A data-driven new energy output modeling method

Abstract

A new energy output modeling method based on data driving, comprising the following steps: acquiring output data of different new energy units in a new energy station; using Lagrange multi-item interpolation method to interpolate and fill in the missing output data; establishing a mathematical model of wind turbine power-wind speed and a mathematical model of photovoltaic unit power-illumination intensity; fitting the parameters of the mathematical model, using the fitted mathematical model and the Laplace criterion to screen out abnormal power data, and using the fitted mathematical model to fill in the abnormal power data; using K-medoids clustering method to classify new energy units; using principal component analysis method to reduce the wind speed / illumination intensity of different categories of new energy units to one dimension as the characteristic data of the new energy station; and establishing a mathematical model between the total power of the new energy station and the wind speed / illumination intensity. The design not only makes the model simple and the calculation small, but also makes the model high in accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of new energy modeling in power systems, and particularly relates to a new energy output modeling method based on data driving. BACKGROUND

[0002] With the proposal of the double carbon target, more and more new energy machines (Distributed Generator, DG) will be connected to the power grid in the future, and the new energy machines mainly include wind turbines (Wind Turbine, WT) and photovoltaic power generation (Photovoltaic, PV). Due to the randomness of new energy output, it is of great significance to establish the output characteristic model of new energy station for power system optimization scheduling.

[0003] The existing technical route is mostly for the output modeling of one or several wind turbines / photovoltaic units in the new energy station, and does not classify the new energy units at different positions, and aggregates the characteristics of different categories of new energy units to obtain the output characteristics of the entire new energy station. In the process of power system scheduling, the output characteristics of the entire new energy station are generally considered, and if the output characteristics of each new energy unit are considered, the model will be too complex, and the calculation amount will also increase, which makes it difficult to realize large-scale power system day-ahead optimization scheduling and day-ahead optimization scheduling. SUMMARY

[0004] The purpose of the present application is to overcome the defects and problems in the prior art that do not consider clustering new energy units at different geographical positions to establish the output characteristic model of new energy station, and to provide a new energy output modeling method based on data driving which considers clustering new energy units at different geographical positions to establish the output characteristic model of new energy station.

[0005] To achieve the above purpose, the technical solution of the present application is: a new energy output modeling method based on data driving, which comprises the following steps:

[0006] S1, a mathematical model of wind turbine power-wind speed and a mathematical model of photovoltaic unit power-illumination intensity are established;

[0007] S2, the parameters of the mathematical model of wind turbine power-wind speed and the mathematical model of photovoltaic unit power-illumination intensity are fitted, the fitted mathematical model and the Laplace criterion are used to screen out abnormal power data, and the fitted mathematical model is used to fill in the abnormal power data;

[0008] S3, the K-medoids clustering method is used to classify the new energy units;

[0009] S4, using principal component analysis to reduce the wind speed / illumination intensity of different categories of new energy units to one dimension as the characteristic data of the new energy station;

[0010] S5, establishing a mathematical model between the total power of the new energy station and the wind speed / illumination intensity.

[0011] Step S1 specifically comprises the following steps:

[0012] S11, obtaining the output data of different new energy units in the new energy station, including the output power and wind speed of different wind turbines, and the output power and illumination intensity of different photovoltaic panels;

[0013] S12, using Lagrange polynomial interpolation method to interpolate and fill in the missing output data of new energy units;

[0014] S13, drawing the probability density function and cumulative probability distribution function of wind speed of wind turbine at different times, and the probability density function and cumulative probability distribution function of illumination intensity of photovoltaic unit;

[0015] S14, establishing a mathematical model of wind turbine power-wind speed and a mathematical model of photovoltaic unit power-illumination intensity.

[0016] In step S14, the mathematical model of wind turbine power-wind speed is:

[0017]

[0018] In the formula, P WT is the active power output by the wind turbine, P WT，N is the rated power output by the wind turbine, v is the environmental wind speed, v ci is the cut-in wind speed of the wind turbine, v c2 is the cut-out wind speed of the wind turbine, v n is the rated wind speed of the wind turbine, v co is the cut-out wind speed of the wind turbine, and are the parameters to be fitted in the mathematical model;

[0019] The mathematical model of wind turbine power-wind speed satisfies the following conditions:

[0020]

[0021] The mathematical model of photovoltaic unit power-illumination intensity is:

[0022]

[0023] In the formula, P PV is the active power of the photovoltaic unit, P PV，Nis the rated power of the photovoltaic unit, I is the light intensity under natural conditions, I max is the maximum value of the light intensity under natural conditions.

[0024] In step S2, after the fitting process using the existing data, the difference between the fitted predicted data and the true value is used to determine whether the output power of the new energy unit is an abnormal value:

[0025]

[0026]

[0027]

[0028]

[0029] wherein, and are the predicted powers of the new energy unit, the wind turbine and the photovoltaic unit at time t, f WT (·) and f PV (·) are the power models of the wind turbine and the photovoltaic unit fitted, v t is the environmental wind speed at time t, I t is the light intensity under natural conditions at time t, μ DG and σ DG are the error and standard deviation of the new energy unit fitted prediction, T is the total time of the data, is the average value of the output power of the new energy unit, P t，DG is the actual output power of the new energy unit at time t.

[0030] If the predicted power at time t satisfies the following formula, the data is an abnormal value data:

[0031]

[0032] In step S3, the new energy units at different geographical positions are classified into a class based on the wind speed / light intensity, and the similarity index is used as the index for judgment, and the calculation formula of the similarity index is:

[0033]

[0034] D ij = |max(x i )-max(x j )|

[0035]

[0036] wherein, x t，i and respectively are the wind speed / illumination intensity of the ith new energy unit at the tth time and the average wind speed / illumination intensity of the ith new energy unit, x t，j and respectively are the wind speed / illumination intensity of the jth new energy unit at the tth time and the average wind speed / illumination intensity of the jth new energy unit, R(x i , x j ) is the similarity value between the ith new energy unit and the jth new energy unit, x i and x j respectively are the wind speed / illumination intensity of the ith new energy unit and the jth new energy unit, and respectively are the standard deviations of the wind speed / illumination intensity of the ith new energy unit and the jth new energy unit, T is the total number of times of data, D ij and G ij respectively are the fluctuation amplitude feature quantity and the fluctuation amplitude feature quantity, max(·) and min(·) are respectively the maximum value function and the minimum value function;

[0037] Suppose the selected clustering center is Each new energy unit selects the group according to the maximum similarity index, and the formula for updating the clustering center is:

[0038]

[0039]

[0040] In the formula, k is the iteration number of clustering, n is the number of clustering centers, is the wind speed / illumination intensity data set of the new energy unit belonging to the ith class when the clustering number is n and in the kth iteration, and arg max(·) is the state variable x j that makes the function value maximum;

[0041] The new clustering center is obtained Then it is judged whether the following criterion is satisfied:

[0042]

[0043] In the formula, is the 2-norm, σ M is the upper limit of the clustering iteration error allowed;

[0044] When the above criterion is not satisfied, the clustering categories of the new energy units are re-determined and the clustering centers are updated, and the iteration is continuously performed; when the above criterion is satisfied, the clustering center obtained by K-medoids when the clustering number is n is determined as M i，n ; after the above criterion is satisfied, the number of clustering centers is gradually increased, and it is judged whether the following criterion is satisfied:

[0045]

[0046] In the formula, R(x j , M i，n ) is the similarity value between the jth new energy unit and the clustering center, sigma n is the upper limit of the clustering convergence error;

[0047] If the above criterion is not met, the clustering center is increased and the clustering center is randomly selected again; if the above criterion is met, the different new energy units are divided into n classes.

[0048] In step S4, the principal component analysis method is used to reduce the dimension to obtain the index factor with the highest contribution, and the characteristic wind speed / light intensity of the new energy station is obtained.

[0049] In step S5, the power of all new energy units is summed to obtain the total power of the new energy station, and a mathematical model between the total power of the new energy station and the wind speed / light intensity is established through the total power of the new energy station and the characteristic wind speed / light intensity.

[0050] Compared with the prior art, the beneficial effects of the present application are:

[0051] In the new energy output modeling method based on data driving, the data screening method can effectively screen out the abnormal value of the new energy unit power, and can obtain the power-wind speed mathematical model of the wind turbine and the power-illumination intensity mathematical model of the photovoltaic unit; based on the double-weighted correlation coefficient index, the K-medoids clustering method is used to classify the new energy units, which can effectively classify the new energy units in similar geographical positions; the characteristic quantity extracted based on the principal component analysis method can effectively represent the wind speed / illumination intensity characteristics of the entire new energy station; the above design not only makes the model simple and the calculation amount small, but also makes the model accurate. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 It is a flowchart of the new energy output modeling method based on data driving of the present application.

[0053] Figure 2 It is the probability density distribution diagram and the cumulative probability distribution diagram of the wind turbine in the present application.

[0054] Figure 3 It is the power-wind speed fitting model and abnormal value screening schematic diagram of the wind turbine in the present application.

[0055] Figure 4 It is the K-medoids clustering flowchart in the present application.

[0056] Figure 5is the average similarity change curve of the wind turbine in the application.

[0057] Figure 6 is the mathematical model of the power-wind speed of the wind farm station in the application. DETAILED DESCRIPTION

[0058] The application is further described in detail in the following description and specific embodiments in conjunction with the accompanying drawings.

[0059] Referring to Figures 1 to 6 A new energy output modeling method based on data driving, the method comprising the following steps:

[0060] S1, establishing a mathematical model of wind turbine power-wind speed and a mathematical model of photovoltaic unit power-illumination intensity;

[0061] S2, fitting the parameters of the mathematical model of wind turbine power-wind speed and the mathematical model of photovoltaic unit power-illumination intensity, using the fitted mathematical model and the Relyada criterion to screen out abnormal power data, and using the fitted mathematical model to fill in the abnormal power data;

[0062] S3, using the K-medoids clustering method to classify new energy units;

[0063] S4, using principal component analysis to reduce the wind speed / illumination intensity of different categories of new energy units to one dimension as the feature data of the new energy station;

[0064] S5, establishing a mathematical model between the total power of the new energy station and the wind speed / illumination intensity.

[0065] Step S1 specifically comprises the following steps:

[0066] S11, obtaining the output data of different new energy units in the new energy station, including the output power and wind speed of different wind turbines, and the output power and illumination intensity of different photovoltaic panels;

[0067] S12, using the Lagrange polynomial interpolation method to interpolate and fill in the missing output data of the new energy units;

[0068] S13, drawing the probability density function and cumulative probability distribution function of the wind speed of the wind turbine at different times, and the probability density function and cumulative probability distribution function of the illumination intensity of the photovoltaic unit;

[0069] S14, establishing a mathematical model of wind turbine power-wind speed and a mathematical model of photovoltaic unit power-illumination intensity.

[0070] In step S14, the mathematical model of wind turbine power-wind speed is:

[0071]

[0072] In the formula, P WT P represents the active power output of the wind turbine. WT，N v is the rated power output of the wind turbine, v is the ambient wind speed, v ci v is the cut-in wind speed of the wind turbine. c2 To change the wind speed, v n For the rated wind speed, v co This refers to the cut-off wind speed of the wind turbine. and These are the parameters to be fitted to the mathematical model;

[0073] The mathematical model of wind turbine power-wind speed satisfies the following conditions:

[0074]

[0075] The mathematical model for photovoltaic unit power versus solar irradiance is as follows:

[0076]

[0077] In the formula, P PV P represents the active power of the photovoltaic unit. PV，N I is the rated power of the photovoltaic unit, and I is the solar irradiance under natural conditions. max This represents the maximum light intensity under natural conditions.

[0078] In step S2, after fitting the existing data, the difference between the fitted predicted data and the actual value is used to determine whether the output power of the new energy unit is an outlier.

[0079]

[0080]

[0081]

[0082]

[0083] In the formula, and f represents the predicted power output of the renewable energy unit, wind turbine, and photovoltaic unit at time t, respectively. WT (·) and f PV (·) represent the power models fitted to the wind turbine and photovoltaic unit, respectively, v t Let I be the ambient wind speed at time t. t Let μ be the light intensity under natural conditions at time t. DG and σ DGRespectively, the error and standard deviation of the new energy unit fitting prediction, T is the total time number of data, P is the average value of the new energy unit output power, t，DG P is the actual output power of the new energy unit at time t;

[0084] If the predicted power at time t satisfies the following formula, the data is abnormal value data:

[0085]

[0086] In step S3, the new energy units in different geographical positions are classified into a category based on wind speed / illumination intensity, and the index for judgment is the similarity index, and the calculation formula of the similarity index is:

[0087]

[0088] D ij =|max(x i )-max(x j )|

[0089]

[0090] In the formula, x t，i and are the wind speed / illumination intensity of the ith new energy unit at time t and the average wind speed / illumination intensity of the ith new energy unit, x t，j and are the wind speed / illumination intensity of the jth new energy unit at time t and the average wind speed / illumination intensity of the jth new energy unit, R(x i , x j ) is the similarity value between the ith new energy unit and the jth new energy unit, x i and x j are the wind speed / illumination intensity of the ith new energy unit and the jth new energy unit, and are the standard deviations of the wind speed / illumination intensity of the ith new energy unit and the jth new energy unit, T is the total time number of data, D ij and G ij are the fluctuation amplitude characteristic quantity and the fluctuation amplitude characteristic quantity, max(·) and min(·) are the maximum value function and the minimum value function respectively;

[0091] Suppose the selected clustering center is Each new energy unit selects the group according to the maximum similarity index, and the formula for updating the clustering center is:

[0092]

[0093]

[0094] wherein k is the iteration number of clustering, n is the number of clustering centers, is the wind speed / illumination intensity data set of the new energy unit belonging to the i-th cluster when the number of clusters is n and in the k-th iteration, arg max(·) is to find the state variable x that makes the function value maximum j ;

[0095] new clustering centers are obtained whether the following criterion is satisfied is judged

[0096]

[0097] wherein is the 2-norm, σ M is the upper limit of clustering iteration error;

[0098] When the above criterion is not satisfied, the clustering categories of the new energy units are re-determined and the clustering centers are updated, and iteration is continuously carried out; when the above criterion is satisfied, the clustering centers obtained by K-medoids when the number of clusters is n are determined as M i，n ; when the above criterion is satisfied, the number of clustering centers is gradually increased, and whether the following criterion is satisfied is judged:

[0099]

[0100] wherein R(x j , M i，n ) is the similarity value between the j-th new energy unit and the clustering center, σ n is the upper limit of clustering convergence error;

[0101] If the above criterion is not satisfied, the clustering center is increased and the clustering center is randomly selected again; if the above criterion is satisfied, the different new energy units are divided into n classes.

[0102] In step S4, the principal component analysis method is used to reduce the dimension to obtain the index factor with the highest contribution degree, and the characteristic wind speed / illumination intensity of the new energy station is obtained.

[0103] In step S5, the power of all new energy units is summed to obtain the total power of the new energy station, and a mathematical model between the total power of the new energy station and the wind speed / illumination intensity is established through the total power of the new energy station and the characteristic wind speed / illumination intensity.

[0104] The principle of the application is explained as follows:

[0105] In view of the defects of the prior art, the purpose of the present application is to provide a new energy output modeling method based on data driving, aiming to solve the problem that the prior art does not consider clustering new energy units in different geographical positions to establish the output characteristic model of the entire new energy station. The present application first obtains the output of different units in the new energy station, including the output power and wind speed of different wind generators and the output power and light intensity of different photovoltaic panels; the missing data is interpolated using the Lagrange interpolation; the mathematical model of the wind turbine output power-wind speed and the mathematical model of the photovoltaic output power-light intensity are established according to existing experience; the mathematical model parameters are fitted using the obtained data; the fitted model and the Laplace criterion are used to remove the abnormal output data of the new energy, and the power value obtained by fitting is used to fill in; the correlation coefficient matrix of the wind speed of different wind turbines and the light intensity of different photovoltaic panel blocks is established and double weighted to obtain the similarity index of different wind turbines and photovoltaic panel blocks; the K-medoids is used to cluster the wind turbines and photovoltaic panel blocks to obtain the fluctuation curves of different types of typical wind speed / light intensity; the characteristic wind speed / light intensity of the new energy station is obtained by dimension reduction using the principal component analysis method; and the mathematical model between the total power of the new energy station and the wind speed / light intensity is established.

[0106] Embodiment:

[0107] Referring to Figure 1 A new energy output modeling method based on data driving, the method comprising the following steps:

[0108] S1, a mathematical model of wind turbine power-wind speed and a mathematical model of photovoltaic unit power-light intensity are established; specifically comprising the following steps:

[0109] S11, the output data of different new energy units in the new energy station is obtained, including the output power and wind speed of different wind generators and the output power and light intensity of different photovoltaic panels;

[0110] S12, the missing output data of the new energy unit is interpolated and filled in using the Lagrange interpolation method;

[0111] S13, the probability density function and the cumulative probability distribution function of the wind speed of the wind turbine at different times, and the probability density function and the cumulative probability distribution function of the light intensity of the photovoltaic unit are plotted; Figure 2 The probability density distribution graph and the cumulative probability distribution graph of a certain wind turbine in the wind farm at the 23rd time are shown;

[0112] S14, a mathematical model of wind turbine power-wind speed and a mathematical model of photovoltaic unit power-light intensity are established;

[0113] The mathematical model of wind turbine power-wind speed is:

[0114]

[0115] P = P WT P is the active power output by the wind turbine, P WT，N P is the rated power output by the wind turbine, v is the ambient wind speed, v ci v is the cut-in wind speed of the wind turbine, v c2 v is the cut-out wind speed of the wind turbine, v n v is the rated wind speed of the wind turbine, v co v is the cut-out wind speed of the wind turbine, and are the parameters to be fitted in the mathematical model;

[0116] The mathematical model of the wind turbine power-wind speed satisfies the following conditions:

[0117]

[0118] The mathematical model of the photovoltaic power-illumination intensity is:

[0119]

[0120] P = P PV P is the active power output by the photovoltaic unit, P PV，N P is the rated power output by the photovoltaic unit, I is the illumination intensity under natural conditions, I max I is the maximum light intensity under natural conditions;

[0121] S2, the parameters of the mathematical model of the wind turbine power-wind speed and the mathematical model of the photovoltaic power-illumination intensity are fitted, the fitted mathematical model is used to screen abnormal power data with the Relyda criterion, and the fitted mathematical model is used to fill in the abnormal power data;

[0122] After the fitting process is performed using the existing data, whether the output power of the new energy unit is an abnormal value is determined according to the difference between the fitted predicted data and the true value:

[0123]

[0124]

[0125]

[0126]

[0127] P = P and are the predicted powers of the new energy unit, the wind turbine and the photovoltaic unit at time t, respectively, wherein the new energy unit includes the wind turbine and the photovoltaic unit; f WT (·) and fPV (·) are the power models fitted for wind turbines and photovoltaic units respectively, v t is the ambient wind speed at time t, I t is the ambient light intensity at time t; μ DG and σ DG are the error and standard deviation of the fitted prediction of new energy units respectively, T is the total number of data, is the average output power of new energy units, P t，DG is the actual output power of new energy units at time t;

[0128] If the predicted power at time t satisfies the following formula, the data is abnormal value data:

[0129]

[0130] Figure 3 is the fitting of the power-wind speed model of a certain wind turbine, abnormal power data is screened out through fitting and the Relyada criterion, and the abnormal power data is filled in by using the fitted mathematical model; From Figure 3 It can be seen that most of the abnormal data are large wind speed and power of 0, indicating that the wind turbine is in planned downtime or downtime due to failure at that time;

[0131] S3, classify new energy units by using K-medoids clustering method; see Figure 4 for specific process.

[0132] Based on wind speed / light intensity, new energy units in different geographical locations are classified into a class, and the index for judgment is the double weighted similarity index-Pearon correlation coefficient, including amplitude weighting and fluctuation amplitude weighting, and the calculation formula of the similarity index is:

[0133]

[0134] D ij = |max(x i )-max(x j )|

[0135]

[0136] In the formula, x t，i and are the wind speed / light intensity of the ith new energy unit at time t and the average wind speed / light intensity of the ith new energy unit, x t，j and are the wind speed / light intensity of the jth new energy unit at time t and the average wind speed / light intensity of the jth new energy unit, R(xi , x j is the similarity value between the ith new energy unit and the jth new energy unit, x i and x j are the wind speed / illumination intensity of the ith new energy unit and the jth new energy unit, respectively, and are the standard deviations of the wind speed / illumination intensity of the ith new energy unit and the jth new energy unit, respectively, T is the total number of data time, and D ij and G ij are the fluctuation amplitude feature quantity and the fluctuation amplitude feature quantity, respectively, and max(·) and min(·) are the maximum value function and the minimum value function, respectively;

[0137] Suppose the selected clustering center is Each new energy unit selects the group according to the maximum similarity index, and the formula for updating the clustering center is:

[0138]

[0139]

[0140] In the formula, k is the iteration number of clustering, n is the number of clustering centers, is the wind speed / illumination intensity data set of the new energy unit belonging to the ith class when the clustering number is n and in the kth iteration, and arg max(·) is the state variable x j that makes the function value maximum;

[0141] The new clustering center is obtained, and it is judged whether the following criterion is satisfied:

[0142]

[0143] In the formula, is the 2-norm, and σ M is the upper limit of the clustering iteration error allowed;

[0144] When the above criterion is not satisfied, the clustering categories of the new energy units are re-determined and the clustering centers are updated, and the iteration is continuously performed; when the above criterion is satisfied, the clustering center obtained by K-medoids when the clustering number is n is determined as M i，n ; after the above criterion is satisfied, the number of clustering centers is gradually increased, and it is judged whether the following criterion is satisfied:

[0145]

[0146] In the formula, R(x j , M i，n ) is the similarity value between the jth new energy unit and the clustering center, and σn To cluster convergence criterion error upper limit;

[0147] If the above criterion is not met, increase the cluster center and randomly select the cluster center again; if the above criterion is met, different new energy units are divided into n classes;

[0148] Figure 5 The curve of the average similarity of wind turbine with the number of cluster centers is σ n When σ = 0.85, the cluster number of wind turbine is determined to be five classes;

[0149] S4, using principal component analysis to reduce the dimension of wind speed / light intensity of different categories of new energy units to one dimension as the characteristic data of new energy station;

[0150] Using principal component analysis to reduce the dimension to get the highest contribution index factor and get the characteristic wind speed / light intensity of new energy station;

[0151] S5, establish a mathematical model between the total power of new energy station and wind speed / light intensity;

[0152] Sum the power of all new energy units to get the total power of new energy station, and establish a mathematical model between the total power of new energy station and wind speed / light intensity through the total power of new energy station and characteristic wind speed / light intensity, Figure 6 The mathematical model of the total power of a certain regional wind power station and wind speed is shown.

Claims

1. A data-driven new energy output modeling method, characterized in that, The method comprises the following steps: S1, establishing a mathematical model of wind turbine power-wind speed and a mathematical model of photovoltaic unit power-illumination intensity; specifically comprising the following steps: S11, obtaining output data of different new energy units in the new energy station, including output power and wind speed of different wind turbines, and output power and illumination intensity of different photovoltaic panels; S12, using Lagrange polynomial interpolation method to interpolate and complete the missing output data of the new energy units; S13, drawing the probability density function and cumulative probability distribution function of the wind speed of the wind turbine and the probability density function and cumulative probability distribution function of the illumination intensity of the photovoltaic unit at different times; S14, establishing a mathematical model of wind turbine power-wind speed and a mathematical model of photovoltaic unit power-illumination intensity; The mathematical model of wind turbine power-wind speed is: ； wherein Pout is the active power output by the wind turbine, Pnom is the rated power output by the wind turbine, Venv is the ambient wind speed, Vcutin is the cut-in wind speed of the wind turbine, Vbreak is the break wind speed, Vnom is the rated wind speed, Vcutout is the cut-out wind speed of the wind turbine, , , and are the parameters of the mathematical model to be fitted; The mathematical model of wind turbine power-wind speed satisfies the following conditions: ； The mathematical model of photovoltaic unit power-illumination intensity is: ； wherein Ppv is the active power of the photovoltaic unit, Ppv is the active power of the photovoltaic unit, I is the light intensity under natural conditions, I is the light intensity under natural conditions, S2, fitting the parameters of the mathematical model of wind turbine power-wind speed and the mathematical model of photovoltaic unit power-illumination intensity, using the fitted mathematical model and the Laplace criterion to screen out abnormal power data, and using the fitted mathematical model to complete the abnormal power data; After the fitting process using the existing data, the difference between the fitted predicted data and the true value is used to determine whether the output power of the new energy unit is an abnormal value: ； ； ； ； In the formula, , and They are respectively The predicted power output of new energy units, wind turbines, and photovoltaic units at all times. and The power models fitted to wind turbines and photovoltaic units are shown below. for Real-time ambient wind speed, for Light intensity under natural conditions at any given time and These represent the error and standard deviation of the fitting prediction for new energy power units, respectively. The total number of time points in the data. This represents the average output power of the new energy generating units. for The actual output power of the new energy unit at all times; If The data is abnormal value data if the power predicted at the moment satisfies the following formula: ； S3, using K-medoids clustering method to classify the new energy units; S4, using principal component analysis method to reduce the wind speed / illumination intensity of different categories of new energy units to one dimension as the feature data of the new energy station; S5, establishing a mathematical model between the total power of the new energy station and the wind speed / illumination intensity.

2. The new energy output modeling method based on data driving according to claim 1, wherein: In step S3, the new energy units at different geographical positions are classified into a category based on wind speed / illumination intensity, and the similarity index is used as the judgment index, and the calculation formula of the similarity index is: ； ； ； In the formula, and The first Taiwan's new energy units in the first Wind speed / light intensity at the first moment and the second moment Average wind speed / solar intensity of Taiwan's new energy power units and The first Taiwan's new energy units in the first Wind speed / light intensity at the first moment and the second moment Average wind speed / solar intensity of Taiwan's new energy power units For the first Taiwan's new energy units and the first Similarity values ​​between new energy power generation units in Taiwan and The first Taiwan New Energy Unit and the first Wind speed / solar intensity of Taiwan's new energy units and The first Taiwan New Energy Unit and the first Standard deviation of wind speed / solar intensity for Taiwan's new energy power units The total number of time points in the data. and These are the fluctuation amplitude characteristic and the fluctuation range characteristic, respectively. and These are the maximum value function and the minimum value function, respectively. Assuming the selected clustering center is Each new energy unit selects the group according to the maximum similarity index, and the formula for updating the clustering center is: ； ； In the formula, is the number of iterations of clustering, is the number of cluster centers, is the number of clusters when the number of clusters is and the wind speed / illumination intensity data set of the new energy unit belonging to the th cluster at the th iteration, is the state variable for finding the maximum function value ; obtaining new cluster centers post-determining whether the following criterion is satisfied: ； wherein is the 2-norm, is the upper limit of the clustering iteration error allowance; When the above criterion is not satisfied, the cluster category of the new energy unit is re-determined and the cluster center is updated, and iteration is continuously carried out; when the above criterion is satisfied, the cluster number is determined as The cluster center obtained by K-medoids is When the above criterion is satisfied, the cluster center number is gradually increased, and it is judged whether the following criterion is satisfied: ； In the formula, is the first The similarity value between the Taixin energy unit and the cluster center, is the upper limit of the clustering convergence error If the above criterion is not satisfied, the cluster center is increased and the cluster center is randomly selected again; if the above criterion is satisfied, different new energy units are divided into classes.

3. The new energy output modeling method based on data driving according to claim 2, wherein: In step S4, the principal component analysis method is used to reduce the dimension to obtain the index factor with the highest contribution degree, and the characteristic wind speed / illumination intensity of the new energy station is obtained.

4. The new energy output modeling method based on data driving according to claim 3, wherein: In step S5, the power of all new energy units is summed to obtain the total power of the new energy station, and a mathematical model between the total power of the new energy station and the wind speed / illumination intensity is established.

Images