CN116681154A - Photovoltaic power calculation method based on EMD-AO-DELM - Google Patents

Abstract

The invention discloses a photovoltaic power calculation method based on EMD-AO-DELM, including S1: obtaining meteorological factors, the meteorological factors include cloud amount, air temperature, air pressure, humidity and total radiation, and calculating the total radiation, cloud amount and photovoltaic power It is highly positively correlated, and relative humidity and atmospheric pressure are negatively correlated, so cloud cover and total radiation are selected as the initial input data of DELM; S2: Establish the calculation model of AO-DELM, and use the initial input weight of DELM as the initial population of the AO algorithm location, and set the fitness function as the sum of the mean square errors of the training set and the test set; S3: Establish a calculation model of EMD‑AO‑DELM, use EMD to decompose the photovoltaic power curve, so that the original environmental signal The fluctuations or trends of different scales are decomposed step by step, and the AO-DELM modeling analysis is performed on the decomposed IMF components; S4: verify the validity and accuracy of the calculation model; the present invention realizes budget photovoltaic by inputting fewer parameters. Power is more precise.

Description

A photovoltaic power calculation method based on EMD-AO-DELM

technical field

The invention belongs to the technical field of photovoltaic power calculation, and in particular relates to a photovoltaic power calculation method based on EMD-AO-DELM.

Background technique

As a new energy power generation, photovoltaic power generation is widely used at present. However, the power output of photovoltaic power generation is affected by various factors, including environmental factors such as solar radiation intensity, temperature, humidity and cloud cover. Changes in these factors may lead to fluctuations in the power output of the photovoltaic power system, which will adversely affect the security and stability of the power system, and also bring challenges to the photovoltaic grid-connected scheduling process.

The prediction technology for photovoltaic power generation is mainly divided into direct prediction method and indirect prediction method. The former is to train and learn photovoltaic historical data information, and predict future power through prediction algorithms. The latter adopts a step-by-step forecast method, which can be divided into two parts: future solar radiation and future power prediction. The comparative document discloses a short-term power prediction method of photovoltaic power plants based on EMD and ELM, and proposes a combined power prediction method based on empirical mode decomposition (EMD) and extreme learning machine (ELM), but fails to fully consider the output power of photovoltaic power plants. environmental factors. The prior art proposes an improved BP neural network photovoltaic prediction method based on the Fireworks Algorithm (FWA), which has good accuracy for short-term photovoltaic data prediction. However, the BP neural network has the problems of slow convergence and easy to fall into local extremum. The existing technology discloses a photovoltaic power prediction method based on CNN-BiLSTM, and proposes a prediction algorithm combining convolutional neural network (CNN) and bidirectional long short-term memory network (BILSTM), but the test sample type of this method is too single , failing to test the prediction accuracy under different seasons and weather conditions.

There are many factors affecting the efficiency of photovoltaic power generation, which are mainly divided into subjective factors and objective factors. Among them, subjective factors include the model parameters of photovoltaic array panels, inclination and orientation of photovoltaic panels, etc.; objective factors include uncontrollable meteorological factors such as temperature, humidity, cloud cover, precipitation, and irradiance of light, and meteorological factors often play a decisive role. . Too many input sample factors will reduce the prediction accuracy and make the prediction model complex and redundant.

Contents of the invention

The purpose of the present invention is to provide a photovoltaic power calculation method based on EMD-AO-DELM in order to solve the problems that too many input sample factors in the existing photovoltaic power prediction method will reduce the prediction accuracy and make the prediction model complex and redundant.

The present invention achieves the above object through the following technical solutions: comprising the following steps:

S1: Obtain meteorological factors. Meteorological factors include cloud amount, air temperature, air pressure, humidity and total radiation. It is calculated that total radiation and cloud amount are highly positively correlated with photovoltaic power, and relative humidity and atmospheric pressure are negatively correlated. Therefore, cloud amount and total These two items of radiation are used as the initial input data of DELM;

S2: Establish the calculation model of AO-DELM, use the initial input weight of DELM as the initial population position of the AO algorithm, and set the fitness function as the sum of the mean square errors of the training set and the test set;

S3: Establish the calculation model of EMD-AO-DELM, use EMD to decompose the photovoltaic power generation curve, so as to decompose the fluctuations or trends of different scales in the original environmental signal step by step, and perform AO-DELM on the decomposed IMF components respectively DELM modeling analysis, and then the prediction results of each IMF component are superimposed and summed to obtain the final prediction value;

S4: Verify the validity and accuracy of the calculation model.

Further, said step S1 includes: S11: introducing Pearson (Pearson correlation coefficient) correlation coefficient, by formula It is calculated to measure whether the two data sets are on the same line, and to measure the linear relationship between the fixed-distance variables. Among them, r>1 means that there is a positive correlation between the two, and r<1 means that there is a positive correlation between the two Negative correlation; xi and yi respectively represent the i-th value of the two factors; /> and /> Represents the mean of the two factors, respectively.

Further, the step S2 includes: the calculation formula is fitness=MSE(train)+MSE(test).

Further, the step S2 includes: S21: Perform data cleaning, and propose abnormal values caused by some sampling errors in the historical photovoltaic power data; S22: Perform normalization processing on the cleaned sample data; S23: Initialize the parameters of the AO algorithm, including the population size, the maximum number of iterations T, the exploration and development parameters ɑ, δ; S24: initialize the population position X, the initial population fitness, and the best individual; S25: expand the exploration stage and shrink the exploration in order stage, expand the development stage, shrink the development stage, and continuously update the population position; S26: calculate the fitness of the updated population, obtain the current best individual position and fitness, and compare the current best individual with the best found in the t generation Individual fitness, retaining a better individual position; S27: Judging whether the maximum number of iterations or the solution condition is reached, if yes, then output the optimal value, if not, return to step S25; S28: Input the final optimized weight value result into the DELM model.

Further, the step S3 includes: S31: use EMD to decompose the historical photovoltaic data to obtain a set of IMF components; S32: establish an AO-DELM model for each IMF component, and predict each component; S33: superimpose each The prediction results of the subsequences and verify the accuracy of the model predictions.

Further, the step S4 includes: using both MAPE and RMSE as error indicators, and the calculation formula of MAPE is RMSE calculation formula is /> Among them, yoi is the i-th real value in the sample, and ypi is the i-th predicted value in the sample. The smaller the two values, the higher the accuracy.

Beneficial effects: the invention has reasonable design, simple and stable structure, strong practicability, and has the following beneficial effects:

1. Among the influencing factors of photovoltaic power generation, the total solar radiation and cloud cover are positively correlated with photovoltaic power, which play a key role in the final prediction results; air pressure and humidity are negatively correlated with photovoltaic power, which is not suitable for actual power prediction. as input data;

2. In view of the volatility and randomness of photovoltaic power, the historical photovoltaic power data is decomposed by EMD. predict better;

3. The prediction performance of the method in this paper is better than that of AO-DELM and DELM models in the four seasons of a year. Among them, the prediction accuracy is the highest in the two quarters of S2 and S3, which is related to the local weather conditions, and the weather is relatively stable. Seasons with a high proportion of sunny days have higher forecast accuracy.

Description of drawings

Fig. 1 is the structural representation of ELM of the present invention;

Fig. 2 is the structural representation of ELM-AE of the present invention;

Fig. 3 is the structure schematic diagram of DELM of the present invention;

Fig. 4 is a flow chart of AO-DELM structure of the present invention;

Fig. 5 is a schematic diagram of the power EMD decomposition sequence in the first quarter of the present invention;

Fig. 6 is the combined prediction flowchart of EMD and AO-DELM of the present invention;

Fig. 7 is the power prediction result of the present invention in different quarters;

Fig. 8 is the prediction error of each algorithm in the first quarter of the present invention;

Fig. 9 is a flowchart of the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention.

Embodiment one:

Empirical Mode Decomposition (EMD) is a signal decomposition method based on the local characteristics of the signal.

It absorbs the multi-resolution advantages of wavelet transform and overcomes the difficulties of selecting wavelet base and determining the decomposition scale in wavelet transform, so it is more suitable for the analysis of nonlinear and non-stationary signals, and it is an adaptive signal decomposition method. EMD assumes that any complex signal is composed of simple eigenmode functions (IMFs). And each IMF component is independent of each other. EMD can decompose time series data of different scales or trends, decompose it into its components step by step, and generate a series of data sequences with the same scale characteristics from a series of data sequences with the same scale characteristics. Through these data sequences, non-stationary nonlinear The data were transformed into stationary linear data. Compared with the original data sequence, the decomposed sequence has higher regularity. This is of great help in identifying hidden relationships, which can improve the accuracy of predictions.

Its specific decomposition steps are as follows:

For the initial time series x(t), take all its maximum and minimum points, connect all the maximum values as the upper envelope, and connect all the minimum points as the lower envelope, record m(t) as The mean of the upper and lower envelopes. Subtract the original sequence x(t) from the mean value m(t) to obtain the first component h1(t)=x(t)-m(t);

Treat h1(t) as the initial time series, record m1(t) as the mean value of the upper and lower envelopes of h1(t), repeat step (1), and obtain the second component h2(t);

Repeat the above steps n times until hn(t) is an eigenmode function or the remaining component rn(t) is monotonic, and the decomposition process is terminated;

So far, the initial time series x(t) can be expressed by the sum of n eigenmode components hi(t) and a residual component rn(t), the formula is:

Principle of Tianying Algorithm: The mathematical model is briefly described as follows:

Step 1: Expand Exploration

In this method, the Aquila bird flies high above the ground, extensively exploring the hunting space, and then resorts to a vertical dive once the Aquila bird has identified the area of its prey. The mathematical representation of this behavior is written as:

Among them, Xbest(t) represents the best position obtained so far, XM(t) represents the average position of all Aquila birds in the current iteration, and t and T are the current iteration and the maximum number of iterations, respectively. N is the population size and rand is a random number between 0 and 1.

Step 2: Narrowing down the exploration phase

This is the most common hunting method used by Aquila birds. After it descends in a selected area and flies around the prey, it uses a short-distance gliding method to attack the prey. The location update formula is expressed as:

X(t+1)＝X _best (t)×LF(D)+X _R (t)+(yx)×rand

Among them, XR(t) represents the random position of the Aquila bird, D is the dimension size, and LF(D) represents the Levy flight function, which is expressed as follows:

Among them, s and β are constants equal to 0.01 and 1.5 respectively, and u and v are random numbers between 0 and 1. y and x are used to render the spiral in the search and are calculated as follows:

Among them, r1 refers to the number of search cycles between 1 and 20, D1 is composed of integers from 1 to dimension size D, and ω is equal to 0.005.

Step 3: Expanding the Development Phase

In the third stage, when the area of prey has been roughly determined, the Aquila bird descends vertically for an initial attack. AO utilizes selected areas to approach and attack prey. This behavior manifests itself as follows:

X(t+1)＝(X _best (t)-X _M (t))×α-rand+((U _B -L _B )×rand+L _B )×δ

where α and δ are development adjustment parameters fixed at 0.1, and UB and LB are upper and lower bounds.

Step 4: Narrowing down the development phase

In this method, the Aquila bird chases the prey according to the trajectory of the prey's escape, and then attacks the prey on the ground. The mathematical representation of this behavior is as follows:

where X(t) is the current position and QF(t) represents the quality function value used to balance the search strategy. G1 represents the motion parameter of the eagle in the process of tracking the prey, which is a random number between [-1, 1]. G2 represents the flight slope when chasing prey, decreasing linearly from 2 to 0.

ELM:

The extreme learning machine (Extreme Learning Machine, ELM) model consists of three parts, namely: input layer, hidden layer and output layer, which is a typical (Single-hidden Layer Feed-forward NeuralNetwork, SLFN) single hidden layer The feed-forward neural network has the advantages of fast learning speed and strong generalization ability.

The model structure of ELM is shown in Figure 1, where the input layer contains q nodes, the hidden layer contains n nodes, the output layer contains e nodes, and the activation function of the hidden layer is g(x). The commonly used function is Sigmoid , Hard-lim, Sin, etc.

Suppose the samples are xi∈RN×Rq, yi∈RN×Re(i=1,2,...,N), where the output of the hidden layer is formula (10), the output matrix of the hidden layer and the output of the ELM network The relationship between them can be expressed by formula (11).

h=g(ax+b)

h(xi)V=yi, i=1,2,...,N

in,

Among them, a _i =[a _i1 ,a _i2 ,…,a _in ] ^T is the weight connecting the i-th input node and the hidden layer, bj is the threshold of the j-th hidden node, v _j =[v _j1 ,v _j2 ,...v _jn ] ^T is the weight connecting the jth hidden node to the output layer. H is the hidden layer output matrix of the neural network. The input weight aij and the threshold bj of the hidden layer are randomly selected; the output weight V can be obtained by solving the equation system.

Obtaining output weights using ELM can be divided into three steps.

Randomly select a value between 0 and 1 to set the input weight aij and the threshold bj of the hidden layer;

Calculate the hidden layer output matrix H; calculate the output weight V=H ⁺ Y.

where H+ represents the generalized inverse matrix of the output matrix H.

Different from the traditional gradient-based feed-forward neural network algorithm, the hidden layer of the extreme learning machine network randomly generates input weights and thresholds during the training process. Therefore, the calculation of the output weight can only use the generalized inverse matrix theory. However, ELM is a single hidden layer structure, and its ability to capture the effective characteristics of the data is insufficient when facing input data with a large amount of data and high dimensions. Therefore, more scholars use the DELM algorithm, as a derivative algorithm of ELM, to solve the problem that the extreme learning machine with only one hidden layer cannot capture the effective characteristics of the data.

Extreme Learning Machine-Autoencoder (ELM-AE):

Autoencoder (ELM-AE) is an artificial neural network module commonly used in the field of deep learning, which is an unsupervised way to learn the structure of samples. Its main feature is that the output of the network is consistent with the input results. The ELM-AE model, like ELM, also consists of an input layer, a hidden layer and an output layer. Its model structure is shown in Figure 2. The weights and thresholds of the hidden layer nodes of the constructed ELM-AE are randomly generated during the training process, and have orthogonality, so that the generalization ability of the ELM-AE has been optimized to a certain extent. . In order to further improve the generalization ability and robustness of the model, regularization parameters are introduced in the process of solving the weight coefficients. The objective function is set as:

Suppose given N different samples, xi∈Rn×Rq(i=1, 2, N), the output of ELM-AE hidden layer can be expressed as formula h=g(ax+b), then the hidden layer The mathematical relationship between the output matrix and the output of the output layer can be expressed as Among them, i=1,2,...,N, for the equal-dimensional ELM-AE representation, the calculation method of the output weight V is: V=H ^{- 1} X where, H is the output matrix of the ELM-AE hidden layer, X are the input and output matrices of the ELM-AE.

Deep extreme learning machine (DELM) builds a multi-layer network structure by stacking extreme learning machine-autoencoder (ELM-AE) to improve the expressive ability of the network. It is a new structure combining extreme learning machine and automatic encoder.

DELM applies ELM-AE to train the model layer by layer. The numerical relationship between the output of the i hidden layer and the output of the (i-1) hidden layer can be expressed by the following formula: H _i =g((v _i ) ^T H _i-1 )

DELM (Deep Extreme Learning Machine):

ELM-AE is used to construct the basic unit of deep extreme learning machine DELM, and then use the output weight of ELM-AE to initialize the whole DELM. The idea of DELM is to make the output infinitely close to the original input by minimizing the reconstruction error, and to learn the advanced features of the original data through layer-by-layer iterative training.

ELM-AE maps the input to hidden layer feature vectors at the encoder, and reconstructs the original input from the feature vectors at the decoder. From a structural point of view, DELM is equivalent to connecting multiple ELMs. Compared with ELM, DELM can more comprehensively capture sample features and improve the accuracy of processing high-dimensional input. DELM performs unsupervised training and learning layer by layer through ELM-AE, and finally connects to the last output layer for supervised training. The parameters of the system do not need to be adjusted simultaneously. The structure of the DELM network is shown in Figure 3. The input weights of each hidden layer of DELM are initialized by ELM-AE, and hierarchical unsupervised training is performed. During this whole process, DELM does not require reverse fine-tuning.

Assuming that the model has Y hidden layers, according to the ELM-AE theory mentioned above, the weight matrix V1 can be obtained through the input data X, and then the output matrix H1 of the hidden layer can be obtained. Then H1 is used as the input and target output of the next ELM-AE. By analogy with layer-by-layer training, the output weight matrix VY of the Y layer and the output matrix HY of the hidden layer can be obtained. The output weight of each ELM-AE is used to initialize the entire DELM. In the ELM-AE training process, the input layer weights and thresholds are randomly generated orthogonal random matrices; at the same time, the ELM-AE unsupervised training process uses the least squares method to update parameters. In this process, only the weight parameters of the output layer are updated, while the weights and thresholds of the input layer remain unchanged, and the influence of random input weights and random thresholds of each ELM-AE will affect the prediction accuracy of DELM. Since the initial weight plays a more critical role in the prediction results of the entire model. Therefore, this paper uses the AO algorithm to optimize the input weight of DELM.

In this paper, the global optimization ability of Tianying optimization algorithm can be used to find the input weight of deep extreme learning machine when the training error is small, so as to improve the generalization ability of deep extreme learning machine and improve the prediction accuracy of DELM.

Factor selection:

As shown in Figure 4-9, there are many factors affecting the efficiency of photovoltaic power generation, which are mainly divided into subjective factors and objective factors. Among them, subjective factors include the model parameters of photovoltaic array panels, inclination and orientation of photovoltaic panels, etc.; objective factors include uncontrollable meteorological factors such as temperature, humidity, cloud cover, precipitation, and irradiance of light, and meteorological factors often play a decisive role. . Too many input sample factors will reduce the prediction accuracy and make the prediction model complex and redundant. In order to explore the correlation between meteorological factors and photovoltaic power, in order to select the optimal factor as input, the Pearson correlation coefficient is introduced here, which is used to measure whether the two data sets are on a line, and it is used to measure the linearity between fixed-distance variables relation.

Among them, r>1 means that there is a positive correlation between the two, and r<1 means that there is a negative correlation between the two; xi and yi respectively represent the i-th value of the two factors; and /> Represents the mean of the two factors, respectively.

According to the Pearson correlation coefficient, sort out the following table.

It can be seen from the table that the total radiation and cloud amount are highly positively correlated with photovoltaic power, and the relative humidity and atmospheric pressure are negatively correlated. Therefore, cloud amount and total radiation are selected as the initial input data of DELM.

Establishment of AO-DELM model:

The main idea of the AO-DELM model is: use the initial input weight of DELM as the initial population position of the AO algorithm; and set the fitness function as the sum of the mean square errors of the training set and the test set, which is expressed as follows:

fitness=MSE(train)+MSE(test)

The process of AO-DELM prediction model is as follows:

Carry out data cleaning, eliminate outliers caused by some sampling errors in historical photovoltaic power data, and normalize the cleaned sample data;

Initialize the parameters of the AO algorithm, including the population size, the maximum number of iterations T, the exploration and development parameters ɑ, δ;

Initialize the population position X, the initial population fitness, and the best individual;

Expand the exploration phase, shrink the exploration phase, expand the development phase, and shrink the development phase in sequence, and continuously update the population position;

Calculate the fitness of the updated population, obtain the current best individual position and fitness, and compare the current best individual with the best individual fitness found in the tth generation, and retain the better individual position;

Determine whether the maximum number of iterations or the solution condition is reached, if yes, output the optimal value, if not, return to step 5;

Input the final optimized weight value results into the DELM model, and its structure flow chart is shown in Figure 4

Establishment of EMD-AO-DELM model

Photovoltaic power data is nonlinear and non-stationary discrete data. The traditional linear time-series model method has great limitations, and direct predictive modeling of it has a large error. Therefore, EMD is used to decompose the power curve of photovoltaic power generation, so as to decompose the fluctuations or trends of different scales in the original environmental signal step by step. The AO-DELM modeling analysis is carried out on the decomposed IMF components, and then the prediction results of each IMF component are superimposed and summed to obtain the final prediction value.

The specific steps are:

Using EMD to decompose the photovoltaic historical data to obtain a set of IMF components;

Establish the AO-DELM model for each IMF component to predict each component;

Superimpose the prediction results of each subsequence and verify the accuracy of the model prediction;

The EMD-AO-DELM prediction model is shown in Fig. 6.

The EMD method is used to decompose the historical photovoltaic data into four IMF components with different characteristics and a residual Res. The IMF sequences of the photovoltaic power in the first quarter are shown in Figure 5. The IMF component can reflect the local characteristics of the original data, better reflect its periodic items, random items and trend items, and accurately reflect the characteristics of the original data.

Among them, IMF1-IMF2 presents an unstable and oscillating curve, which belongs to the random item; IMF3-IMF4 presents a smooth, decreasing frequency, and periodic trend, which belongs to the trend item. Therefore, EMD decomposition can highlight the local characteristics of the original photovoltaic power sequence.

Evaluation indicators:

In order to verify the validity and accuracy of this forecasting model, both MAPE and RMSE are used as error indicators.

Among them, yoi is the i-th actual value (observed) in the sample, and ypi is the i-th predicted value (predicted) in the sample. The smaller the two values, the higher the accuracy.

Embodiment 2: A photovoltaic power calculation method based on EMD-AO-DELM in this embodiment 2 is an improvement on the basis of the above-mentioned embodiment. The technical content disclosed in the above-mentioned embodiment will not be described repeatedly. The disclosed in the above-mentioned embodiment The content also belongs to the content disclosed in the second embodiment

As shown in Figures 4-9, an embodiment of the present invention: a photovoltaic power calculation method based on EMD-AO-DELM, including the following steps: S1: abnormal data cleaning: during the actual sampling process of photovoltaic power plants, some abnormal data. Abnormal data will lead to poor fitting of the predictive model and weakened generalization ability. Therefore, it must be cleaned.

The detection of abnormal data adopts the principle of 3σ criterion: where X represents the initial data of photovoltaic power, Indicates the average value of the initial PV power data. In statistics, the 3σ criterion is the percentage within a normal distribution that is less than one standard deviation, two standard deviations, and three standard deviations from the mean. The more precise figures are 68.27%, 95.45%, and 99.73%.

Data normalization: The range of the same data unit is different, which affects the fitting speed of the model and is not conducive to model training. In order to improve the prediction accuracy of the model, the initial photovoltaic power data is normalized, and the normalized power value is kept between [0, 1]. Among them, under the normalization formula: Among them, yi represents the normalized data, and xi represents the original power data value.

In this embodiment, the simulation results: Matlab R2022a is used to simulate and analyze the proposed control strategy. In the prediction model established in this paper, the number of hidden layers of the DELM model is set to 2; the number of hidden layer nodes is 5, 5 respectively ; The maximum number of iterations of AO is 200; the population size is 20; the regularization coefficient is set to infinity. For the sake of fairness, the parameter hidden layer of the DELM model used as a comparison is 2; the number of hidden layer nodes is 5 and 5 respectively. Divide the annual data into four quarters on average, divide the data of each quarter into a training set and a test set at a rate of 19:1, and perform normalization processing. The simulation verification is carried out for each quarter separately, and the test set is taken from the time period with sunshine from 8:00 to 19:00 every day, with a step size of 1 hour. Among them, the prediction results from the first quarter to the fourth quarter are shown in Figure 7.

In this embodiment shown in conjunction with Fig. 1-4:

Based on Figure 7 and the above table, the prediction accuracy is the highest in the two quarters of S2 and S3, mainly because the sunshine is relatively stable in the two quarters of S2 and S3, and there are mostly sunny days. However, the average solar radiation is lower than that of the S1 season, resulting in a lower overall power generation peak value than that of the S1 season. The relatively poor prediction accuracy of S1 and S2 is due to the unstable weather in these two quarters, with more showers and cloudy weather, especially in the S2 season. The photovoltaic power in the four quarters showed a typical normal distribution, which was in line with the actual power situation, and the forecast model was accurate.

In summary, through the simulation diagram and the error analysis in Figure 8, the EMD-AO-DELM model mentioned in this paper has significantly improved prediction accuracy and better model stability compared with the initial DELM model and AO-DELM model. The index is obviously better than the other two algorithms. It can be competent for actual photovoltaic power forecasting requirements and better cooperate with photovoltaic grid-connected dispatching work.

In this example, the forecast situation of photovoltaic power generation in four quarters of a year is analyzed, the correlation analysis is carried out on the analysis of the influencing factors of photovoltaic power generation, and the EMD decomposition of the historical photovoltaic power generation power is carried out, and each IMF component is input into AO-DELM model, and finally sum the results of each component to get the prediction result. Furthermore, a photovoltaic power prediction model based on EMD-AO-DELM is proposed. Through the analysis of simulation results, the following conclusions are obtained:

Among the influencing factors of photovoltaic power generation, the total solar radiation and cloud cover are positively correlated with photovoltaic power, which play a key role in the final prediction results; air pressure and humidity are negatively correlated with photovoltaic power, and should not be used as inputs in actual power prediction data.

Aiming at the characteristics of fluctuation and randomness of photovoltaic power, the historical photovoltaic power data is decomposed by EMD, and each component is independent of each other, respectively predicted, and finally superimposed and summed. The experiment proves that the prediction effect is better after adopting the EMD decomposition method.

The prediction performance of the method in this paper is better than that of the AO-DELM and DELM models in the four quarters of the year. Among them, the prediction accuracy is the highest in the two seasons of S2 and S3, which is related to the local weather conditions. The weather is relatively stable, and the forecast accuracy is higher in quarters with a high proportion of sunny days.

It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Accordingly, the embodiments should be regarded in all points of view as exemplary and not restrictive, the scope of the invention being defined by the appended claims rather than the foregoing description, and it is therefore intended that the scope of the invention be defined by the appended claims rather than by the foregoing description. All changes within the meaning and range of equivalents of the elements are embraced in the present invention. Any reference sign in a claim should not be construed as limiting the claim concerned.

In addition, it should be understood that although this specification is described according to implementation modes, not each implementation mode only includes an independent technical solution, and this description in the specification is only for clarity, and those skilled in the art should take the specification as a whole , the technical solutions in the various embodiments can also be properly combined to form other implementations that can be understood by those skilled in the art.

Claims (6)

1. A photovoltaic power calculation method based on EMD-AO-DELM is characterized in that: the method comprises the following steps:

s1: the method comprises the steps of obtaining meteorological factors, wherein the meteorological factors comprise cloud cover, air temperature, air pressure, humidity and total radiation, calculating that the total radiation and the cloud cover are highly positively correlated with photovoltaic power, and the relative humidity and the atmospheric pressure are negatively correlated, so that the cloud cover and the total radiation are selected as DELM initial input data;

s2: establishing an AO-DELM calculation model, taking the DELM initial input weight as an initial population position of an AO algorithm, and setting an fitness function as the sum of mean square errors of a training set and a test set;

s3: establishing an EMD-AO-DELM calculation model, decomposing a photovoltaic power generation power curve by adopting the EMD, so as to decompose different scale fluctuation or trend existing in an original environment signal step by step, respectively carrying out AO-DELM modeling analysis on decomposed IMF components, and then carrying out superposition summation on prediction results of the IMF components to obtain a final prediction value;

s4: and verifying the validity and accuracy of the calculation model.

2. The EMD-AO-DELM based photovoltaic power calculation method according to claim 1, wherein: the step S1 includes: s11: introducing Pearson correlation coefficient through the formulaCalculating a linear relationship between distance variables for measuring whether two data sets are on a line, wherein r>1 represents positive correlation between the two, r<1 represents that the two are in negative correlation; xi and yi represent the values of the ith of two factors, respectively; />And->Each representing an average of 2 factors.

3. The EMD-AO-DELM based photovoltaic power calculation method according to claim 1, wherein: the step S2 includes: the calculation formula is fitness=mse (train) +mse (test).

4. A method of EMD-AO-DELM based photovoltaic power calculation according to claim 3, wherein: the step S2 includes: s21: data cleaning is carried out, and abnormal values caused by errors in sampling some historical photovoltaic power data are proposed; s22: normalizing the cleaned sample data; s23: initializing AO algorithm parameters, including population scale, maximum iteration times T, and exploring and developing parameters alpha and delta; s24, initializing a population position X, initial population fitness and optimal individuals; s25: the method comprises the steps of sequentially carrying out an expansion exploration phase, a reduction exploration phase, an expansion development phase and a reduction development phase, and continuously updating the population positions; s26: calculating the fitness of the updated population to obtain the current optimal individual position and fitness, comparing the current optimal individual with the optimal individual fitness found until the t generation, and reserving the optimal individual position; s27: judging whether the maximum iteration times or solving conditions are reached, if so, outputting an optimal value, and if not, returning to the step S25; s28: and inputting the final optimized weight value result into the DELM model.

5. The EMD-AO-DELM based photovoltaic power calculation method according to claim 1, wherein: the step S3 comprises the following steps: s31: decomposing the photovoltaic historical data by adopting EMD to obtain a group of IMF components; s32: respectively establishing an AO-DELM model for each IMF component, and predicting each component; s33: and superposing the prediction results of the subsequences and verifying the accuracy of model prediction.

6. The EMD-AO-DELM based photovoltaic power calculation method according to claim 1, wherein: the step S4 includes: adopts both MAPE and RMSE as error indexes, and the MAPE calculation formula is as followsThe calculation formula of RMSE is +.>Where yoi is the i-th true value in the sample, ypi is the i-th predicted value in the sample, and the smaller the two values, the higher the accuracy.