A wind power combination prediction method

Abstract

The present invention discloses a wind power combination prediction method, comprising the following steps: first, taking a weight sequence as a variable, taking the weighted combination of a single wind power prediction value and the minimum difference between the real value as the target, constructing a weight sequence optimization configuration model, and using an intelligent algorithm to obtain the weight sequence configuration result; second, taking the wind power output value of two adjacent time periods and the weight sequence configuration result of the corresponding time period obtained based on the weight sequence optimization configuration model as input and output, training and obtaining a weight sequence prediction model; finally, using the trained weight sequence prediction model and the single wind power prediction model to predict the wind power in the future time period. The wind power combination prediction method provided by the present invention predicts wind power based on the weight sequence optimization configuration model and the weight sequence prediction model, simplifies the calculation process while ensuring the accuracy of the calculation result, and is conducive to improving the wind power grid-connected capacity and improving the stability of the wind power grid-connected system operation.

Description

Technical Field

[0001] The invention relates to the technical field of wind power prediction in wind farms, and in particular to a wind power combination prediction method. Background Art

[0002] According to incomplete statistics, by 2015, the cumulative installed capacity of wind power in my country had reached 58,249MW. However, due to the volatility and randomness of wind energy resources, wind power output has a relatively low controllability, which may bring many disadvantages to the safe and stable control of the power system and its optimized dispatching, and on the other hand, it will also cause a sharp drop in the friendly grid-connected capacity of wind power. High-precision wind power prediction is one of the effective measures to improve its friendly grid-connected capacity and ensure the safety and stability of its grid-connected system.

[0003] At present, the wind power prediction methods at home and abroad are mainly divided into the following two categories:

[0004] (1) Single wind power prediction methods, such as duration method, auto-regressive and moving average model (ARMA) method, Kalman filter method, artificial neural network method, wavelet analysis method, support vector machine (SVM) regression method, etc. However, due to the inherent limitations of various single prediction methods, their prediction accuracy is low and their generalization ability is also poor;

[0005] (2) Wind power combination prediction methods. For example, Reference 1 (Chen Ning, Sha Qian, Tang Yi, et al. Wind power combination prediction method based on cross entropy theory [J]. Proceedings of the CSEE, 2012, 32(4): 29-34) introduces cross entropy into the wind power prediction process and determines the weights of different single wind power prediction methods based on the degree of pairwise interaction between different information sources. For example, Reference 2 (Liu Chun, Fan Gaofeng, Wang Weisheng, et al. Combined prediction model for wind farm output power [J]. Power System Technology, 2009(13): 74-79) uses three methods, namely, equal weight averaging method, covariance optimization method and time-varying weight coefficient method, to determine the weights of different single wind power prediction methods. Although the above-mentioned existing wind power combination prediction methods have overcome the inherent limitations of single wind power prediction methods, it is difficult to balance the problem between the amount of calculation and the actuarial accuracy in the process of obtaining the weights of each single wind power prediction method. For example, Reference 1 uses a disk algorithm to calculate the weights of each single wind power prediction method. Although the calculation results are relatively ideal, the calculation process is relatively complicated. The three methods of equal weight averaging method, covariance optimization method and time-varying weight coefficient given in Reference 2 reduce the complexity of the calculation process, but it is difficult to guarantee the accuracy of the calculation results. Summary of the invention

[0006] The purpose of the present invention is to provide a wind power combination prediction method, which simplifies the calculation process while ensuring the accuracy of the calculation results, overcomes the low prediction accuracy and poor generalization ability of the existing single wind power prediction method and the problem that it is difficult to balance the calculation amount and actuarial accuracy in the existing wind power combination prediction method, which is beneficial to improving the wind power grid-connected capacity and improving the stability of the wind power grid-connected system operation.

[0007] To achieve the above object, the present invention provides the following solutions:

[0008] A wind power combination prediction method, the method comprising the following steps:

[0009] S1, constructing a weight sequence optimization configuration model: taking the weight sequence as a variable and the minimum difference between the weighted combination of a single wind power forecast value and the actual value as the goal, construct a weight sequence optimization configuration model, and use an intelligent algorithm to obtain the weight sequence configuration result;

[0010] S2, establishing a weight sequence prediction model: taking the wind power output values ​​of two adjacent time periods and the weight sequence configuration results of the corresponding time periods obtained based on the above-mentioned weight sequence optimization configuration model as input and output, training and obtaining the weight sequence prediction model;

[0011] S3, predicting wind power: using the trained weight sequence prediction model and the single wind power prediction model to predict the wind power in the future period.

[0012] Furthermore, in step S1, a weight sequence optimization configuration model is constructed with the weight sequence as a variable and the minimum difference between the weighted combination of a single wind power forecast value and the actual value as the target, and an intelligent algorithm is used to obtain the weight sequence configuration result, which specifically includes:

[0013] S101, the actual value of wind power output P in period k-1 k-1 As samples, x row and column vectors are formed, and n single wind power prediction methods are used to predict the wind power output values ​​of adjacent k time periods respectively. are all y row and column vectors, and the wind power output combination prediction matrix p' for this period is obtained from this k , p' k is a matrix of y rows and n columns, as shown in Formula 1:

[0014]

[0015] The combined prediction matrix p' k Divide by row to get y n-column row vectors, the division is shown in formula 2:

[0016]

[0017] S102, with n rows and y columns matrix w k is a variable, which is divided into columns to obtain y n-row and column vectors Calculate and obtain the combined wind power output forecast result p during the k period k , p k is an n-order square matrix, such as

[0018] As shown in formula 3:

[0019]

[0020] S103: The real value sequence P of wind power output in period k k Combined forecast result with wind power output p k Make a difference at the corresponding time point and take the minimum sum of their absolute values ​​as the goal. The weight sequence optimization configuration model constructed is as follows

[0021] As shown in formula 4:

[0022]

[0023] S104, using an intelligent algorithm to solve Formula 4, where the row and column vectors are obtained is the weight sequence.

[0024] Furthermore, in step S2, the wind power output values ​​of two adjacent time periods and the weight sequence configuration results of the corresponding time periods obtained based on the above-mentioned weight sequence optimization configuration model are used as input and output to train and obtain the weight sequence prediction model, which specifically includes:

[0025] S201, according to the weight sequence optimization configuration model and its solution method described in steps S101-S103, calculate the weight matrix w of time period k and time period k+1 k With w k+1 ;

[0026] S202, the actual value of wind power output P in period k k And the weight matrix w of the corresponding period k As input, the actual value of wind power output P in period k+1 k+1 And the weight matrix w of the corresponding period k+1 As the output, the BP neural network is trained to obtain the weight sequence prediction model.

[0027] Furthermore, in step S3, the trained weight sequence prediction model and the single wind power prediction model are used to predict the wind power in the future period, specifically including:

[0028] S301, the actual value of wind power output P in period k+1k+1 As samples, the above n single wind power prediction methods are used to predict the k+2 time periods. And use the above formula 1 and formula 2 to calculate y n-column row vectors

[0029] S302, still taking the actual wind power output value P in period k+1 k+1 The weighted sequence prediction model described in steps S201-S202 is used as a sample to calculate and obtain the weighted prediction matrix w for the k+2 period. k+2 , divide it by column to get y n-row and column vectors

[0030] S303: the wind power prediction results obtained by the n single wind power prediction methods are After weighted combination, the wind power output value p in the k+2 period is obtained k+2 , as shown in Formula 5:

[0031]

[0032] S304, repeating the above steps S301 to S303, predicting and obtaining the wind power output value in the future period.

[0033] Furthermore, the intelligent algorithm is one or more of a GA algorithm, a PSO algorithm and a chaos algorithm.

[0034] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the wind power combination prediction method provided by the present invention first obtains the weight sequence by establishing a weight sequence optimization configuration model; then, the wind power output value of two adjacent time periods is used as input, and the weight sequence is used as output, and the weight sequence prediction model is obtained by training on the basis of the neural network model; finally, the trained weight sequence prediction model and the single wind power prediction model are used to predict the wind power in the future time period, and the inherent regularity between the reasonable combinations of the single wind power prediction results in the adjacent time periods is fully explored. While reducing the problem of large amount of calculation of the weight sequence optimization configuration in the wind power combination prediction, the accuracy of the wind power prediction is also guaranteed, and the method has a strong application value, which is used to improve the wind power grid-connected capacity and improve the stability of the wind power grid-connected system operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] 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 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 labor.

[0036] Figure 1 It is a schematic diagram of the flow chart of the wind power combination prediction method of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] The purpose of the present invention is to provide a wind power combination prediction method, which simplifies the calculation process while ensuring the accuracy of the calculation results, which is beneficial to improving the wind power grid-connected capacity and improving the stability of the wind power grid-connected system operation.

[0039] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] like Figure 1 As shown, the wind power combination prediction method provided by the present invention comprises the following steps:

[0041] S1, build a weight sequence optimization configuration model: take the weight sequence as a variable, and take the minimum difference between the weighted combination of a single wind power forecast value and the actual value as the goal, build a weight sequence optimization configuration model, and use an intelligent algorithm to obtain the weight sequence configuration result; specifically include:

[0042] S101, the actual value of wind power output P in period k-1 k-1 As samples, x row and column vectors are formed, and n single wind power prediction methods are used to predict the wind power output values ​​of adjacent k time periods respectively. are all y row and column vectors, and the wind power output combination prediction matrix p' for this period is obtained from this k , p' k is a matrix of y rows and n columns, as shown in Formula 1:

[0043]

[0044] The combined prediction matrix p' k Divide by row to get y n-column row vectors, the division is shown in formula 2:

[0045]

[0046] S102, with n rows and y columns matrix w k is a variable, which is divided into columns to obtain y n-row and column vectors Calculate and obtain the combined wind power output forecast result p during the k period k , p k is an n-order square matrix, such as

[0047] As shown in formula 3:

[0048]

[0049] S103: The real value sequence P of wind power output in period k k Combined forecast result with wind power output p k Make a difference at the corresponding time point and take the minimum sum of their absolute values ​​as the goal. The weight sequence optimization configuration model constructed is as follows

[0050] As shown in formula 4:

[0051]

[0052] S104, using an intelligent algorithm to solve Formula 4, where the row and column vectors are obtained is the weight sequence;

[0053] S2, establish a weight sequence prediction model: take the wind power output values ​​of two adjacent time periods and the weight sequence configuration results of the corresponding time period obtained based on the above weight sequence optimization configuration model as input and output, and train to obtain the weight sequence prediction model; specifically including:

[0054] S201, according to the weight sequence optimization configuration model and its solution method described in steps S101-S103, calculate the weight matrix w of time period k and time period k+1 k With w k+1 ;

[0055] S202, the actual value of wind power output P in period k k And the weight matrix w of the corresponding period k As input, the actual value of wind power output P in period k+1 k+1 And the weight matrix w of the corresponding period k+1 For output, the BP neural network is trained to obtain the weight sequence prediction model;

[0056] S3, predicting wind power: using the trained weight sequence prediction model and the single wind power prediction model to predict the wind power in the future period, including:

[0057] S301, the actual value of wind power output P in period k+1 k+1 As samples, the above n single wind power prediction methods are used to predict the k+2 time periods. And use the above formula 1 and formula 2 to calculate y n-column row vectors

[0058] S302, still taking the actual wind power output value P in period k+1 k+1 The weighted sequence prediction model described in steps S201-S202 is used as a sample to calculate and obtain the weighted prediction matrix w for the k+2 period. k+2 , divide it by column to get y n-row and column vectors

[0059] S303: the wind power prediction results obtained by the n single wind power prediction methods are After weighted combination, the wind power output value p in the k+2 period is obtained k+2 , as shown in Formula 5:

[0060]

[0061] S304, repeating the above steps S301 to S303, predicting and obtaining the wind power output value in the future period.

[0062] The step S1 completely constructs a k-period weight sequence optimization configuration model. In this model, the k-period wind power output real value sequence P k Combined forecast result with wind power output p k The goal is to minimize the sum of the absolute values ​​of the differences at the corresponding time points, and the weight sequence matrix w k As variables, the global optimization intelligent algorithm (such as GA algorithm, PSO algorithm and chaos algorithm, etc.) can be used to solve the model and calculate the optimal configuration result of the weight sequence.

[0063] In step S2, first, according to the above-mentioned weight sequence optimization configuration model and its solution method, the weight matrix w of the k period and the k+1 period is calculated. k With w k+1 ; Further, the actual value of wind power output P in period k k And the weight matrix w of the corresponding period k As input, the actual value of wind power output P in period k+1 k+1 And the weight matrix w of the corresponding period k+1 For output, the BP neural network is trained to obtain the weight matrix prediction model.

[0064] In step S3, in order to improve the prediction accuracy and adaptability of the wind power combination prediction method proposed in the present invention, the method can be applied in a rolling manner to the wind power prediction of the future k+3, k+4, and k+5 time periods, that is, repeating the above steps 1 to 3 to predict and obtain the wind power output values ​​of the future time periods.

[0065] The wind power combination prediction method provided by the present invention first obtains the weight sequence by establishing a weight sequence optimization configuration model; then, taking the wind power output values ​​of two adjacent time periods as input and the weight sequence as output, the weight sequence prediction model is trained on the basis of the neural network model to obtain the weight sequence prediction model; finally, the trained weight sequence prediction model and the single wind power prediction model are used to predict the wind power in the future time period, and the inherent regularity between the reasonable combinations of the single wind power prediction results in the adjacent time periods is fully explored. While reducing the problem of large amount of calculation of the weight sequence optimization configuration in the wind power combination prediction, the accuracy of the wind power prediction is also guaranteed. The method has strong application value and is used to improve the wind power grid-connected capacity and improve the stability of the wind power grid-connected system operation.

[0066] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A wind power combination prediction method, characterized in that: The following steps are involved: S1, constructing a weight sequence optimization configuration model: taking the weight sequence as a variable and the minimum difference between the weighted combination of a single wind power forecast value and the actual value as the goal, construct a weight sequence optimization configuration model, and use an intelligent algorithm to obtain the weight sequence configuration result; S2, establishing a weight sequence prediction model: taking the wind power output values ​​of two adjacent time periods and the weight sequence configuration results of the corresponding time periods obtained based on the above-mentioned weight sequence optimization configuration model as input and output, training and obtaining the weight sequence prediction model; S3, predicting wind power: using the trained weight sequence prediction model and the single wind power prediction model to predict the wind power in the future period; In step S1, a weight sequence optimization configuration model is constructed with the weight sequence as a variable and the minimum difference between the weighted combination of a single wind power forecast value and the actual value as the target, and an intelligent algorithm is used to obtain the weight sequence optimization configuration result, which specifically includes: S101, the actual value of wind power output P in period k-1 k-1 As samples, x row and column vectors are formed, and n single wind power prediction methods are used to predict the wind power output values ​​of adjacent k time periods respectively. are all y row and column vectors, and the wind power output combination prediction matrix p′ for this period is obtained from this k , p′ k is a matrix of y rows and n columns, as shown in Formula 1: The combined prediction matrix p′ k Divide by row to get y n-column row vectors, the division is shown in formula 2: S102, with n rows and y columns matrix w k is a variable, which is divided into columns to obtain y n-row and column vectors Calculate and obtain the combined wind power output forecast result p during the k period k , p k is an n-order square matrix, as shown in Formula 3: S103: The real value sequence P of wind power output in period k k Combined forecast result with wind power output p k The difference is made at the corresponding time points, and the minimum sum of the absolute values ​​is taken as the goal. The weight sequence optimization configuration model constructed is shown in Formula 4: S104, using an intelligent algorithm to solve Formula 4, where the row and column vectors are obtained is the weight sequence; In step S2, the wind power output values ​​of two adjacent time periods and the weight sequence configuration results of the corresponding time periods obtained based on the weight sequence optimization configuration model are used as input and output to train and obtain the weight sequence prediction model, which specifically includes: S201, according to the weight sequence optimization configuration model and its solution method described in steps S101-S103, calculate the weight matrix w of time period k and time period k+1 k With w k+1 ; S202, the actual value of wind power output P in period k k And the weight matrix w of the corresponding period k As input, the actual value of wind power output P in period k+1 k+1 And the weight matrix w of the corresponding period k+1 For output, the BP neural network is trained to obtain the weight sequence prediction model; In step S3, the trained weight sequence prediction model and the single wind power prediction model are used to predict the wind power in the future period, which specifically includes: S301, the actual value of wind power output P in period k+1 k+1 As samples, the above n single wind power prediction methods are used to predict the k+2 time periods. And use the above formula 1 and formula 2 to calculate y n-column row vectors S302, still taking the actual wind power output value P in period k+1 k+1 The weighted sequence prediction model described in steps S201-S202 is used as a sample to calculate and obtain the weighted prediction matrix w for the k+2 period. k+2 , divide it by column to get y n-row and column vectors S303: the wind power prediction results obtained by the n single wind power prediction methods are After weighted combination, the wind power output value p in the k+2 period is obtained k+2 , as shown in Formula 5: S304, repeating the above steps S301-S303 to predict and obtain the wind power output value in the future period.

Images