention ]
The invention provides a WOA-VMD-OLS-based photovoltaic power station short-term power generation power prediction method, which is used for solving the technical problems that the prior art cannot adapt to photovoltaic historical data, meteorological data fluctuation is large, prediction precision is low, and an existing prediction model is not stable enough and low in accuracy.
To achieve the object of the present invention, the present invention provides a method comprising the steps of:
a. acquiring historical power generation time series data and historical weather time series data of a photovoltaic power station, preprocessing missing data, and constructing a data set for predicting the power generation of the photovoltaic power station;
b. taking the signal difference average value SDA as an adaptability function, optimizing parameters of the VMD through WOA, and obtaining the optimal decomposition mode number K and penalty factor alpha of the VMD;
c. b, decomposing the optimal decomposition modal number K and the penalty factor alpha obtained in the step b through VMD to obtain a power generation power sub-sequence of the photovoltaic power station and each weather factor sub-sequence;
d. calculating a mutual information entropy value and a Hampel distance H between each power generation power subsequence of the photovoltaic power station and each weather factor subsequence, and selecting the weather factor subsequence meeting H & gt 3 as the input of a power generation power subsequence prediction model of the photovoltaic power station;
e. and establishing an OLS prediction model for each input power sub-sequence of the photovoltaic power station to obtain a predicted value of each power sub-sequence of the photovoltaic power station, and adding the predicted values of the power sub-sequences of the photovoltaic power station to obtain the predicted value of the short-term power of the photovoltaic power station.
In the step a, preprocessing measures of mean value supplementation of the data at the time points before and after the missing data are adopted, namely:
Figure BDA0004162502910000031
in the method, in the process of the invention,
Figure BDA0004162502910000032
data representing an i-th influence factor of the generated power of the photovoltaic power station at the t-th historical moment to be preprocessed; />
Figure BDA0004162502910000033
Data representing an i-th influence factor affecting the generated power of the photovoltaic power plant at the t-1 th historical moment; i epsilon {1,2, …, M }, namely M influence factors influencing the power generated by the photovoltaic power station; n is the original time sequenceColumn data length;
when other data remains unchanged, i.e
Figure BDA0004162502910000034
And when the power generation power prediction data set of the photovoltaic power station after pretreatment is as follows:
Figure BDA0004162502910000035
in the step b, the calculation formula of the signal difference average value SDA is as follows:
Figure BDA0004162502910000036
wherein i is {1,2, …, M }, namely each factor affecting the power generated by the photovoltaic power station;
Figure BDA0004162502910000037
generating a power sequence and a weather sequence for the pretreated photovoltaic power station; />
Figure BDA0004162502910000038
K modal components decomposed for the ith influence factor time series data; />
Figure BDA0004162502910000039
The sum of k modal components after decomposing the ith influence factor time series data; k is the number of decomposed layers; n is the time series data length.
Further, the method for optimizing the parameters of the WOA-VMD comprises the following steps:
b1, inputting historical power generation time series data and historical weather time series data of a photovoltaic power station, setting the range of parameters K and alpha in a VMD algorithm, and initializing various parameters in a WOA model;
b2, according to the position vector of whale population individual [ K alpha ]] 30×2 Sub-sequence data of generating power of photovoltaic power station
Figure BDA0004162502910000041
And weather factor subsequence data->
Figure BDA0004162502910000042
) VMD decomposition is performed;
b3, calculating each individual [ K.alpha. ] in the initial whale population] 30×2 Corresponding adaptation SDA value and recording the optimal individual position, and marking as Y * ＝[K * α * ]；
b4, updating the position of the whale individual;
b5, reserving the updated whale population position as a new round of initial population, and performing loop iteration until the set maximum iteration number Iter_max is reached;
b6, outputting optimal VMD decomposition parameters K and alpha.
In step b1, the input signal is: the preprocessed sub-sequence data of the power generation power of the photovoltaic power station
Figure BDA0004162502910000043
And weather factor subsequence data->
Figure BDA0004162502910000044
The range of the parameter K is K epsilon [2,10 ]]The range of the parameter alpha is alpha epsilon [1,2000 ]]The parameters in the initialized WOA model comprise whale group size pop_size=30, maximum iteration number Iter_max=200, variable number dim=2 and initialized whale group individual position vector [ K alpha ]] 30×2 。
In step b2, the step of decomposing the VMD includes:
b21 defining the kth eigenmode function of the ith time series as
Figure BDA0004162502910000045
It satisfies the following conditions:
Figure BDA0004162502910000046
in the method, in the process of the invention,
Figure BDA0004162502910000047
satisfy->
Figure BDA0004162502910000048
V t i,k For instantaneous amplitude +.>
Figure BDA0004162502910000049
For instantaneous phase, instantaneous frequency
Figure BDA00041625029100000410
b22, generating power sub-sequence data of photovoltaic power station
Figure BDA00041625029100000411
And weather factor subsequence data
Figure BDA0004162502910000051
An augmented lagrangian function L is constructed:
Figure BDA0004162502910000052
wherein t is time,
Figure BDA0004162502910000053
for the pre-processed photovoltaic power station power generation power time sequence and weather time sequence, { u } i,k The (i) represents k modal component sets obtained by VMD decomposition of the ith subsequence; { omega i,k The central frequency set of each modal component; alpha is a variation modal decomposition penalty factor; delta t Is a dirichlet function; lambda (lambda) i Is a Lagrangian operator;
solving by using a multiplier alternating direction algorithm ADMM, and finally carrying out iterative updating and transformation to obtain k modal component sets of the ith subsequence
Figure BDA0004162502910000054
In step b4, the updating the location of the individual includes the steps of:
b41 let z=1;
b42, randomly generating a value, wherein a = a 1 ×(2×rand-1)，
Figure BDA0004162502910000055
rand is a random number of (0, 1);
b43, when the absolute value of A is less than or equal to 1, randomly generating a p value between (0 and 1), and updating the position of the whale individual according to the following formula:
Figure BDA0004162502910000056
wherein: y is Y * ＝[K * α * ]Is the current target position; y (z) and Y (z+1) are expressed as whale positions at the z-th and z+1th iterations; d= |c·y * (z)-Y(z)|，C＝2·rand；D′＝|Y * (z) -Y (z) | represents the distance of whale to prey; b is a constant, defined as the shape of a logarithmic spiral; l is a linear control parameter such that the spiral shape is irregular in order to better search for the optimal solution and satisfy
Figure BDA0004162502910000057
Wherein (1)>
Figure BDA0004162502910000058
rand is a random number of (0, 1);
b44, when |a| >1, updating the position of the whale individual according to the following formula, and reserving the optimal fitness SDA and the corresponding parameter combinations K and α:
Y(z+1)＝Y rand (z)-A·D
wherein Y is rand For randomly selected whale positions, d= |c·y * (z)-Y(z)|，C＝2·rand，Y * ＝[K * α * ]Is the current target position; y (z) is expressed as the whale position at the z-th iteration.
In the step d, the calculation formula of the mutual information entropy value is as follows:
Figure BDA0004162502910000061
in the method, in the process of the invention,
Figure BDA0004162502910000062
k1=1, 2, …, K, the generated power modal component of the kth photovoltaic power plant obtained by VMD decomposition 1 ；
Figure BDA0004162502910000063
Ki=1, 2, … K for the ki modal component of the ith weather effect factor time series as obtained by VMD decomposition i ；
Figure BDA0004162502910000064
And->
Figure BDA0004162502910000065
And (5) entropy of edge information of the power generation power time sequence modal component and the weather factor subsequence modal component of the photovoltaic power station.
In the step d, the calculation formula of the Hampel distance H is as follows:
Figure BDA0004162502910000066
Figure BDA0004162502910000067
wherein H is a Hampel distance matrix; i is a mutual information entropy matrix; i 0.5 Is the median of I; and the k1 st photovoltaic power station power generation power modal component is recorded as R, and each influence factor modal component is recorded as
Figure BDA0004162502910000068
In step e, the formula of the OLS prediction model is:
Figure BDA0004162502910000069
in the method, in the process of the invention,
Figure BDA00041625029100000610
generating power for the photovoltaic power station in the t period of the predicted kth 1 photovoltaic power modal component; η (eta) r Weather influencing factors selected for the r-th +.>
Figure BDA00041625029100000611
Fitting coefficients of (a); />
Figure BDA00041625029100000612
The mode components of weather influence factors are screened out for the r-th period of the t period; c is a residual term;
wherein eta r The method meets the following conditions:
η r ＝(X T X) -1 X T Y；
in the method, in the process of the invention,
Figure BDA0004162502910000071
the short-term power generation predicted value of the photovoltaic power station is as follows:
Figure BDA0004162502910000072
in the method, in the process of the invention,
Figure BDA0004162502910000073
and (5) predicting the generated power of the photovoltaic power station for the t period.
The beneficial effects of the invention are as follows: according to the photovoltaic power station short-term power generation power prediction method based on WOA-VMD-OLS, fluctuation of photovoltaic power historical data and weather data is considered, and aiming at a photovoltaic power historical data subsequence and a weather data subsequence, the parameter decomposition mode number K and the penalty factor alpha of variation mode decomposition (Variational Mode Decomposition, VMD) are optimized through a whale optimization algorithm (Whale Optimization Algorithm, WOA), so that decomposition loss caused by manual experience setting can be reduced. Meanwhile, the photovoltaic power sequence and the weather factor sequence after VMD decomposition are more stable, and the accuracy of a subsequent prediction model is improved. In addition, mutual information entropy (Mutual Information Entropy, MIE) can deeply mine the relevance between the decomposed photovoltaic power and the weather factor subsequence, and the Hampel criterion is adopted for selecting the influence factors, so that accurate and effective input is selected, and the complexity of a prediction model is reduced. The WOA-VMD-OLS prediction model provided by the invention has a good prediction effect under the condition of large power fluctuation, has good adaptability, and provides technical reference for coping with impact on a power grid after grid connection of the photovoltaic.
[ description of the drawings ]
Fig. 1 is a flow chart of the method of the present invention.
FIG. 2 is a flow chart of the WOA-optimized VMD parameters of the present invention.
[ detailed description ] of the invention
The invention provides a short-term power generation power prediction method for a photovoltaic power station based on WOA-VMD-OLS, and the method is further described below with reference to the accompanying drawings and specific embodiments.
As shown in fig. 1, the short-term power generation power prediction method of the photovoltaic power plant based on WOA-VMD-OLS of the present invention comprises the steps of:
s10, acquiring historical power generation power time series data and historical weather time series data of the photovoltaic power station, preprocessing missing data, and constructing a data set for power generation power prediction of the photovoltaic power station.
In the step, the obtained historical generated power time series original data and the historical weather time original data set of the photovoltaic power station are as follows:
Figure BDA0004162502910000081
in the method, in the process of the invention,
Figure BDA0004162502910000082
generating power data of the photovoltaic power station at the t historical moment;
Figure BDA0004162502910000083
data representing an ith influence factor affecting the generated power of the photovoltaic power station at a tth historical time; i epsilon {1,2, …, M }, namely M influence factors influencing the power generated by the photovoltaic power station; n is the original time series data length. In this embodiment, taking 15 minutes as time-series data of a time interval, the influencing factors include historical power generation, temperature, wind speed, humidity, solar radiation and solar diffusion, and the obtained historical data set of the photovoltaic power station is shown in table 1. Of course, the time interval and the influencing factor may be changed according to actual requirements, and the application is not limited thereto.
Table 1 shows historical data of photovoltaic power station
Figure BDA0004162502910000084
Figure BDA0004162502910000091
Taking pretreatment measures of mean value supplementation of data at the front and rear time points of the missing acquired data, namely:
Figure BDA0004162502910000092
specifically, there is a deficiency (NaN in table 1) in the historical data of the factor affecting the solar power generation at the wind speed in table 1, and the presence of the deficiency value and the null value is unfavorable for the prediction accuracy, and the supplemental processing of replacing the deficiency value with the average value of the data at the time points before and after the deficiency value is adopted.
Figure BDA0004162502910000093
When other data remains unchanged, i.e
Figure BDA0004162502910000094
The preprocessed photovoltaic power plant generated power prediction data set is:
Figure BDA0004162502910000095
s20, optimizing parameters of the VMD through WOA to obtain the optimal decomposition modal number K and penalty factor alpha of the VMD.
In the step, the preprocessed sub-sequence data of the power generated by the photovoltaic power station
Figure BDA0004162502910000096
And weather factor subsequence data->
Figure BDA0004162502910000097
And determining the optimal decomposition modal number K of the VMD and the penalty factor alpha parameter value by using the signal difference average value SDA as an adaptability function and using WOA.
The calculation formula of SDA is as follows:
Figure BDA0004162502910000101
wherein i is {1,2, …, M }, namely each factor affecting the power generated by the photovoltaic power station;
Figure BDA0004162502910000102
generating a power sequence and a weather sequence for the pretreated photovoltaic power station; />
Figure BDA0004162502910000103
K modal components (IMFs) decomposed for the ith impact factor time series data; />
Figure BDA0004162502910000104
Then the sum of k modal components after the decomposition of the ith influence factor time series data is obtained; k is the number of decomposed modes; n is the time series data length.
As shown in fig. 2, the steps of WOA-VMD optimization decomposition are:
s201, inputting signals, setting parameters K and alpha ranges in a VMD algorithm, and initializing various parameters in a WOA model.
In this step, the input signal includes sub-sequence data of the generated power of the photovoltaic power plant
Figure BDA0004162502910000105
And weather factor subsequence data->
Figure BDA0004162502910000106
The range of the parameter K is K epsilon [2,10 ]]The range of the parameter alpha is alpha epsilon [1,2000 ]]. Parameters in the initialized WOA model include whale size pop_size=30, maximum number of iterations iter_max=200, number of variables (dim=2), and initialized whale individual position vector [ kα] 30×2 。
S202, according to the position vector [ K alpha ] of whale group individual] 30×2 Sub-sequence data of generating power of photovoltaic power station
Figure BDA0004162502910000107
And weather factor subsequence data->
Figure BDA0004162502910000108
VMD decomposition is performed.
In this step, the VMD decomposition step includes:
s2021 defining the kth eigenmode function of the ith time series as
Figure BDA0004162502910000109
The formula is:
Figure BDA00041625029100001010
in the method, in the process of the invention,
Figure BDA00041625029100001011
satisfy->
Figure BDA00041625029100001012
V t i,k For instantaneous amplitude +.>
Figure BDA00041625029100001013
For instantaneous phase, instantaneous frequency
Figure BDA00041625029100001014
S2022 is used for generating power sub-sequence data of photovoltaic power station
Figure BDA0004162502910000111
And weather factor subsequence data->
Figure BDA0004162502910000112
An augmented lagrangian function L is constructed:
Figure BDA0004162502910000113
wherein t is time,
Figure BDA0004162502910000114
for the pre-processed photovoltaic power station power generation power time sequence and weather time sequence, { u } i,k The (i) represents k modal component sets obtained by VMD decomposition of the ith subsequence; { omega i,k The central frequency set of each modal component; alpha is a variation modal decomposition penalty factor; delta t Is a dirichlet function; lambda (lambda) i Is a lagrangian.
Solving by using a multiplier alternating direction algorithm ADMM, and finally carrying out iterative updating and transformation to obtain k modal components of the ith subsequence
Figure BDA0004162502910000115
S203, calculating each individual [ K alpha ] in the initial whale group] 30×2 Corresponding adaptation SDA value and recording the optimal individual position, and marking as Y * ＝[K * α * ]Wherein, represent the optimum. In this embodiment, K and α corresponding to the minimum SDA value are selected as the optimal individual positions by comparing the SDA values in step S202.
S204, updating the position of the whale individual.
In the step, the specific steps of updating the individual positions of whales are as follows:
let z=1;
randomly generating a value, a=a 1 ×(2×rand-1)，
Figure BDA0004162502910000116
rand is a random number of (0, 1);
when the A is less than or equal to 1, randomly generating a p value between (0 and 1), and updating the position of the whale individual according to the following formula to obtain the following formula:
Figure BDA0004162502910000117
wherein: y is Y * ＝[K * α * ]Is the current target position; y (z) and Y (z+1) are expressed as whale positions at the z-th and z+1th iterations; d= |c·y * (z)-Y(z)|，C＝2·rand；D′＝|Y * (z) -Y (z) | represents the distance of whale to prey; b is a constant, defined as the shape of a logarithmic spiral; l is a linear control parameter, such that the spiral shape is irregular, in order to better search for the optimal solution,
Figure BDA0004162502910000121
rand is a random number of (0, 1).
When |A| >1, updating the position of the whale individual according to the following formula, reserving the optimal fitness SDA and the corresponding parameter combinations K and alpha,
Y(z+1)＝Y rand (z)-A·D
wherein Y is rand Is a randomly selected whale position.
S205, reserving the updated whale population position as a new initial population, and performing loop iteration until the set maximum iteration number Iter_max is reached.
S206, outputting optimal VMD decomposition parameters K and alpha.
In the step, the optimal whale individual is output, namely the optimal power generation power decomposition mode number K of the photovoltaic power station 1 Value, i weather influence factor time series optimal decomposition modal number K i And a corresponding penalty factor α.
Specific examples are shown in table 2. Table 2 shows the VMD decomposition weather factor results.
TABLE 2 weather factor decomposition results
Weather factor The number of modal components (IMFs) obtained by decomposition, i.e. the best decomposition K i Value of
Temperature (i=2) 5
Wind speed (i=3) 8
Humidity (i=4) 7
Solar radiation (i=5) 8
Solar diffusion (i=6) 9
S30, decomposing by utilizing the VMD to obtain a power generation power sub-sequence of the photovoltaic power station and each weather factor sub-sequence.
In the step, according to the optimal decomposition mode number K and the penalty factor alpha obtained in the step S20, the VMD decomposition is utilized to obtain the power generation sub-sequence of the photovoltaic power station
Figure BDA0004162502910000122
And the weather factor subsequences->
Figure BDA0004162502910000123
And S40, calculating a mutual information entropy value and a Hampel distance H between each power generation power subsequence of the photovoltaic power station and each weather factor subsequence, and selecting the weather factor subsequences meeting H & gt 3 as the input of a power generation power subsequence prediction model of the photovoltaic power station.
In the step, the calculation formula of the mutual information entropy value is as follows:
Figure BDA0004162502910000131
in the method, in the process of the invention,
Figure BDA0004162502910000132
k1=1, 2, …, K, the generated power modal component of the kth photovoltaic power plant obtained by VMD decomposition 1 ；/>
Figure BDA0004162502910000133
Ki=1, 2, … K for the ki modal component of the ith weather effect factor time series as obtained by VMD decomposition i ；
Figure BDA0004162502910000134
And->
Figure BDA0004162502910000135
Is photovoltaic power generationThe edge information entropy of the station generation power time sequence modal component and the weather factor sub-sequence modal component.
The calculation formula of Hampel distance H is as follows:
Figure BDA0004162502910000136
Figure BDA0004162502910000137
wherein H is a Hampel distance matrix; i is a mutual information entropy matrix; i 0.5 Is the median of I; and the k1 st photovoltaic power station power generation power modal component is recorded as R, and each influence factor modal component is recorded as
Figure BDA0004162502910000138
Specifically, taking table 3 as an example, it shows the influence factor screening results.
TABLE 3 influence factor screening results
Figure BDA0004162502910000139
Figure BDA0004162502910000141
S50, for each light Fu Gonglv sub-sequence, an OLS prediction model is established, a predicted value of each light Fu Gonglv sub-sequence is obtained, and then the predicted values of each light Fu Gonglv sub-sequence are added to obtain the predicted value of the short-term power generation power of the photovoltaic power station.
In this step, an OLS prediction model is built for each light Fu Gonglv sub-sequence and its corresponding input, and the formula of the model is:
Figure BDA0004162502910000142
in the method, in the process of the invention,
Figure BDA0004162502910000143
generating power for the photovoltaic power station in the t period of the predicted kth 1 photovoltaic power modal component; η (eta) r Weather influencing factors selected for the r-th +.>
Figure BDA0004162502910000144
Fitting coefficients of (a); />
Figure BDA0004162502910000145
The mode components of weather influence factors are screened out for the r-th period of the t period; c is a residual term;
wherein:
η r ＝(X T X) -1 X T Y；
in the method, in the process of the invention,
Figure BDA0004162502910000146
and superposing each group of prediction results to obtain the short-term power generation predicted value of the photovoltaic power station, wherein the predicted value is as follows:
Figure BDA0004162502910000147
in the method, in the process of the invention,
Figure BDA0004162502910000148
and (5) predicting the generated power of the photovoltaic power station for the t period.
According to the method and the device, under the condition that fluctuation of photovoltaic power historical data and weather data is considered, aiming at the photovoltaic power historical data subsequence and the weather data subsequence, the parameter decomposition mode number K and the penalty factor alpha of the VMD are optimized through WOA, so that decomposition loss caused by artificial experience setting can be reduced. Meanwhile, the photovoltaic power sequence and the weather factor sequence after VMD decomposition are more stable, and the accuracy of a subsequent prediction model is improved. In addition, the mutual information entropy value can deeply mine the relevance between the decomposed photovoltaic power and the weather factor subsequence, and the Hampel criterion is adopted for selecting the influence factors, so that accurate and effective input is selected, and the complexity of a prediction model is reduced. The WOA-VMD-OLS prediction model has good prediction effect under the condition of larger power fluctuation, has good adaptability, and provides technical reference for coping with impact brought to a power grid after grid connection of the photovoltaic.
Although the present invention has been disclosed by the above embodiments, the scope of the present invention is not limited thereto, and modifications, substitutions, etc. made to the above components will fall within the scope of the claims of the present invention without departing from the spirit of the present invention.

Claims (10)

1. A photovoltaic power station short-term generation power prediction method based on WOA-VMD-OLS is characterized by comprising the following steps:
a. acquiring historical power generation time series data and historical weather time series data of a photovoltaic power station, preprocessing missing data, and constructing a data set for predicting the power generation of the photovoltaic power station;
b. taking the signal difference average value SDA as an adaptability function, optimizing parameters of the VMD through WOA, and obtaining the optimal decomposition mode number K and penalty factor alpha of the VMD;
c. b, decomposing the optimal decomposition modal number K and the penalty factor alpha obtained in the step b through VMD to obtain a power generation power sub-sequence of the photovoltaic power station and each weather factor sub-sequence;
d. calculating mutual information entropy values and Hampel distances H between each subsequence of the power generation power of the photovoltaic power station and each subsequence of the weather factors, and selecting the subsequences of the weather factors meeting H & gt 3 as the input of a prediction model of the subsequences of the power generation power of the photovoltaic power station;
e. and establishing an OLS prediction model for each power generation sub-sequence of the photovoltaic power station to obtain a predicted value of the power generation sub-sequence of each photovoltaic power station, and adding the predicted values of the power generation sub-sequences of each photovoltaic power station to obtain the predicted value of the short-term power generation of the photovoltaic power station.
2. The WOA-VMD-OLS based photovoltaic power plant short-term generation power prediction method according to claim 1, wherein in step a, preprocessing measures of mean replenishment of the missing data at the previous and subsequent time points are taken, namely:
Figure FDA0004162502900000011
in the method, in the process of the invention,
Figure FDA0004162502900000012
data representing an i-th influence factor of the generated power of the photovoltaic power station at the t-th historical moment to be preprocessed; />
Figure FDA0004162502900000013
Data representing an i-th influence factor affecting the generated power of the photovoltaic power plant at the t-1 th historical moment; i epsilon {1,2, …, M }, namely M influence factors influencing the power generated by the photovoltaic power station; n is the original time sequence data length;
when other data remains unchanged, i.e
Figure FDA0004162502900000014
And when the power generation power prediction data set of the photovoltaic power station after pretreatment is as follows:
Figure FDA0004162502900000021
3. the WOA-VMD-OLS based short-term power generation prediction method of a photovoltaic power plant according to claim 2, wherein in step b, the calculation formula of the signal difference average SDA is:
Figure FDA0004162502900000022
wherein i is {1,2, …, M }, namely each factor affecting the power generated by the photovoltaic power station;
Figure FDA0004162502900000023
generating a power sequence and a weather factor sequence for the pretreated photovoltaic power station; />
Figure FDA0004162502900000024
K modal components decomposed for the ith influence factor time series data; />
Figure FDA0004162502900000025
The sum of k modal components after decomposing the ith influence factor time series data; k is the number of decomposed layers; n is the time series data length.
4. A WOA-VMD-OLS based photovoltaic power plant short term generation power prediction method according to claim 3, characterized in that the method of WOA-VMD optimizing parameters comprises the steps of:
b1, inputting historical power generation time series data and historical weather time series data of a photovoltaic power station, setting the range of K and alpha parameters in a VMD algorithm, and initializing various parameters in a WOA model;
b2, according to the position vector of whale population individual [ K alpha ]] 30×2 Sub-sequence data of generating power of photovoltaic power station
Figure FDA0004162502900000026
And weather factor subsequence data->
Figure FDA0004162502900000027
) VMD decomposition is performed;
b3, calculating each individual [ K.alpha. ] in the initial whale population] 30×2 Corresponding adaptation SDA value and recording the optimal individual position, and marking as Y * ＝[K * α * ]；
b4, updating the position of the whale individual;
b5, reserving the updated whale population position as a new round of initial population, and performing loop iteration until the set maximum iteration number Iter_max is reached;
b6, outputting optimal VMD decomposition parameters K and alpha.
5. The WOA-VMD-OLS based photovoltaic power plant short-term generation power prediction method according to claim 4, wherein in step b1, the input signal is: the preprocessed sub-sequence data of the power generation power of the photovoltaic power station
Figure FDA0004162502900000031
And weather factor subsequence data->
Figure FDA0004162502900000032
The range of the parameter K is K epsilon [2,10 ]]The range of the parameter alpha is alpha epsilon [1,200 ], and the parameters in the initialized WOA model comprise whale group size pop_size=30, maximum iteration number Iter_max=200, variable number dim=2 and initialized whale group individual position vector [ K alpha ]] 30×2 。
6. The WOA-VMD-OLS based photovoltaic power plant short-term generation power prediction method according to claim 4, wherein in step b2, the step of VMD decomposition comprises:
b21 defining the kth eigenmode function of the ith time series as
Figure FDA0004162502900000033
It satisfies the following conditions:
Figure FDA0004162502900000034
in the method, in the process of the invention,
Figure FDA0004162502900000035
satisfy->
Figure FDA0004162502900000036
Figure FDA0004162502900000037
For instantaneous amplitude +.>
Figure FDA0004162502900000038
For instantaneous phase, instantaneous frequency->
Figure FDA0004162502900000039
b22, generating power sub-sequence data of photovoltaic power station
Figure FDA00041625029000000310
And weather factor subsequence data
Figure FDA00041625029000000311
An augmented lagrangian function L is constructed:
Figure FDA00041625029000000312
wherein t is time,
Figure FDA00041625029000000313
for the pre-processed photovoltaic power station power generation power time sequence and weather time sequence, { u } i，k The (i) represents k modal component sets obtained by VMD decomposition of the ith subsequence; { omega i,k The central frequency set of each modal component; alpha is a variation modal decomposition penalty factor; delta t Is a dirichlet function; lambda (lambda) i Is a Lagrangian operator;
solving by using an ADMM (adm-m) algorithm, and finally performing iterative updating and transformation to obtain the firstK sets of modal components of i subsequences
Figure FDA0004162502900000041
7. The WOA-VMD-OLS based photovoltaic power plant short-term generation power prediction method according to claim 6, wherein in step b4, the updating the location of the individual comprises the steps of:
b41 let z=1;
b42, randomly generating a value, wherein a = a 1 ×(2×rand-1)，
Figure FDA0004162502900000042
rand is a random number of (0, 1);
b43, when the absolute value of A is less than or equal to 1, randomly generating a p value between (0 and 1), and updating the position of the whale individual according to the following formula to obtain the following formula:
Figure FDA0004162502900000043
wherein: y is Y * ＝[K * α * ]Is the current target position; y (z) and Y (z+1) are expressed as whale positions at the z-th and z+1th iterations; d= |c·y * (z)-Y(z)|，C＝2·rand；D′＝|Y * (z) -Y (z) | represents the distance of whale to prey; b is a constant, defined as the shape of a logarithmic spiral; l is a linear control parameter such that the spiral shape is irregular in order to better search for the optimal solution and satisfy
Figure FDA0004162502900000044
Wherein (1)>
Figure FDA0004162502900000045
rand is a random number of (0, 1);
b44, when |a| >1, updating the position of the whale individual according to the following formula, and reserving the optimal fitness SDA and the corresponding parameter combinations K and α:
Y(z+1)＝Y rand (z)-A·D
wherein Y is rand For randomly selected whale positions, d= |c·y * (z)-Y(z)|，C＝2·rand，Y * ＝[K * α * ]Is the current target position; y (z) is expressed as the whale position at the z-th iteration.
8. The WOA-VMD-OLS-based photovoltaic power plant short-term power generation prediction method according to claim 1, wherein in step d, the calculation formula of the mutual information entropy value is:
Figure FDA0004162502900000051
in the method, in the process of the invention,
Figure FDA0004162502900000052
the (k 1) th photovoltaic power station generated power modal component obtained by VMD decomposition>
Figure FDA0004162502900000053
A kth modal component obtained by VMD decomposition for the ith weather effect factor time series,/->
Figure FDA0004162502900000054
And->
Figure FDA0004162502900000055
And (5) entropy of edge information of the power generation power time sequence modal component and the weather factor subsequence modal component of the photovoltaic power station.
9. The WOA-VMD-OLS based photovoltaic power plant short-term power generation prediction method according to claim 1, wherein in step d, the calculation formula of the Hampel distance H is:
Figure FDA0004162502900000056
Figure FDA0004162502900000057
wherein H is a Hampel distance matrix; i is a mutual information entropy matrix; i 0.5 Is the median of I; and the k1 st photovoltaic power station power generation power modal component is recorded as R, and each influence factor modal component is recorded as
Figure FDA0004162502900000058
10. The WOA-VMD-OLS based photovoltaic power plant short-term generation power prediction method according to claim 1, wherein in step e, the formula of the OLS prediction model is:
Figure FDA0004162502900000059
in the method, in the process of the invention,
Figure FDA00041625029000000510
generating power for the photovoltaic power station in the t period of the predicted kth 1 photovoltaic power modal component; η (eta) r Weather influencing factors selected for the r-th +.>
Figure FDA00041625029000000511
Fitting coefficients of (a); />
Figure FDA00041625029000000512
The mode components of weather influence factors are screened out for the r-th period of the t period; c is a residual term;
wherein eta r The method meets the following conditions:
η r ＝(X T X) -1 X T Y；
in the method, in the process of the invention,
Figure FDA0004162502900000061
the short-term power generation predicted value of the photovoltaic power station is as follows:
Figure FDA0004162502900000062
in the method, in the process of the invention,
Figure FDA0004162502900000063
and (5) predicting the generated power of the photovoltaic power station for the t period.
CN202310353336.5A 2023-04-04 2023-04-04 Photovoltaic power station short-term power generation power prediction method based on WOA-VMD-OLS Pending CN116307240A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310353336.5A CN116307240A (en) 2023-04-04 2023-04-04 Photovoltaic power station short-term power generation power prediction method based on WOA-VMD-OLS

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310353336.5A CN116307240A (en) 2023-04-04 2023-04-04 Photovoltaic power station short-term power generation power prediction method based on WOA-VMD-OLS

Publications (1)

Publication Number Publication Date
CN116307240A true CN116307240A (en) 2023-06-23

Family

ID=86787039

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310353336.5A Pending CN116307240A (en) 2023-04-04 2023-04-04 Photovoltaic power station short-term power generation power prediction method based on WOA-VMD-OLS

Country Status (1)

Country Link
CN (1) CN116307240A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116738187A (en) * 2023-08-08 2023-09-12 山东航宇游艇发展有限公司 Ship gas power dynamic prediction method and system based on artificial intelligence
CN117498352A (en) * 2023-12-29 2024-02-02 西安热工研究院有限公司 Wind speed prediction method and device based on energy storage auxiliary black start capability

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116738187A (en) * 2023-08-08 2023-09-12 山东航宇游艇发展有限公司 Ship gas power dynamic prediction method and system based on artificial intelligence
CN116738187B (en) * 2023-08-08 2023-10-24 山东航宇游艇发展有限公司 Ship gas power dynamic prediction method and system based on artificial intelligence
CN117498352A (en) * 2023-12-29 2024-02-02 西安热工研究院有限公司 Wind speed prediction method and device based on energy storage auxiliary black start capability
CN117498352B (en) * 2023-12-29 2024-04-16 西安热工研究院有限公司 Wind speed prediction method and device based on energy storage auxiliary black start capability

Similar Documents

Publication Publication Date Title
CN110414045B (en) Short-term wind speed prediction method based on VMD-GRU
CN116307240A (en) Photovoltaic power station short-term power generation power prediction method based on WOA-VMD-OLS
CN110969290B (en) Runoff probability prediction method and system based on deep learning
CN111027775A (en) Step hydropower station generating capacity prediction method based on long-term and short-term memory network
CN110717610B (en) Wind power prediction method based on data mining
CN110648017A (en) Short-term impact load prediction method based on two-layer decomposition technology
CN111915092A (en) Ultra-short-term wind power prediction method based on long-time and short-time memory neural network
CN113344288B (en) Cascade hydropower station group water level prediction method and device and computer readable storage medium
CN114757427A (en) Autoregressive corrected LSTM intelligent wind power plant ultra-short term power prediction method
CN109242136A (en) A kind of micro-capacitance sensor wind power Chaos-Genetic-BP neural network prediction technique
CN114022311A (en) Comprehensive energy system data compensation method for generating countermeasure network based on time sequence condition
CN105956708A (en) Grey correlation time sequence based short-term wind speed forecasting method
CN115310669A (en) Ultra-short-term wind power prediction method based on secondary decomposition and IWOA-LSSVM
CN116702937A (en) Photovoltaic output day-ahead prediction method based on K-means mean value clustering and BP neural network optimization
CN115759465A (en) Wind power prediction method based on multi-target collaborative training and NWP implicit correction
Zhou et al. Short-term wind power prediction based on EMD-LSTM
Phan et al. Application of a new Transformer-based model and XGBoost to improve one-day-ahead solar power forecasts
CN115456286A (en) Short-term photovoltaic power prediction method
CN115860232A (en) Steam load prediction method, system, electronic device and medium
CN115907131A (en) Method and system for building electric heating load prediction model in northern area
CN115660038A (en) Multi-stage integrated short-term load prediction based on error factors and improved MOEA/D-SAS
CN115511657A (en) Wind power output and photovoltaic output evaluation method based on combined prediction model
CN115018156A (en) Short-term wind power prediction method
CN114897274A (en) Method and system for improving time sequence prediction effect
CN113112085A (en) New energy station power generation load prediction method based on BP neural network

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