Method, device and storage medium for short-term power generation prediction of photovoltaic power station based on TLBO-Elman

Abstract

The present invention discloses a method, device and storage medium for predicting short-term power generation of photovoltaic power stations based on TLBO-Elman, and obtains the historical day with the highest similarity according to the meteorological data and weather type of the forecast day; decomposes the photovoltaic power generation data of the historical day, and superimposes the obtained IMF components representing the short-term law as the short-term forecast intermediate frequency data; normalizes the meteorological data of the forecast day and the short-term forecast intermediate frequency data, and inputs the power generation prediction model corresponding to the weather type of the forecast day to predict the short-term forecast power of photovoltaic power generation on the forecast day; the power generation prediction model is obtained by training the Elman neural network based on the meteorological data and photovoltaic power generation data of the historical power generation day, and the initial weight and threshold of the Elman neural network are obtained by using the TLBO optimization algorithm. The present invention can realize the effective prediction of the short-term power of photovoltaic power generation under different meteorological conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation prediction, and in particular to a method, device and storage medium for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman. Background Art

[0002] Since the beginning of the new century, environmental protection and energy crises have become increasingly severe. Solar power generation has been widely used around the world due to its advantages of cleanliness, renewable energy and easy distributed promotion. However, photovoltaic power generation is closely related to meteorological factors. It has strong intermittent, volatility and randomness. Large-scale photovoltaic power generation grid connection not only has a strong impact on the power grid, but also affects the quality of electric energy, and also has an adverse impact on the stable operation and effective dispatch of the power system. Reasonable prediction of photovoltaic power generation can effectively reduce the impact of grid connection on the power grid, improve the stability of power station operation, and enhance the main grid's acceptance of photovoltaic power. Therefore, effective prediction of photovoltaic power generation is of great significance to promoting the application of solar power generation and the safe, efficient and economical operation of the entire power grid.

[0003] Existing photovoltaic power generation prediction methods are mainly divided into two categories: indirect prediction method and direct prediction method. The indirect prediction method is based on the physical model of the photovoltaic power generation system. Its model is complex and has high requirements for the accuracy of meteorological information. When the parameters of the photovoltaic power station components change, the accuracy of the prediction results is significantly reduced. The direct prediction method is based on meteorological information and historical power generation information of the photovoltaic power station. Its model is relatively simple and does not need to consider the parameters of the power station components. It has high accuracy and strong adaptability. However, most of the existing direct prediction technologies are prone to fall into local optimality due to the randomness of the initial weights and thresholds, and the prediction ability is poor under the situation of meteorological mutations. Therefore, it is necessary to study high-precision short-term power prediction technology for photovoltaic power stations. Summary of the invention

[0004] The present invention provides a TLBO-Elman-based photovoltaic power station short-term power generation prediction method, device and storage medium to solve the problem of low estimation accuracy of conventional photovoltaic power station short-term power generation prediction methods.

[0005] In a first aspect, a method for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman is provided, comprising:

[0006] Obtain the meteorological data for the forecast day and determine the weather type for the forecast day;

[0007] According to the meteorological data of the forecast day and its weather type, obtain the historical day with the highest similarity to the forecast day;

[0008] The photovoltaic power generation data of the most similar historical days are decomposed, and the IMF components representing short-term rules obtained by the decomposition are superimposed as short-term forecast intermediate frequency data;

[0009] The meteorological data of the forecast day and the medium-frequency data of the short-term forecast are normalized, and then input into the power generation prediction model corresponding to the weather type of the forecast day to obtain the normalized photovoltaic power generation prediction power; wherein, the power generation prediction model corresponding to each weather type is obtained by training the Elman neural network based on the meteorological data and photovoltaic power generation data of the historical power generation days of each weather type, and the initial weights and thresholds of the Elman neural network are obtained by using the TLBO optimization algorithm;

[0010] The photovoltaic power generation forecast power is denormalized to obtain the short-term photovoltaic power generation forecast power on the forecast day.

[0011] Furthermore, the power generation prediction model corresponding to each weather type is obtained by the following method:

[0012] According to the weather type, historical days corresponding to each weather type are extracted from historical days to construct a training sample set. Each sample in the training sample set includes the meteorological feature vector, short-term forecast medium-frequency data and photovoltaic power generation data of the historical day; wherein the meteorological feature vector is obtained based on the meteorological data;

[0013] The meteorological characteristic vector, short-term forecast medium-frequency data and photovoltaic power generation data of each sample are normalized;

[0014] Taking the normalized meteorological feature vector and short-term forecast intermediate frequency data as input and the normalized photovoltaic power generation data as output, the Elman neural network is trained based on the training sample sets corresponding to each weather type, and the power generation prediction models corresponding to each weather type are obtained respectively.

[0015] Among them, the initial weights and thresholds of the Elman neural network are obtained using the TLBO optimization algorithm.

[0016] Furthermore, the receiving layer, hidden layer, and output layer of the Elman neural network can be calculated by the following formula:

[0017] Undertaking layer:

[0018] x c (k) = αx c (k-1)+h(k-1)

[0019] Hidden layer:

[0020] h(k)=f(w 1 X(k-1)+w 3 xc (k)+b1)

[0021] Output layer:

[0022] Y(k)=g(w 2 x(k)+b2)

[0023] Among them, X(k) is the input of the input layer at time k; x c (k) is the output of the receiving layer at time k; h(k) is the output of the hidden layer at time k; Y(k) is the output of the output layer at time k; g(·) is the output layer neuron transfer function; f(·) is the hidden layer neuron transfer function; b1 is the hidden layer threshold; b2 is the output layer threshold; α is the self-connection feedback gain factor; w 1 is the weight between the input layer and the hidden layer; w 2 is the weight between the hidden layer and the output layer; w 3 is the weight between the hidden layer and the connecting layer;

[0024] The network error E is calculated by the calculated output Y(k) and the true value y(k):

[0025]

[0026] Where N represents the total length of the predicted data;

[0027] The weight w between the hidden layer and the output layer 2 The correction value Δw 2 Calculated by the following formula:

[0028]

[0029] Where η is the learning rate; i = 1, 2, ..., m; j = 1, 2, ..., n, m and n are the number of nodes in the output layer and the hidden layer respectively; g'(·) is the derivative of the transfer function g(·) of the output layer neuron;

[0030] The weight w between the input layer and the hidden layer 1 The correction value Δw 1 Calculated by the following formula:

[0031]

[0032] Where q = 1, 2, …, r, r represents the number of nodes in the input layer; f'(·) is the derivative of the transfer function f(·) of the hidden layer neuron;

[0033] The weight w between the receiving layer and the hidden layer 3 The correction value Δw 3 Calculated by the following formula:

[0034]

[0035] Where, l = 1, 2, ..., n;

[0036] During the training process, if the network error E is less than the set error ε or the maximum training step is reached, the iteration is terminated.

[0037] Furthermore, the TLBO optimization algorithm is used to obtain the initial weights and thresholds of the Elman neural network, specifically including:

[0038] The purpose of optimizing the initial weights and thresholds of the Elman neural network is achieved by imitating the teacher's teaching and the student's learning behavior. It mainly includes two stages: teaching and learning:

[0039] Teaching stage:

[0040] L i,new =L i,old +s i (O new -T F O i )

[0041] T F =round[1+round(0,1){2-1}]

[0042] Where, L i,new and L i,old Indicates the ability level of students before and after teaching; new and O i Indicates the teacher level and average level; T F is the teaching factor; s i is a random number between [0,1];

[0043] Learning stage:

[0044]

[0045] Where, L i and L j represents different learners; d(·) is the objective function.

[0046] Furthermore, the step of obtaining the historical day with the highest similarity to the predicted day based on the meteorological data of the predicted day and its weather type specifically includes:

[0047] According to the weather type of the forecast day, select the historical days with the same weather type as the forecast day;

[0048] The similarity between the meteorological feature vector constructed based on the meteorological data of the forecast day and the meteorological feature vectors of the selected historical days is calculated, and the historical day with the highest similarity is selected.

[0049] Furthermore, the similarity is calculated by the following formula:

[0050]

[0051] Among them, F i represents the similarity between the forecast day and the i-th historical day; M represents the dimension of the meteorological feature vector; ω i (k) represents the grey correlation coefficient between the kth meteorological characteristic vector component of the forecast day and the ith historical day, and its calculation formula is as follows:

[0052]

[0053] in, represents the kth component of the meteorological feature vector of the forecast day, represents the kth component of the meteorological characteristic vector of the i-th historical day, and ρ represents the resolution coefficient with a value between [0,1].

[0054] Furthermore, the photovoltaic power generation data of the historical day with the highest similarity is decomposed and processed, and the IMF components representing the short-term rules obtained by the decomposition are superimposed as the short-term prediction intermediate frequency data, specifically including:

[0055] The photovoltaic power generation data of the most similar historical days are decomposed by EEMD to obtain multiple IMF components and a residual component;

[0056] The maximum run length and number of runs of each IMF component are tested and calculated by the run test method, and the intermediate frequency IMF component is screened out according to the maximum run length and number of runs;

[0057] The screened intermediate frequency IMF components are superimposed as short-term forecast intermediate frequency data.

[0058] Furthermore, the meteorological data include average solar irradiance, maximum temperature, average temperature, average humidity and average wind speed; and the weather types include sunny, rainy and cloudy.

[0059] In the second aspect, a short-term power generation prediction device for a photovoltaic power station based on TLBO-Elman is provided, comprising:

[0060] The data acquisition module is used to obtain the meteorological data of the forecast day and determine the weather type of the forecast day;

[0061] The historical day acquisition module is used to obtain the historical day with the highest similarity to the forecast day based on the meteorological data of the forecast day and its weather type;

[0062] The intermediate frequency data acquisition module is used to decompose the photovoltaic power generation data of the most similar historical day, and superimpose the IMF components representing the short-term rules obtained by the decomposition process as the short-term forecast intermediate frequency data;

[0063] The photovoltaic power generation prediction module is used to normalize the meteorological data of the prediction day and the short-term prediction medium-frequency data, and then input the power generation prediction model corresponding to the weather type of the prediction day to obtain the normalized photovoltaic power generation prediction power; wherein, the power generation prediction model corresponding to each weather type is obtained by training the Elman neural network based on the meteorological data and photovoltaic power generation data of the historical power generation days of each weather type, and the initial weights and thresholds of the Elman neural network are obtained by using the TLBO optimization algorithm;

[0064] The anti-normalization module is used to perform anti-normalization processing on the photovoltaic power generation prediction power to obtain the short-term prediction power of photovoltaic power generation on the prediction day.

[0065] In a third aspect, a computer-readable storage medium is provided, which stores a computer program, and when the computer program is executed by a processor, the short-term power generation prediction method of a photovoltaic power station based on TLBO-Elman is implemented as described above.

[0066] Beneficial Effects

[0067] The present invention proposes a method, device and storage medium for short-term power generation prediction of photovoltaic power stations based on TLBO-Elman. The method obtains meteorological data and determines the weather type. Then, based on the meteorological data and the weather type, the historical days with the highest similarity are selected by similarity. The photovoltaic power generation data of the historical days with the highest similarity are decomposed, and the intermediate frequency components are extracted to construct short-term prediction intermediate frequency data. Then, the meteorological data of the prediction day and the short-term prediction intermediate frequency data of the historical days with the highest similarity are normalized, and then input into the power generation prediction model corresponding to the weather type of the prediction day, so as to predict the short-term photovoltaic power generation of the prediction day. The power generation prediction model is obtained by training the Elman neural network based on the meteorological data and photovoltaic power generation data of the historical power generation days of each weather type, and the initial weights and thresholds of the Elman neural network are obtained by using the TLBO optimization algorithm. The present invention can realize the effective prediction of short-term photovoltaic power generation under different meteorological conditions, and provide a basis for the extended research work on the grid connection of photovoltaic power generation and the stable operation of power stations. 。 BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0069] Figure 1 The present invention provides a flow chart of a method for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman. DETAILED DESCRIPTION

[0070] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0071] The present invention provides a method, device and storage medium for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman, which aims to solve the problem that the estimation results of conventional short-term power generation prediction methods of photovoltaic power stations are not accurate. Before predicting the photovoltaic power generation on the forecast day, it is necessary to first construct a power generation prediction model corresponding to each weather type, and the specific method is as follows.

[0072] The irradiance intensity I, temperature T, humidity H, wind speed W and photovoltaic power generation data P of the photovoltaic power station to be predicted over a long period of time are measured and recorded using instruments such as irradiance meters, thermometers, hygrometers and wind speed testers. The sampling interval is 5 minutes. At the same time, all recorded historical days are classified into three weather conditions: sunny, rainy and cloudy.

[0073] Calculate or filter the average solar irradiance I for each historical day based on recorded meteorological data a , maximum temperature T max , average temperature T a , average humidity H a and the average wind speed W a And other information, and construct the meteorological feature vector X * =[I a ,T a ,T max ,H a ,W a ].

[0074] The photovoltaic power generation data of each historical day is decomposed by EEMD (Ensemble Empirical Mode Decomposition) to obtain multiple IMF components and a residual component. The specific process includes:

[0075] The Gaussian white noise e i (t) is added to the photovoltaic power generation data information sequence P(t), and the new mixed signal after adding Gaussian white noise can be expressed by the following formula:

[0076] P i (t) = P (t) + e i (t)

[0077] Among them, i represents the number of times Gaussian white noise is added to the original signal;

[0078] Using the empirical mode decomposition technique, the new sequence P i (t) is decomposed to obtain V IMF components c j (t) and a residual component r0(t):

[0079]

[0080] Repeat the above process N times, each time adding new Gaussian white noise to the original photovoltaic power generation data information sequence P(t) and finding the corresponding IMF component;

[0081] The calculated IMF components are averaged, and the IMF components of the original photovoltaic power generation data information sequence P(t) can be expressed by the following equation:

[0082]

[0083] Finally, the original photovoltaic power generation data information sequence P(t) is decomposed into:

[0084]

[0085] The maximum run length and number of runs of each IMF component obtained by decomposing the calculated photovoltaic power generation data are tested by the run test method, and the intermediate frequency IMF component is screened out according to the maximum run length and number of runs. The specific process is:

[0086] For each IMF component c j (t), are all time series, assuming that they are represented by the sequence {G(t)} (t=1,2,…,p), and their average value is G a , then the timing symbol is defined as:

[0087]

[0088] Where, i = 1, 2, ..., p;

[0089] Continuous 1 or 0 represents a run, and the number of 1 or 0 in a run represents the run length. The larger the maximum run length, the more stable the data and the lower the frequency. Then, the maximum run length and the number of runs can be compared with the preset threshold to screen out the intermediate frequency IMF components that can represent the short-term law of photovoltaic power generation, and these IMF components can be superimposed as short-term prediction intermediate frequency data.

[0090] Based on this, sunny day training sample sets, rainy day training sample sets and cloudy day training sample sets can be extracted according to the weather types of recorded historical days, and each sample includes the meteorological feature vector of the historical day, short-term prediction medium-frequency data and photovoltaic power generation data.

[0091] Take the construction of a power generation prediction model for sunny days as an example to illustrate:

[0092] The meteorological feature vector, short-term forecast intermediate frequency data and photovoltaic power generation data of each sample in the sunny training sample set are normalized;

[0093] With the normalized meteorological feature vector and short-term forecast intermediate frequency data as input and the normalized photovoltaic power generation data as output, the Elman neural network is trained based on the sunny day training sample set to obtain the power generation prediction model corresponding to sunny days.

[0094] More specifically, the number of nodes r, m and n in the input layer, output layer and hidden layer of the Elman neural network is determined according to the number of components in the meteorological feature vector and the photovoltaic power generation data. The input of the Elman neural network consists of the meteorological feature vector and the short-term forecast intermediate frequency data, and its output is the photovoltaic power generation data (which is a time series of power generation at different times). Given the parameters of the Elman neural network training function, network learning function and node transfer function, the receiving layer, hidden layer and output layer of the Elman neural network can be calculated by the following formula:

[0095] Undertaking layer:

[0096] x c (k) = αx c (k-1)+h(k-1)

[0097] Hidden layer:

[0098] h(k)=f(w 1 X(k-1)+w 3 x c (k)+b1)

[0099] Output layer:

[0100] Y(k)=g(w 2 x(k)+b2)

[0101] Among them, X(k) is the input of the input layer at time k; x c (k) is the output of the receiving layer at time k; h(k) is the output of the hidden layer at time k; Y(k) is the output of the output layer at time k; g(·) is the output layer neuron transfer function; f(·) is the hidden layer neuron transfer function; the output layer neuron transfer function and the hidden layer neuron transfer function can use the Signmoid function, tansig function or logsig function; b1 is the hidden layer threshold; b2 is the output layer threshold; α is the self-connection feedback gain factor; w 1 is the weight between the input layer and the hidden layer; w 2 is the weight between the hidden layer and the output layer; w 3 is the weight between the hidden layer and the connecting layer;

[0102] The network error E is calculated by the calculated output Y(k) and the true value y(k):

[0103]

[0104] Where N represents the total length of the predicted data;

[0105] The weight w between the hidden layer and the output layer 2 The correction value Δw 2 Calculated by the following formula:

[0106]

[0107] Where η is the learning rate; i = 1, 2, ..., m; j = 1, 2, ..., n, m and n are the number of nodes in the output layer and the hidden layer respectively; g'(·) is the derivative of the transfer function g(·) of the output layer neuron;

[0108] The weight w between the input layer and the hidden layer 1 The correction value Δw 1 Calculated by the following formula:

[0109]

[0110] Where q = 1, 2, …, r, r represents the number of nodes in the input layer; f'(·) is the derivative of the transfer function f(·) of the hidden layer neuron;

[0111] The weight w between the receiving layer and the hidden layer 3 The correction value Δw 3 Calculated by the following formula:

[0112]

[0113] Where, l = 1, 2, ..., n;

[0114] During the training process, the relationship between the network error E and the set error ε is determined. If E>ε, the hidden layer and output layer calculations are returned; otherwise, the iteration is terminated. If the maximum training step is reached, the iteration is terminated.

[0115] The number of nodes in the input layer and output layer of the Elman neural network is determined by the input and output data. The number of nodes in the hidden layer is determined by the number of nodes in the output layer and the input layer. The optimal number of nodes in the hidden layer is usually obtained by the following methods:

[0116]

[0117] Where C is the adjustment constant, and its value range is [1,10].

[0118] During implementation, the TLBO optimization algorithm is used to obtain the initial weights and thresholds of the Elman neural network, including:

[0119] The purpose of optimizing the initial weights and thresholds of the Elman neural network is achieved by imitating the teacher's teaching and the student's learning behavior. It mainly includes two stages: teaching and learning:

[0120] Teaching stage:

[0121] L i,new =L i,old +s i (O new -T F O i )

[0122] T F =round[1+round(0,1){2-1}]

[0123] Where, L i,new and L i,old Indicates the ability level of students before and after teaching; new and O i Indicates the teacher level and average level; T F is the teaching factor; s i is a random number between [0,1];

[0124] Learning stage:

[0125]

[0126] Where, L i and L j represents different learners; d(·) is the objective function.

[0127] The process of obtaining the corresponding power generation prediction models based on the rainy day training sample set and the cloudy day training sample set can refer to the aforementioned process of obtaining the corresponding power generation prediction model for sunny days based on the sunny day training sample set, which will not be repeated here.

[0128] On the basis of obtaining the power generation prediction models corresponding to the above-mentioned weather types, short-term power generation prediction of the photovoltaic power station can be performed, and the following embodiments are provided for illustration.

[0129] Example 1

[0130] This embodiment provides a method for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman, including:

[0131] S1: Obtain the meteorological data of the forecast day and determine the weather type of the forecast day;

[0132] S2: Based on the meteorological data of the forecast day and its weather type, obtain the historical day with the highest similarity to the forecast day;

[0133] S3: Decompose the photovoltaic power generation data of the historical day with the highest similarity, and superimpose the IMF components representing the short-term rules obtained by the decomposition process as the short-term forecast intermediate frequency data;

[0134] S4: normalizing the meteorological data of the forecast day and the short-term forecast medium-frequency data, and then inputting them into the power generation prediction model corresponding to the weather type of the forecast day to obtain the normalized photovoltaic power generation prediction power;

[0135] S5: Denormalize the photovoltaic power generation prediction power to obtain the short-term photovoltaic power generation prediction power on the prediction day.

[0136] In step S2, the step of obtaining the historical day with the highest similarity to the predicted day based on the meteorological data of the predicted day and its weather type specifically includes:

[0137] According to the weather type of the forecast day, select the historical days with the same weather type as the forecast day;

[0138] The similarity between the meteorological feature vector constructed based on the meteorological data of the forecast day and the meteorological feature vectors of the selected historical days is calculated, and the historical day with the highest similarity is selected.

[0139] The similarity is calculated by the following formula:

[0140]

[0141] Among them, F i represents the similarity between the predicted day and the i-th historical day; M represents the dimension of the meteorological feature vector, and in this embodiment, M is 5; ω i(k) represents the grey correlation coefficient between the kth meteorological characteristic vector component of the forecast day and the ith historical day, and its calculation formula is as follows:

[0142]

[0143] in, represents the kth component of the meteorological feature vector of the forecast day, represents the kth component of the meteorological characteristic vector of the i-th historical day, and ρ represents the resolution coefficient with a value between [0,1].

[0144] Example 2

[0145] This embodiment provides a short-term power generation prediction device for a photovoltaic power station based on TLBO-Elman, including:

[0146] The data acquisition module is used to obtain the meteorological data of the forecast day and determine the weather type of the forecast day;

[0147] The historical day acquisition module is used to obtain the historical day with the highest similarity to the forecast day based on the meteorological data of the forecast day and its weather type;

[0148] The intermediate frequency data acquisition module is used to decompose the photovoltaic power generation data of the most similar historical day, and superimpose the IMF components representing the short-term rules obtained by the decomposition process as the short-term forecast intermediate frequency data;

[0149] The photovoltaic power generation prediction module is used to normalize the meteorological data of the forecast day and the short-term forecast medium-frequency data, and then input the power generation prediction model corresponding to the weather type of the forecast day to obtain the normalized photovoltaic power generation prediction power;

[0150] The anti-normalization module is used to perform anti-normalization processing on the photovoltaic power generation prediction power to obtain the short-term prediction power of photovoltaic power generation on the prediction day.

[0151] Example 3

[0152] This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the above-mentioned TLBO-Elman-based photovoltaic power station short-term power generation prediction method is implemented.

[0153] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0154] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0155] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0156] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0157] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0158] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0159] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman, characterized in that: include: Obtain the meteorological data for the forecast day and determine the weather type for the forecast day; According to the meteorological data of the forecast day and its weather type, obtain the historical day with the highest similarity to the forecast day; The photovoltaic power generation data of the most similar historical days are decomposed, and the IMF components representing short-term rules obtained by the decomposition are superimposed as short-term forecast intermediate frequency data; The meteorological data of the forecast day and the medium-frequency data of the short-term forecast are normalized, and then input into the power generation prediction model corresponding to the weather type of the forecast day to obtain the normalized photovoltaic power generation prediction power; wherein, the power generation prediction model corresponding to each weather type is obtained by training the Elman neural network based on the meteorological data and photovoltaic power generation data of the historical power generation days of each weather type, and the initial weights and thresholds of the Elman neural network are obtained by using the TLBO optimization algorithm; The photovoltaic power generation prediction power is denormalized to obtain the short-term photovoltaic power generation prediction power on the prediction day; The receiving layer, hidden layer, and output layer of the Elman neural network can be calculated by the following formula: Undertaking layer: x c (k)=αx c (k-1)+h(k-1) Hidden layer: h(k)=f(w 1 X(k-1)+w 3 x c (k)+b1) Output layer: Y(k)=g(w 2 x(k)+b2) Among them, X(k) is the input of the input layer at time k; x c (k) is the output of the receiving layer at time k; h(k) is the output of the hidden layer at time k; Y(k) is the output of the output layer at time k; g(·) is the output layer neuron transfer function; f(·) is the hidden layer neuron transfer function; b1 is the hidden layer threshold; b2 is the output layer threshold; α is the self-connection feedback gain factor; w 1 is the weight between the input layer and the hidden layer; w 2 is the weight between the hidden layer and the output layer; w 3 is the weight between the hidden layer and the receiving layer; The network error E is calculated by the calculated output Y(k) and the true value y(k): Where N represents the total length of the predicted data; The weight w between the hidden layer and the output layer 2 The correction value Δw 2 Calculated by the following formula: Where η is the learning rate; i = 1, 2, ..., m; j = 1, 2, ..., n, m and n are the number of nodes in the output layer and the hidden layer respectively; g'(·) is the derivative of the transfer function g(·) of the output layer neuron; The weight w between the input layer and the hidden layer 1 The correction value Δw 1 Calculated by the following formula: Where q = 1, 2, …, r, r represents the number of nodes in the input layer; f'(·) is the derivative of the transfer function f(·) of the hidden layer neuron; The weight w between the receiving layer and the hidden layer 3 The correction value Δw 3 Calculated by the following formula: Where, l = 1, 2, ..., n; During the training process, if the network error E is less than the set error ε or the maximum training step is reached, the iteration is terminated.

2. The method for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman according to claim 1 is characterized in that: The power generation prediction model corresponding to each weather type is obtained by the following method: According to the weather type, historical days corresponding to each weather type are extracted from historical days to construct a training sample set. Each sample in the training sample set includes the meteorological feature vector, short-term forecast medium-frequency data and photovoltaic power generation data of the historical day; wherein the meteorological feature vector is obtained based on the meteorological data; Normalize the meteorological characteristic vector, short-term forecast medium-frequency data and photovoltaic power generation data of each sample; Taking the normalized meteorological feature vector and short-term forecast intermediate frequency data as input and the normalized photovoltaic power generation data as output, the Elman neural network is trained based on the training sample sets corresponding to each weather type, and the power generation prediction models corresponding to each weather type are obtained respectively. Among them, the initial weights and thresholds of the Elman neural network are obtained using the TLBO optimization algorithm.

3. The method for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman according to claim 2 is characterized in that: The TLBO optimization algorithm is used to obtain the initial weights and thresholds of the Elman neural network, including: The purpose of optimizing the initial weights and thresholds of the Elman neural network is achieved by imitating the teacher's teaching and the student's learning behavior. It mainly includes two stages: teaching and learning: Teaching stage: L i,new =L i,old +s i (O new -T F O i ) T F =round[1+round(0,1){2-1}] Where, L i,new and L i,old Indicates the ability level of students before and after teaching; new and O i Indicates the teacher level and average level; T F is the teaching factor; s i is a random number between [0,1]; Learning stage: Where, L i and L j represents different learners; d(·) is the objective function.

4. The method for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman according to claim 1 is characterized in that: The step of obtaining the historical day with the highest similarity to the forecast day according to the meteorological data of the forecast day and its weather type specifically includes: According to the weather type of the forecast day, select the historical days with the same weather type as the forecast day; The similarity between the meteorological feature vector constructed based on the meteorological data of the forecast day and the meteorological feature vectors of the selected historical days is calculated, and the historical day with the highest similarity is selected.

5. The method for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman according to claim 4 is characterized in that: The similarity is calculated by the following formula: Among them, F i represents the similarity between the forecast day and the i-th historical day; M represents the dimension of the meteorological feature vector; ω i (k) represents the grey correlation coefficient between the kth meteorological characteristic vector component of the forecast day and the ith historical day, and its calculation formula is as follows: in, represents the kth component of the meteorological feature vector of the forecast day, represents the kth component of the meteorological characteristic vector of the i-th historical day, and ρ represents the resolution coefficient with a value between [0,1].

6. The method for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman according to claim 1, characterized in that: The photovoltaic power generation data of the historical day with the highest similarity is decomposed and processed, and the IMF components representing the short-term rules obtained by the decomposition are superimposed as the short-term prediction intermediate frequency data, specifically including: The photovoltaic power generation data of the most similar historical days are decomposed by EEMD to obtain multiple IMF components and a residual component; The maximum run length and number of runs of each IMF component are tested and calculated by the run test method, and the intermediate frequency IMF component is screened out according to the maximum run length and number of runs; The screened intermediate frequency IMF components are superimposed as short-term forecast intermediate frequency data.

7. The method for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman according to claim 1, characterized in that: The meteorological data include average solar irradiance, maximum temperature, average temperature, average humidity and average wind speed; the weather types include sunny, rainy and cloudy.

8. A photovoltaic power station short-term power generation prediction device based on TLBO-Elman, characterized in that: include: The data acquisition module is used to obtain the meteorological data of the forecast day and determine the weather type of the forecast day; The historical day acquisition module is used to obtain the historical day with the highest similarity to the forecast day based on the meteorological data of the forecast day and its weather type; The intermediate frequency data acquisition module is used to decompose the photovoltaic power generation data of the most similar historical day, and superimpose the IMF components representing the short-term rules obtained by the decomposition process as the short-term forecast intermediate frequency data; The photovoltaic power generation prediction module is used to normalize the meteorological data of the prediction day and the short-term prediction medium-frequency data, and then input the power generation prediction model corresponding to the weather type of the prediction day to obtain the normalized photovoltaic power generation prediction power; wherein, the power generation prediction model corresponding to each weather type is obtained by training the Elman neural network based on the meteorological data and photovoltaic power generation data of the historical power generation days of each weather type, and the initial weights and thresholds of the Elman neural network are obtained by using the TLBO optimization algorithm; The anti-normalization module is used to perform anti-normalization processing on the photovoltaic power generation prediction power to obtain the short-term prediction power of photovoltaic power generation on the prediction day; The receiving layer, hidden layer, and output layer of the Elman neural network can be calculated by the following formula: Undertaking layer: x c (k)=αx c (k-1)+h(k-1) Hidden layer: h(k)=f(w 1 X(k-1)+w 3 x c (k)+b1) Output layer: Y(k)=g(w 2 x(k)+b2) Among them, X(k) is the input of the input layer at time k; x c (k) is the output of the receiving layer at time k; h(k) is the output of the hidden layer at time k; Y(k) is the output of the output layer at time k; g(·) is the output layer neuron transfer function; f(·) is the hidden layer neuron transfer function; b1 is the hidden layer threshold; b2 is the output layer threshold; α is the self-connection feedback gain factor; w 1 is the weight between the input layer and the hidden layer; w 2 is the weight between the hidden layer and the output layer; w 3 is the weight between the hidden layer and the receiving layer; The network error E is calculated by the calculated output Y(k) and the true value y(k): Where N represents the total length of the predicted data; The weight w between the hidden layer and the output layer 2 The correction value Δw 2 Calculated by the following formula: Where η is the learning rate; i = 1, 2, ..., m; j = 1, 2, ..., n, m and n are the number of nodes in the output layer and the hidden layer respectively; g'(·) is the derivative of the transfer function g(·) of the output layer neuron; The weight w between the input layer and the hidden layer 1 The correction value Δw 1 Calculated by the following formula: Where q = 1, 2, …, r, r represents the number of nodes in the input layer; f'(·) is the derivative of the transfer function f(·) of the hidden layer neuron; The weight w between the receiving layer and the hidden layer 3 The correction value Δw 3 Calculated by the following formula: Where, l = 1, 2, ..., n; During the training process, if the network error E is less than the set error ε or the maximum training step is reached, the iteration is terminated.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for predicting short-term power generation of a photovoltaic power station based on TLBO-Elman is implemented as described in any one of claims 1 to 7.

Images