Photovoltaic cluster power prediction method, device, equipment and medium

Abstract

The present invention discloses a photovoltaic cluster power prediction method, which relates to the technical field of power system control, and is used to solve the inaccuracy of existing power prediction results. The method includes the following steps: constructing multiple correlation analysis objects; calculating the power correlation coefficient corresponding to the correlation analysis object under each correlation coefficient; performing weighted calculation on the correlation coefficients of multiple analysis objects to obtain the mean value of the correlation coefficient, and selecting a cluster power prediction representative power station according to the mean value of the correlation coefficient; inputting the historical power and meteorological data of the cluster power prediction representative power station into an integrated power prediction model to obtain a representative power station power prediction result; and performing extrapolation based on the power prediction result to obtain a photovoltaic cluster total power prediction result. The present invention also discloses a photovoltaic cluster power prediction device, an electronic device, and a computer storage medium. The present invention performs correlation analysis on a single power station-cluster and performs power prediction through an integrated model, so that the power result is accurate and reliable.

Description

Technical Field

[0001] The present invention relates to the field of power system control technology, and in particular to a photovoltaic cluster power prediction method, device, equipment and medium based on multi-feature extraction. Background Art

[0002] As new energy sources have accounted for a higher proportion of the power generation energy structure, the use of solar energy as a renewable clean energy has also increased. However, the solar energy grid connection process has high requirements for the safety, flexible scheduling and new energy absorption capacity of the power system; in order to maintain the stable operation of the power system, accurate power prediction of solar energy is required.

[0003] At present, my country's photovoltaic power generation is mainly centralized photovoltaic and distributed photovoltaic. Among them, distributed photovoltaic can better solve the problem of photovoltaic power station consumption, so it has gradually been more widely used. With the increase in the capacity of photovoltaic power stations in clusters, it is necessary to predict the power of clusters to ensure power generation efficiency and safety. At present, cluster power prediction is divided into accumulation method, extrapolation method and statistical upscaling method. Among them, accumulation method and extrapolation method have strict data requirements and high data complexity. In addition, the prediction accuracy of a single power plant and a single region will have a barrel effect on the entire cluster. Although the statistical upscaling method only needs to select the representative power station in the cluster for prediction and extrapolate the results using the weight coefficient, and predict the power of the entire cluster; however, it is easy to affect the cluster prediction power due to incomplete consideration of factors in the selection process of the representative power station. Therefore, in order to further improve the prediction accuracy of photovoltaic cluster power, it is urgent to further consider more factors to select representative power stations in the district, and optimize the prediction scheme and model. Summary of the invention

[0004] In order to overcome the deficiencies of the prior art, one of the objectives of the present invention is to provide a photovoltaic cluster power prediction method, which performs multi-feature extraction through correlation coefficient analysis and then performs power prediction.

[0005] One of the purposes of the present invention is achieved by the following technical solution:

[0006] A photovoltaic cluster power prediction method comprises the following steps:

[0007] Construct correlation analysis objects for multiple single power station clusters;

[0008] Calculate the power correlation coefficient corresponding to the correlation analysis object under each preset correlation coefficient respectively;

[0009] Performing weighted calculation on the correlation coefficients of the plurality of analysis objects to obtain a mean value of the correlation coefficients, and selecting a representative power station for cluster power prediction according to the mean value of the correlation coefficients;

[0010] Inputting the historical power and meteorological data of the cluster power forecast representative power station into the power forecast model integrated with LSTM and 1DCNN to obtain the power forecast result of the power forecast representative power station;

[0011] The power prediction result of the representative power station is extrapolated to obtain the total power prediction result of the photovoltaic cluster.

[0012] Furthermore, the correlation analysis objects include single power station-other single power stations, single power station-total cluster, single power station-cluster excluding corresponding single power station.

[0013] Furthermore, the preset correlation coefficients include Pearson correlation coefficient, Kendall correlation coefficient and Spearman correlation coefficient, and the power correlation coefficient is obtained by substituting the power of a single power station, the total power of a cluster and the total power of a cluster without a single power station into the preset correlation coefficient calculation formula.

[0014] Furthermore, when the correlation analysis object is a single power station-other single power stations, after calculating the power correlation coefficient, it also includes: calculating the mean of multiple correlation coefficients of different single power stations in the cluster-other single power stations to obtain the overall characteristics of the correlation analysis object.

[0015] Further, weighted calculation is performed on the correlation coefficients of the plurality of analysis objects to obtain a mean value of the correlation coefficients, and a representative power station for cluster power prediction is selected according to the mean value of the correlation coefficients, including:

[0016] According to the correlation coefficient, a weight standard is calculated by Shapley algorithm;

[0017] Calculating the Shapley value of the correlation coefficient according to the weight standard;

[0018] Calculating the weight of the correlation coefficient according to the Shapley value;

[0019] According to the weights, the mean value of the correlation coefficient is calculated, and the power plant with the largest mean value of the correlation coefficient is selected as the representative power plant for cluster power prediction.

[0020] Furthermore, the prediction model is an LSTM-1DCNN power prediction model integrated by CatBoost.

[0021] Furthermore, the power prediction result of the representative power station is extrapolated to obtain the total power prediction result of the photovoltaic cluster, and the extrapolation satisfies the formula:

[0022]

[0023] Among them, P set is the cluster power prediction, Pref is the reference power station power, C set is the total capacity of the cluster, C ref is the reference power station capacity, P ave is the mean correlation coefficient.

[0024] A second object of the present invention is to provide a photovoltaic cluster power prediction device, which constructs objects for a power station cluster and then performs correlation analysis on each analysis object to obtain a power prediction result.

[0025] The second object of the present invention is achieved by the following technical solution:

[0026] A photovoltaic cluster power prediction device, comprising:

[0027] A construction module is used to construct correlation analysis objects of multiple single power station clusters;

[0028] An analysis module, used to respectively calculate the power correlation coefficient corresponding to the correlation analysis object under each preset correlation coefficient; perform weighted calculation on the correlation coefficients of multiple analysis objects to obtain a mean value of the correlation coefficients, and select a representative power station for cluster power prediction according to the mean value of the correlation coefficients;

[0029] The prediction module is used to input the historical power and meteorological data of the representative power station of the cluster power prediction into the power prediction model integrated with LSTM and 1DCNN to obtain the power prediction result of the representative power station; and to extrapolate the power prediction result of the representative power station to obtain the total power prediction result of the photovoltaic cluster.

[0030] A third object of the present invention is to provide an electronic device for performing one of the objects of the invention, which includes a processor, a storage medium and a computer program, wherein the computer program is stored in the storage medium, and when the computer program is executed by the processor, the above-mentioned photovoltaic cluster power prediction method is implemented.

[0031] A fourth object of the present invention is to provide a computer-readable storage medium storing one of the objects of the invention, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned photovoltaic cluster power prediction method is implemented.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] The present invention realizes a more accurate division of power station clusters through cluster correlation analysis objects of single power stations; calculates the power correlation coefficient through multiple correlation coefficients, considers multiple features of the power station, and performs feature extraction and power prediction; the integrated LSTM-1DCNN network model increases the prediction accuracy; finally, extrapolates the correlation coefficient to obtain the cluster total power prediction result. The present invention considers multiple factors of the power station, including multiple features and multiple analysis objects, so that the result is more reliable, and also selects an optimized prediction model to make the prediction accuracy higher. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a flow chart of a photovoltaic cluster power prediction method according to Embodiment 1;

[0035] Figure 2 Schematic diagram of the LSTM neural network structure of the first embodiment;

[0036] Figure 3 1DCNN neural network structure diagram of the first embodiment;

[0037] Figure 4 is a flow chart of weighted calculation of correlation coefficient in Embodiment 3;

[0038] Figure 5 It is a schematic diagram of the structure of the power prediction model of the representative power station of the photovoltaic cluster in the fourth embodiment;

[0039] Figure 6 is a comparison chart of representative power station prediction results of different prediction schemes under different weather types in Example 4;

[0040] Figure 7 This is a result diagram of selecting representative power stations of a photovoltaic cluster in Embodiment 4;

[0041] Figure 8 is a structural block diagram of a photovoltaic cluster power prediction device according to Embodiment 5;

[0042] Fig. 9 It is a structural block diagram of an electronic device of the sixth embodiment. DETAILED DESCRIPTION

[0043] The present invention will be described in more detail below in conjunction with the accompanying drawings. It should be noted that the following description of the present invention with reference to the accompanying drawings is only illustrative and not restrictive. The various embodiments can be combined with each other to form other embodiments not shown in the following description.

[0044] Embodiment 1

[0045] Embodiment 1 provides a photovoltaic cluster power prediction method, which fully considers the various characteristics of the representative power stations within the cluster and combines multi-feature extraction to perform power prediction. The spatial output characteristic correlation coefficient between the power stations within the cluster is calculated, and then the correlation coefficient is weighted to select the representative power station with the highest correlation; and the representative power station power is predicted through the integrated LSTM-1DCNN network model, and finally the cluster power is derived using the correlation coefficient and capacity ratio to improve the prediction accuracy.

[0046] Please refer to Figure 1 As shown, a photovoltaic cluster power prediction method includes the following steps:

[0047] S1. Construct correlation analysis objects of multiple single power station clusters;

[0048] In this embodiment, the analysis objects include:

[0049] (1) Construct correlation analysis objects between a single power station in the cluster and other single power stations;

[0050] (2) Construct correlation analysis objects for individual power stations within the cluster and the overall cluster;

[0051] (3) Construct a correlation analysis object for a single power plant in the cluster and the entire power plant group excluding the power plant itself.

[0052] S2. Calculate the power correlation coefficient corresponding to the correlation analysis object under each preset correlation coefficient respectively;

[0053] The preset correlation coefficient is not limited in this embodiment and can be set according to the data volume and accuracy requirements. For example, the Pearson correlation coefficient, Kendall correlation coefficient, etc. are used to substitute the single power station power, the cluster total power and the cluster total power without a single power station into the preset correlation coefficient to calculate the power correlation coefficient.

[0054] It should be noted that there will be multiple values ​​for each correlation coefficient calculated between a single power station in a cluster and the rest of the single power stations. Therefore, after calculating the power correlation coefficient, it also includes: calculating the mean of multiple correlation coefficients of different single power stations between the single power station in the cluster and the rest of the single power stations to obtain the total characteristics of the correlation analysis object.

[0055] S3, performing weighted calculation on the correlation coefficients of the plurality of analysis objects to obtain a mean value of the correlation coefficients, and selecting a representative power station for cluster power prediction according to the mean value of the correlation coefficients;

[0056] The weights in S3 can be set according to actual conditions, and the calculation of weights and correlation coefficients can be implemented by some allocation algorithms or weight algorithms, which are not specifically limited in this embodiment.

[0057] S4, inputting the historical power and meteorological data of the cluster power forecast representative power station into the power forecast model integrated with LSTM and 1DCNN to obtain the power forecast result of the power forecast representative power station;

[0058] The above LSTM (Long Short-Term Neural Network) is a special recurrent neural network (RNN) composed of multiple units, each of which includes three gating systems and a memory unit. This network can effectively solve the gradient vanishing and gradient exploding problems in the long sequence training process. The three gating systems are the forget gate, input gate and output gate, specifically:

[0059] Forget Gate:

[0060] F t =sigmoid(w xf x t +w hf h t-1 +b f ),

[0061] Input Gate:

[0062] I t =sigmoid(w xi x t +w hi h t-1 +b i ),

[0063] Output Gate:

[0064] O t =sigmoid(w xo x t +w ho h t-1 +b o ),

[0065] Memory unit status value:

[0066]

[0067] The output of LSTM at time t:

[0068]

[0069] In the above formula, x t is the input variable, h t is the output at time t, w is the weight of the gate control unit, b is the bias value of the gate control unit, F t is the output of the forget gate, I t is the input gate output, O t is the output gate output, Ct is the state of the previous moment, sigmoid and tanh are activation functions respectively. For the specific structure of LSTM neural network, please refer to Figure 2 As shown in Figure 2, the network can predict the future power of the photovoltaic cluster by inputting historical power data and using the historical power data of the representative power station.

[0070] The basic structure of 1DCNN (one-dimensional convolutional neural network) in S4 mainly consists of input layer, convolution layer, pooling layer, fully connected layer and output layer. Please refer to the basic structure of the network Figure 3 As shown, including:

[0071] Input layer: The input dimension of 1DCNN can be one-dimensional or two-dimensional data. The figure shows two-dimensional data.

[0072] Convolutional layer: This layer receives the original data from the input layer and uses the convolution kernel to perform a single-dimensional convolution operation on it. The calculation principle satisfies: The network parameters are trained by feature extraction; in order to match the input data type, the convolution kernel selects a one-dimensional structure. i t is the tth convolution result of the i-th layer; is the jth convolution kernel weight of the i-th layer; X i t is the tth input of the i-th layer; b i t is the tth input bias of the i-th layer. f(x) generally uses the ReLU activation function, which can effectively avoid the gradient diffusion problem.

[0073] Pooling layer: This layer is used to compress the data transmitted from the previous layer to further realize feature extraction, and does not change the number of data layers after convolution; common pooling layers include maximum pooling, average pooling, global average pooling and global maximum pooling layers. The most commonly used is the maximum pooling layer, and its basic principle meets the following requirements: Among them, y i is the output of the i-th layer; x i k is the kth input of the i-th layer.

[0074] Fully connected layer: The data from different layers of the previous network are connected and expanded into a one-dimensional vector form, and the data features are classified and fitted. Finally, they are sent to the output layer, and the network weights are adjusted through the back propagation algorithm to output the optimal parameters.

[0075] Output layer: This layer is the last layer of the network and outputs the final classification or prediction results.

[0076] The network can use the historical irradiance, temperature, and humidity data representing the power station as input to predict future power.

[0077] In summary, this embodiment uses an integrated power prediction model of LSTM and 1DCNN, which can combine the advantages of the two models, realize power prediction through multiple feature data such as historical irradiance, temperature, humidity data, historical power data, etc., and improve prediction accuracy.

[0078] S5: extrapolating the power prediction of the representative power station of the cluster to obtain a total power prediction result of the photovoltaic cluster.

[0079] S4 is to predict the representative power plants within the cluster, and S5 is to use the power of the representative power plants in the cluster to extrapolate the total power of the cluster.

[0080] Extrapolation refers to finding the value outside a function from its existing value. Specifically, the power forecast is extrapolated to obtain the total power forecast of the photovoltaic cluster, and the extrapolation satisfies the formula:

[0081]

[0082] Among them, P set is the cluster power prediction, P ref is the reference power station power, C set is the total capacity of the cluster, C ref is the reference power station capacity, P ave is the mean correlation coefficient.

[0083] Embodiment 2

[0084] Embodiment 2 is an explanation and description of an implementation method of the power correlation analysis in Embodiment 1.

[0085] In this embodiment, the preset correlation coefficients in S2 are selected as Pearson correlation coefficient, Kendall correlation coefficient and Spearman correlation coefficient. Of course, the selection of correlation coefficients can be adaptively changed according to the actual data volume and actual needs. Specifically:

[0086] The Pearson correlation coefficient is used to determine the correlation between random variables through the similarity between waveforms, and then reflect the linear correlation between random variables. It is defined as follows:

[0087]

[0088] Where E(X) and E(Z) are the mathematical expectations of random variables X and Z respectively, D(X) and D(Z) are the variances, and cov(X,Z) is the covariance between random variables X and Z.

[0089] Kendall's correlation coefficient is used to determine the correlation between random variables considering the consistency of variation and is defined as follows:

[0090] Suppose a two-dimensional random variable (X, Z) is (x 1 ,z 1 ) and (x 2 ,z 2 ). If (x 1 -x 2 )(z 1 -z 2 )>0, the two sets of observed values ​​are called harmonious; if (x 1 -x 2 )(z 1 -z 2 )<0, the two sets of observed values ​​are said to be discordant. 1 ,Z 1 ) and (X 2 ,Z 2 ) are independent two-dimensional random variables and have the same distribution as (X,Z), then ρ k Yes (X 1 ,Z 1 ) and (X 2 ,Z 2 ) and (X 1 ,Z 1 ) and (X 2 ,Z 2 ), the formula is as follows:

[0091] ρ k =P[(X 1 -X 2 )(Z 1 -Z 2 )>0]-P[(X 1 -X 2 )(Z 1 -Z 2 )<0].

[0092] The Spearman correlation coefficient is a nonparametric rank statistical parameter whose correlation depends on the monotonicity of the two test variables. It is defined as follows: Assume (X 1 ,Z 1 )、(X 2 ,Z 2 ) and (X 3 ,Z 3 ) are independent two-dimensional random variables and have the same distribution as (X,Z). s (Spearman correlation coefficient) is (X 1 ,X 2 ) and (Z 1 ,Z 3 ) and (X1 ,X 2 ) and (Z 1 ,Z 3 ), the formula is as follows:

[0093] ρ s =3{P[(X 1 -X 2 )(Z 1 -Z 3 )>0]-P[(X 1 -X 2 )(Z 1 -Z 3 )<0]}.

[0094] It should be noted that the above is an explanation of the principle of three correlation coefficients. In the above description, Z and X both represent power series data of two opposing relationships in different correlation analysis objects. In this embodiment, they are specifically: single power station power, cluster total power, and cluster total power without a single power station. The above three formulas are used to calculate the correlation coefficients of the correlation analysis objects of the single power station in the cluster constructed in Example 1 S1 - the remaining single power stations, the single power station - the cluster, and the single power station - the cluster without the corresponding single power station.

[0095] Embodiment 3

[0096] Example 3 is an explanation and description of an implementation method for calculating the correlation coefficient in Example 1.

[0097] In this embodiment, the correlation coefficient is weightedly calculated by Shapley algorithm to obtain the average power correlation coefficient of each power station in the cluster, and the representative power station for power prediction in the cluster is selected based on the value. Please refer to Example 2 for the specific calculation of the correlation coefficient, which will not be repeated in this embodiment.

[0098] For details, please refer to Figure 4 As shown, weighted calculation is performed on the correlation coefficients of the plurality of analysis objects to obtain a mean value of the correlation coefficients, and a representative power station for cluster power prediction is selected according to the mean value of the correlation coefficients, including:

[0099] S31, calculating a weight standard by using the Shapley algorithm according to the correlation coefficient;

[0100] The calculation of S31 is based on the difference between the correlation coefficient and 1 as the weight calculation standard, and the formula satisfies:

[0101]

[0102] S32, calculating the Shapley value of the correlation coefficient according to the weight standard;

[0103] The calculation of Shapley value satisfies:

[0104]

[0105] S33, calculating the weight of the correlation coefficient according to the Shapley value;

[0106] The weight calculation satisfies:

[0107]

[0108] Among them, σ p , σ k and σ s are the weights of Pearson correlation coefficient, Kendall correlation coefficient and Spearman correlation coefficient respectively.

[0109] S34. Calculate the mean value of the correlation coefficient according to the weights, and select the power plant with the largest mean value of the correlation coefficient as the representative power plant for cluster prediction.

[0110] The above mean calculation satisfies:

[0111]

[0112] Embodiment 4

[0113] The fourth embodiment explains and illustrates the integrated power model, and illustrates the effect in combination with specific experimental results.

[0114] Please refer to Figure 5 As shown, the prediction model is the LSTM-1DCNN power prediction model integrated by CatBoost.

[0115] CatBoost is an improved algorithm based on Gradient Boosting Decision Tree (GBDT). GBDT uses CART regression decision tree as the base learner, trains a series of learners through the Boosting method, and accumulates the outputs of all learners as the final result. The basic principles of the algorithm are as follows:

[0116] For a given training set D = (x k ,y k ) k＝1,2,···n , where x k =(x k 1 ,x k 2 ,···,x k m ) is an m-dimensional input feature, y kis a corresponding numerical label. The strong learner generated by the previous step of training is F t-1 , then the training target of this round is a tree h obtained from the CART tree set H t , so that the expectation E(·) of the loss function L(·) is minimized, that is:

[0117] h t =arg minEL(y, F t-1 (x)+h t (x)

[0118] In the formula, (x, y) is a test sample independent of the training set; GBDT uses the negative gradient of the loss function to fit the CART decision tree h for this round of training. t After T rounds of iterations, the seed model F is obtained as shown below.

[0119]

[0120] In the formula, F 0 represents the initial learner; a t Represents the training step size of the tth round.

[0121] The inheritance algorithm makes the following improvements based on GBDT:

[0122] 1) The ordered boosting technology is introduced to solve the prediction shift problem of GBDT.

[0123] 2) The ordered target statistics (Ordered TS) method is introduced through online learning ideas to directly convert categorical features into numerical statistics, which increases the direct support for categorical features.

[0124] 3) A symmetric binary tree is selected as the base learner, and the same splitting standard is implemented on each layer. This is not easy to cause overfitting, and the stability and prediction speed of the model are also significantly improved.

[0125] Based on this, an LSTM-1DCNN power prediction model integrated with CatBoost is constructed. This model can well integrate the advantages of the two networks, further optimize the prediction results, and improve the prediction accuracy.

[0126] The prediction accuracy of using LSTM or 1DCNN network alone is slightly lower than that of the model after CatBoost integration. Therefore, the prediction model of this embodiment has good accuracy and reliability.

[0127] Combined with the specific methods described in the four embodiments of the present invention, the Pearson correlation coefficient, Kendall correlation coefficient and Spearman correlation coefficient of the spatial output characteristics between power stations within the cluster are calculated, and the Shapley algorithm is used to perform weighted calculation on the correlation coefficient to select the reference power station with the highest correlation; and the representative power station power is predicted by the LSTM-1DCNN network model integrated by CatBoost, and finally the cluster power is derived using the correlation coefficient and capacity ratio. This method effectively solves the problem of insufficient cluster power prediction data, and more reliable prediction results can be obtained by selecting representative power stations and integrating prediction models, further improving the prediction accuracy.

[0128] The beneficial effects of this method will be described below in combination with experimental data.

[0129] Please refer to Figure 5 As shown in the figure, the historical irradiance, temperature, humidity and historical power data of the representative photovoltaic power station are extracted from the photovoltaic database, and the 1DCNN and LSTM are trained respectively using the historical 30-day data. The trained models are integrated using CatBoost to construct an LSTM-1DCNN integrated model to predict the data on the 31st day.

[0130] Figure 6 A comparison chart of representative power station prediction results of different prediction schemes under different weather types according to an embodiment of the present invention is shown, and it can be analyzed that the prediction model adopted in this article has a better effect. Figure 7 This is the result diagram of the selection of the representative power station of the photovoltaic cluster in the experiment. From the results, it can be seen that the No. 4 power station is the representative power station of the cluster. Table 1 is a comparison table of the prediction results under different weather types when different representative power stations are selected. It can be seen that the representative power station selection scheme adopted by this method has a certain reliability; the prediction results of the representative power station selected by this method under different weather types are the most accurate.

[0131] Table 1

[0132] Weather Type No. 1 No. 2 No.3 No. 4 clear 7.92 8.37 3.84 3.06 shower 9.64 6.40 11.01 4.69 partly cloudy 7.94 5.89 3.52 2.22 moderate rain 3.39 3.32 3.36 2.79

[0133] Table 2 is a comparison table of cluster power prediction results under different weather types. It can be seen that the selection of LSTM-1DCNN can not only have good prediction accuracy in sunny weather, but also have good prediction effect in rainy weather, which further proves that the model has certain feasibility and accuracy. Among them, the error of shower weather is relatively low because its fluctuation amplitude is significantly greater than that of the other three weather types, which is not conducive to the learning of 1DCNN network, and thus affects the prediction accuracy of the integrated model.

[0134] Table 2

[0135] Weather Type clear shower partly cloudy moderate rain RMSE / % 3.32 9.70 4.85 2.79

[0136] Embodiment 5

[0137] Embodiment 5 discloses a device corresponding to the photovoltaic cluster power prediction method of the above embodiment, which is a virtual device structure of the above embodiment. Please refer to Figure 8 As shown, including:

[0138] A construction module 310 is used to construct a correlation analysis object of multiple single power station clusters;

[0139] The analysis module 320 is used to respectively calculate the power correlation coefficient corresponding to the correlation analysis object under each preset correlation coefficient; perform weighted calculation on the correlation coefficients of multiple analysis objects to obtain a mean value of the correlation coefficients, and select a representative power station for power prediction according to the mean value of the correlation coefficients;

[0140] The prediction module 330 is used to input the power prediction representative power station into the power prediction model integrated by LSTM and 1DCNN to obtain the power prediction result of the representative power station; and extrapolate the power prediction result to obtain the total power prediction result of the photovoltaic cluster.

[0141] Preferably, the correlation analysis objects include single power station-other single power stations, single power station-total cluster, single power station-cluster excluding corresponding single power station.

[0142] Preferably, the preset correlation coefficients include Pearson correlation coefficient, Kendall correlation coefficient and Spearman correlation coefficient, and the power correlation coefficient is obtained by substituting the power of a single power station, the total power of a cluster and the total power of a cluster without a single power station into the preset correlation coefficient calculation formula.

[0143] Preferably, when the correlation analysis object is a single power station-other single power stations, after calculating the power correlation coefficient, it also includes: calculating the mean of multiple correlation coefficients of different single power stations in the cluster-other single power stations to obtain the total characteristics of the correlation analysis object.

[0144] Preferably, weighted calculation is performed on the correlation coefficients of the plurality of analysis objects to obtain a mean value of the correlation coefficients, and a representative power station for power prediction is selected according to the mean value of the correlation coefficients, including:

[0145] According to the correlation coefficient, a weight standard is calculated by Shapley algorithm;

[0146] Calculating the Shapley value of the correlation coefficient according to the weight standard;

[0147] Calculating the weight of the correlation coefficient according to the Shapley value;

[0148] According to the weights, the mean value of the correlation coefficients is calculated, and the power station with the largest mean value of the correlation coefficients is selected as the predicted representative power station.

[0149] Preferably, the prediction model is an LSTM-1DCNN power prediction model integrated by CatBoost.

[0150] Preferably, the power forecast is extrapolated to obtain the total power forecast of the photovoltaic cluster, and the extrapolation satisfies the formula:

[0151]

[0152] Among them, P set is the cluster power prediction, P ref is the reference power station power, C set is the total capacity of the cluster, C ref is the reference power station capacity, P ave is the mean correlation coefficient.

[0153] Embodiment 6

[0154] Fig. 9 A structural diagram of an electronic device provided in Embodiment 6 of the present invention is shown in FIG. Fig. 9 As shown, the electronic device includes a processor 410, a memory 420, an input device 430, and an output device 440; the number of processors 410 in the computer device can be one or more. Fig. 9 A processor 410 is taken as an example; the processor 410, the memory 420, the input device 430 and the output device 440 in the electronic device can be connected via a bus or other means. Fig. 9 The example of connecting through bus is taken in the following.

[0155] The memory 420 is a computer-readable storage medium that can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the photovoltaic cluster power prediction method in the embodiment of the present invention. The processor 410 executes various functional applications and data processing of the electronic device by running the software programs, instructions and modules stored in the memory 420, that is, the photovoltaic cluster power prediction method of the above-mentioned embodiments 1 to 4 is implemented.

[0156] The memory 420 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 420 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 420 may further include a memory remotely arranged relative to the processor 410, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0157] The input device 430 may be used to receive input user identity information, power station data, etc. The output device 440 may include a display device such as a display screen.

[0158] Embodiment 7

[0159] Embodiment 7 of the present invention further provides a storage medium containing computer executable instructions, and the storage medium can be used for a computer to execute a photovoltaic cluster power prediction method, the method comprising:

[0160] Construct correlation analysis objects for multiple single power station clusters;

[0161] Calculate the power correlation coefficient corresponding to the correlation analysis object under each preset correlation coefficient respectively;

[0162] Performing weighted calculation on the correlation coefficients of the plurality of analysis objects to obtain a mean value of the correlation coefficients, and selecting a representative power station for cluster power prediction according to the mean value of the correlation coefficients;

[0163] Inputting the cluster power forecast representative power station historical power and meteorological data into the power forecast model integrated with LSTM and 1DCNN to obtain power forecast;

[0164] The power prediction result of the representative power station is extrapolated to obtain the total power prediction result of the photovoltaic cluster.

[0165] Of course, the computer executable instructions of a storage medium including computer executable instructions provided in an embodiment of the present invention are not limited to the method operations described above, and can also execute related operations in the photovoltaic cluster power prediction method provided in any embodiment of the present invention.

[0166] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for an electronic device (which can be a mobile phone, a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0167] It is worth noting that in the above-mentioned embodiment of the photovoltaic cluster power prediction method device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0168] For those skilled in the art, various other corresponding changes and deformations can be made according to the technical solutions and concepts described above, and all of these changes and deformations should fall within the protection scope of the claims of the present invention.

Claims

1. A photovoltaic cluster power prediction method, characterized in that: The following steps are involved: Construct correlation analysis objects for multiple single power station clusters; Calculate the power correlation coefficient corresponding to the correlation analysis object under each preset correlation coefficient respectively; The weighted calculation is performed on the correlation coefficients of the plurality of analysis objects to obtain a mean value of the correlation coefficients, and a representative power station for cluster power prediction is selected according to the mean value of the correlation coefficients; wherein the weighted calculation is performed on the correlation coefficients of the plurality of analysis objects to obtain a mean value of the correlation coefficients, and a representative power station for power prediction is selected according to the mean value of the correlation coefficients, including: According to the correlation coefficient, a weight standard is calculated by Shapley algorithm; Calculating the Shapley value of the correlation coefficient according to the weight standard; Calculating the weight of the correlation coefficient according to the Shapley value; According to the weights, a mean value of the correlation coefficient is calculated, and a power plant with the largest mean value of the correlation coefficient is selected as a representative power plant for cluster power prediction; Inputting the historical power and meteorological data of the power forecast representative power station into the power forecast model integrated with LSTM and 1DCNN to obtain the power forecast result of the power forecast representative power station; The power prediction result of the representative power station is extrapolated to obtain the total power prediction result of the photovoltaic cluster.

2. The photovoltaic cluster power prediction method according to claim 1, characterized in that: The correlation analysis objects include single power station-other single power stations, single power station-total cluster, single power station-cluster excluding corresponding single power station.

3. The photovoltaic cluster power prediction method according to claim 1 or 2, characterized in that: The preset correlation coefficients include Pearson correlation coefficient, Kendall correlation coefficient and Spearman correlation coefficient, and the power correlation coefficient is obtained by substituting the power of a single power station, the total power of a cluster and the total power of a cluster without a single power station into the preset correlation coefficient calculation formula.

4. The photovoltaic cluster power prediction method according to claim 2, characterized in that: When the correlation analysis object is a single power station-other single power stations, after calculating the power correlation coefficient, it also includes: calculating the mean of multiple correlation coefficients of different single power stations in the cluster-other single power stations to obtain the total characteristics of the correlation analysis object.

5. The photovoltaic cluster power prediction method according to claim 1, characterized in that: The prediction model is an LSTM-1DCNN power prediction model integrated by CatBoost.

6. The photovoltaic cluster power prediction method according to claim 1, characterized in that: The total power prediction result of the photovoltaic cluster is obtained by extrapolating the power prediction result of the representative power station, and the extrapolation satisfies the formula: ， in, P set For cluster power prediction, P ref is the reference power station power, C set is the total capacity of the cluster, C ref is the reference power station capacity, P ave is the mean correlation coefficient.

7. A photovoltaic cluster power prediction device, characterized in that: It includes: A construction module is used to construct correlation analysis objects of multiple single power station clusters; An analysis module is used to respectively calculate the power correlation coefficient corresponding to the correlation analysis object under each preset correlation coefficient; weighted calculation is performed on the correlation coefficients of multiple analysis objects to obtain a correlation coefficient mean, and a cluster power prediction representative power station is selected according to the correlation coefficient mean; wherein, weighted calculation is performed on the correlation coefficients of multiple analysis objects to obtain a correlation coefficient mean, and a power prediction representative power station is selected according to the correlation coefficient mean, including: According to the correlation coefficient, a weight standard is calculated by Shapley algorithm; Calculating the Shapley value of the correlation coefficient according to the weight standard; Calculating the weight of the correlation coefficient according to the Shapley value; According to the weights, a mean value of the correlation coefficient is calculated, and a power plant with the largest mean value of the correlation coefficient is selected as a representative power plant for cluster power prediction; The prediction module is used to input the historical power and meteorological data of the cluster power prediction representative power station into the power prediction model integrated with LSTM and 1DCNN to obtain the power prediction result of the power prediction representative power station; and extrapolate the power prediction result of the representative power station to obtain the total power prediction result of the photovoltaic cluster.

8. An electronic device comprising a processor, a storage medium and a computer program, wherein the computer program is stored in the storage medium, wherein: When the computer program is executed by a processor, the photovoltaic cluster power prediction method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the photovoltaic cluster power prediction method according to any one of claims 1 to 6 is implemented.

Images