Precision improvement method applied to clustering of photovoltaic power generation system

Abstract

The invention discloses a precision improvement method applied to clustering of a photovoltaic power generation system, and belongs to the technical field of equivalent modeling of the photovoltaic power generation system, and the method comprises the following steps: S1, building a mathematical model of a photovoltaic cell; s2, performing theoretical analysis on the mathematical model of the photovoltaic cell to obtain nonlinear characteristics of the maximum power of the photovoltaic cell, the illumination intensity and the temperature; and S3, introducing a correction weight function based on error loss caused by nonlinear characteristics. By adopting the precision improvement method applied to the clustering of the photovoltaic power generation system, the simulation precision of the aggregation model of the photovoltaic power generation system can be effectively improved.

Description

Technical Field

[0001] The invention relates to the technical field of equivalent modeling of photovoltaic power generation systems, and in particular to a precision improvement method applied to photovoltaic power generation system clustering. Background Art

[0002] Based on the distribution of photovoltaic clusters, the equivalent methods of regional centralized photovoltaics mainly include single-machine equivalent method and multi-machine equivalent method. When the environmental conditions of centralized photovoltaics are relatively consistent, the single-machine equivalent method can usually meet the needs due to its simplicity and low accuracy requirements. This method compares the differential equations of the detailed model and the equivalent model, lists the equations of each component, and then compares the parameters to determine the component parameters of the equivalent model. The equivalent model parameters can usually be directly calculated through the formula. The multi-machine equivalent rule seeks a balance between simulation accuracy and efficiency, classifies photovoltaic power generation units through clustering algorithms, and then calculates the equivalent model. When focusing on the dynamic response of the photovoltaic system, the relevant parameters of the inverter, such as the product of the control parameter and the sensitivity, are often selected as clustering indicators. If the requirements for dynamic accuracy are slightly lower and only the steady-state accuracy is considered, environmental parameters such as light intensity and temperature can be used as clustering indicators.

[0003] However, in the process of clustering the light intensity and temperature of the photovoltaic power generation system, it is found that there is a certain error between the output of the aggregation model and the actual system. For large-scale photovoltaic power generation systems, whose power levels are usually in the range of several million kilowatts to tens of megawatts, a small error can actually have a large gap. Therefore, it is very important to further improve the accuracy of the equivalent model. At present, the solution to this problem in this field is generally to propose a more complex clustering algorithm, or to use deep learning and other methods to improve the simulation accuracy of the aggregation model. The implementation process is extremely complicated and inconvenient to implement. Summary of the invention

[0004] The purpose of the present invention is to provide a precision improvement method for photovoltaic power generation system clustering, which can effectively improve the simulation accuracy of the photovoltaic power generation system aggregation model.

[0005] To achieve the above object, the present invention provides a method for improving the accuracy of photovoltaic power generation system clustering, comprising the following steps:

[0006] Step S1, establishing a mathematical model of a photovoltaic cell;

[0007] Step S2, performing theoretical analysis on the mathematical model of the photovoltaic cell to obtain the nonlinear characteristics of the maximum power of the photovoltaic cell, light intensity and temperature;

[0008] Step S3: introducing a modified weight function based on the error loss caused by the nonlinear characteristics.

[0009] Preferably, the mathematical model of the photovoltaic cell in step S1 is as follows:

[0010]

[0011] Where, I represents the output current of the photovoltaic cell; I sc represents the photocurrent of the photovoltaic panel; V represents the output voltage of the photovoltaic cell; V oc Represents the open circuit voltage of the photovoltaic cell; V m Indicates the maximum power point voltage of the photovoltaic cell; I m Indicates the maximum power point current of the photovoltaic cell; I sc represents the short-circuit current of the photovoltaic cell; a, b, c represent the compensation coefficients; S represents the actual light temperature; T represents the actual working temperature; I scref Represents the reference value of photocurrent; I mref Indicates the maximum power point current reference value of the photovoltaic cell; V ocref Indicates the open circuit voltage reference value of the photovoltaic cell; V mref Indicates the maximum power point voltage reference value of the photovoltaic cell; ΔT = TT ref Indicates the difference between the actual temperature and the reference temperature, T ref Indicates the temperature under standard working conditions, which is 25 degrees Celsius; ΔS = SS ref Represents the difference between actual illumination and reference illumination, S ref Indicates the light temperature under standard working conditions, which is 1000W / m 2 ;

[0012] The maximum power of photovoltaic cells is expressed as: P m =V m I m , change V in the mathematical model of photovoltaic cells m ,I m Substitute into the formula P m =V m I m get:

[0013]

[0014] Among them, P m (S,T) represents the maximum power of the photovoltaic cell as a function of light intensity and temperature.

[0015] Preferably, in step S2, the temperature T is set as a constant to obtain the nonlinear characteristic P of the maximum output power of the photovoltaic cell and the light intensity. m (S) is as follows:

[0016]

[0017] Similarly, when the light intensity is stable, the nonlinear characteristic P of the maximum output power and temperature is obtained. m (T) is as follows:

[0018]

[0019] Preferably, the error loss in step S3 is: m (S), and its second-order derivative is:

[0020]

[0021] Among them, in the interval of light intensity S>0, P″ m (S)>0; get function P m (S) is a strictly convex function. For a strictly convex function, if for any x1, x2 and λ∈[0,1], the following inequality is satisfied:

[0022] f(λx1+(1-λ)x2)<λf(x1)+(1-λ)f(x2);

[0023] when When:

[0024]

[0025] Where f(x) represents an arbitrary strictly convex function; the above formula proves that when two data points are clustered, the maximum power corresponding to the cluster center of the light intensity is smaller than the power mean corresponding to the actual two light intensities.

[0026] Preferably, according to the general form of probability theory of the Jensen inequality:

[0027]

[0028] The expected value of a random variable. represents the expectation of the value of a function of a random variable;

[0029] Prove that the following holds,

[0030]

[0031] Among them, x n represents the nth variable; it indicates that for the clustering of any illumination data point, using the illumination intensity of the cluster center as the mean illumination intensity of the entire cluster will reduce the output power of the aggregation model.

[0032] Preferably, the modified weight function w(S) in step S3 is:

[0033] w(S)=1+a·Sln(e+0.0005*(S-1000));

[0034] Among them, α represents the weight function coefficient.

[0035] Therefore, the present invention adopts the above-mentioned precision improvement method applied to photovoltaic power generation system clustering, and through in-depth analysis of the relationship between the maximum power of photovoltaic cells and light intensity, and the introduction of a modified weight function, it can effectively eliminate the error caused by not considering the nonlinear relationship between light intensity and power in the traditional clustering method. This correction greatly improves the simulation accuracy of the photovoltaic power generation system clustering model, especially when the light intensity changes greatly, it can more accurately predict the power output of the photovoltaic system.

[0036] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of an embodiment of a method for improving accuracy of photovoltaic power generation system clustering according to the present invention;

[0038] Figure 2 It is a maximum power output characteristic diagram of a photovoltaic cell in an embodiment of a precision improvement method for photovoltaic power generation system clustering of the present invention; Figure 2 (a) is the maximum power and illumination characteristic curve; Figure 2 (b) is the maximum power and temperature characteristic curve;

[0039] Figure 3 It is a schematic diagram of two illumination data clustering in an embodiment of a precision improvement method for clustering of photovoltaic power generation systems of the present invention;

[0040] Figure 4 It is a schematic diagram of two temperature data clustering in an embodiment of a precision improvement method for clustering of photovoltaic power generation systems of the present invention. DETAILED DESCRIPTION

[0041] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0042] Unless otherwise defined, technical or scientific terms used in the present invention shall have the common meanings understood by one having ordinary skills in the field to which the present invention belongs.

[0043] Embodiment 1

[0044] In this embodiment, the mathematical model of the photovoltaic cell uses twenty photovoltaic power generation units, the aggregation model uses the K-means algorithm, and the number of aggregation equivalent models is three. Figure 1As shown, the present invention provides a precision improvement method for photovoltaic power generation system clustering, comprising the following steps:

[0045] Step S1, establish a mathematical model of photovoltaic cells; photovoltaic cells are the energy source of the entire photovoltaic power generation system, and their output power directly determines the system grid-connected power. When the system is in stable grid-connected operation, due to the MPPT algorithm, it can be considered that all photovoltaic arrays in the entire photovoltaic field station are working at the maximum power point. Therefore, the study of the maximum power characteristics of photovoltaic cells is the key to analyzing the errors generated in the clustering process. The mathematical model of photovoltaic cells is as follows:

[0046]

[0047] Where, I represents the output current of the photovoltaic cell; I sc represents the photocurrent of the photovoltaic panel; V represents the output voltage of the photovoltaic cell; V oc Represents the open circuit voltage of the photovoltaic cell; V m Indicates the maximum power point voltage of the photovoltaic cell; I m Indicates the maximum power point current of the photovoltaic cell; I sc represents the short-circuit current of the photovoltaic cell; a, b, c represent the compensation coefficients; S represents the actual light temperature; T represents the actual working temperature; I scref Represents the reference value of photocurrent; I mref Indicates the maximum power point current reference value of the photovoltaic cell; V ocref Indicates the open circuit voltage reference value of the photovoltaic cell; V mref Indicates the maximum power point voltage reference value of the photovoltaic cell; ΔT = TT ref Indicates the difference between the actual temperature and the reference temperature, T ref Indicates the temperature under standard working conditions, which is 25 degrees Celsius; ΔS = SS ref Represents the difference between actual illumination and reference illumination, S ref Indicates the light temperature under standard working conditions, which is 1000W / m 2 ;

[0048] The maximum power of photovoltaic cells is expressed as: P m =V m I m , change V in the mathematical model of photovoltaic cells m ,I m Substitute into the formula P m =V m I m get:

[0049]

[0050] Among them, P m(S,T) represents the maximum power of the photovoltaic cell as a function of light intensity and temperature.

[0051] Step S2: Theoretically analyze the mathematical model of the photovoltaic cell to obtain the nonlinear characteristics of the maximum power of the photovoltaic cell, the light intensity and the temperature; assuming that the temperature T is a constant, the nonlinear characteristics P of the maximum output power of the photovoltaic cell and the light intensity are obtained. m (S) is as follows:

[0052]

[0053] From the above formula, we can see that when the temperature T is constant and the coefficient k1 is a constant, there is an obvious nonlinear relationship between the maximum power of the photovoltaic cell and the light intensity. Similarly, when the light intensity is stable, the nonlinear characteristic P of the maximum output power and temperature can be obtained. m (T) is as follows:

[0054]

[0055] It can be seen that the maximum power of photovoltaic cells has a nonlinear relationship with light intensity and temperature, and this nonlinearity depends on the characteristics of the photovoltaic cell model and is affected by the compensation coefficients a, b, and c, and has nothing to do with the specific photovoltaic cell model. Therefore, a certain model of photovoltaic cell parameters can be selected to study the nonlinear output characteristics of its maximum power, and it is applicable to all photovoltaic cell models.

[0056] When the temperature is 25℃, m (S) and light intensity of 1000W / m 2 P m (T) is expressed as follows:

[0057]

[0058] To more intuitively represent the relationship, the characteristic curve of the maximum power of photovoltaic cells, light intensity and power can be obtained according to the formula as follows: Figure 2 shown.

[0059] Step S3: introducing a modified weight function based on the error loss caused by the nonlinear characteristics.

[0060] When performing clustering, the clustering algorithm defines the cluster center of each cluster as the Euclidean center of the data points in the cluster. For example, the basic principle of the K-means algorithm is to minimize the sum of the squares of the Euclidean distances from the data points to their corresponding cluster centers to achieve data clustering. This data clustering method assumes that all data have the same weight, so it is suitable for linear processing.

[0061] However, in the clustering of photovoltaic systems, there is a nonlinear relationship between the power of photovoltaic cells and the light intensity and temperature. This nonlinear characteristic will introduce errors in the clustering process. Figure 3 As shown in the figure, taking the parameters of a 100KW photovoltaic cell array as an example, assuming that light intensities A and B are two points belonging to the same cluster in a certain cluster, and assuming that the temperatures of the two points are the same, then according to the relationship between the maximum power curve of the photovoltaic cell and the light intensity, its maximum power is obtained, and at the same time, the cluster center is obtained, that is, the midpoint C of the two light intensities.

[0062] The light intensity at point A and point B is 850W / m 2 and 1150W / m 2 The corresponding maximum powers are 83.22kW and 118.985kW respectively, and the total power is 202.178kW.

[0063] If the aggregation equivalent model is used, replace points A and B with a light intensity of 1000W / m 2 The equivalent point C is selected, and the capacity of the photovoltaic cell is adjusted to twice the original. The maximum power at point C is 201.448kW. Compared with the sum of the actual power at points A and B, the aggregate equivalent model introduces a power loss of 730W. It can be seen that the clustering method simplifies the calculation while causing a certain degree of power deviation.

[0064] Through mathematical analysis, we discuss the clustering of two or more data. For the function P m (S), and its second-order derivative is:

[0065]

[0066] From the above formula, we can see that in the range of light intensity S>0, P″ m (S)>0; get function P m (S) is a strictly convex function. For a strictly convex function, if for any x1, x2 and λ∈[0,1], the following inequality is satisfied:

[0067] f(λx1+(1-λ)x2)<λf(x1)+(1-λ)f(x2);

[0068] when When:

[0069]

[0070] Among them, f(x) represents an arbitrary strictly convex function. The above formula proves that when two data points are clustered, the maximum power corresponding to the cluster center of the light intensity is smaller than the power mean corresponding to the actual two light intensities.

[0071] According to the general form of probability theory of the Jensen inequality:

[0072]

[0073] The expected value of a random variable. represents the expectation of the value of a function of a random variable;

[0074] Prove that the following holds,

[0075]

[0076] Among them, x n represents the nth variable. The above formula shows that for the clustering of any illumination data point, using the illumination intensity of the cluster center as the mean illumination intensity of the entire cluster will reduce the output power of the aggregation model.

[0077] Similarly, the influence of temperature clustering process on power error is analyzed. Assuming that data points D and E belong to the same cluster in the cluster, their illumination intensity is the same but the temperature is different, and the characteristics of their maximum power and temperature are as follows: Figure 4 shown.

[0078] Depend on Figure 4 It can be seen that when two data points with temperatures of 15°C and 25°C are aggregated and equivalent, a power error of 27W will be generated, which is much smaller than the effect of light intensity on the power error. In addition, in actual photovoltaic systems, the difference in light intensity between different photovoltaic units is usually much larger than the temperature difference, and after clustering, the temperature difference will be further reduced. Therefore, in the clustering process, the effect of light intensity on clustering accuracy is mainly considered, and the error caused by temperature data on clustering is ignored.

[0079] In the clustering modeling of photovoltaic power generation systems, the clustering results of light intensity directly affect the accuracy of the aggregation model. However, since the maximum power output of photovoltaic cells has a nonlinear relationship with light intensity, and existing clustering algorithms such as K-means usually assume that the relationship between data points is linear, it leads to non-negligible errors in the clustering process, especially under non-uniform light conditions. This error can make the power output of the aggregation model significantly lower than that of the detailed model. To reduce this error, a data preprocessing method based on weight function to correct light intensity is proposed.

[0080] The design of error correction is based on the characteristic function of the maximum power of photovoltaic cells and light intensity. The light intensity is preprocessed by constructing a weight function to enhance the representativeness of the data in clustering. m(T), construct a light intensity preprocessing weight function w(S). The minimum value of this function is 1. As the light intensity increases, it maintains the same nonlinear increase characteristic as the maximum power characteristic of the photovoltaic cell, thereby increasing the weight of strong light intensity. According to this rule, the expression of the weight function can be obtained as follows:

[0081] w(S)=1+α·Sln(e+0.0005*(S-1000));

[0082] Among them, α represents the weight function coefficient.

[0083] Based on the aggregation model results obtained by the K-means clustering algorithm, the illumination intensity of the entire system is 500W / m 2 Up to 1500W / m 2 , the average light intensity is 1000W / m 2 , the accuracy after clustering is 94.8%, that is, clustering produces a simulation error of about 5%. Then, we can consider providing 5% compensation at the average light intensity of the system to calculate the weight coefficient, that is:

[0084] w(1000)≤1.05;

[0085] The weight function coefficient α is calculated as 0.0005, and the expression of w(S) is:

[0086] w(S)=1+0.0005·Sln(e+0.0005*(S-1000));

[0087] The weight function is used to preprocess the light intensity before clustering to correct the power loss caused by the nonlinear characteristics of the maximum power of photovoltaic cells and light intensity. This can compensate for the results of the photovoltaic aggregation equivalent model and effectively improve its accuracy.

[0088] Therefore, the present invention adopts the above-mentioned accuracy improvement method applied to photovoltaic power generation system clustering, which can effectively improve the simulation accuracy of the photovoltaic power generation system aggregation model.

[0089] It is worth noting that the contents not elaborated in detail in the present invention are all prior art and are well known to those skilled in the art.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A precision improvement method for photovoltaic power generation system clustering, characterized in that: The following steps are involved: Step S1, establishing a mathematical model of a photovoltaic cell; Step S2, performing theoretical analysis on the mathematical model of the photovoltaic cell to obtain the nonlinear characteristics of the maximum power of the photovoltaic cell, light intensity and temperature; Step S3: introducing a modified weight function based on the error loss caused by the nonlinear characteristics.

2. The accuracy improvement method for photovoltaic power generation system clustering according to claim 1, characterized in that: The mathematical model of the photovoltaic cell in step S1 is as follows: Where, I represents the output current of the photovoltaic cell; I sc represents the photocurrent of the photovoltaic panel; V represents the output voltage of the photovoltaic cell; V oc Represents the open circuit voltage of the photovoltaic cell; V m Indicates the maximum power point voltage of the photovoltaic cell; I m Indicates the maximum power point current of the photovoltaic cell; I sc represents the short-circuit current of the photovoltaic cell; a, b, c represent the compensation coefficients; S represents the actual light temperature; T represents the actual working temperature; I scref Represents the reference value of photocurrent; I mref Indicates the maximum power point current reference value of the photovoltaic cell; V ocref Indicates the open circuit voltage reference value of the photovoltaic cell; V mref Indicates the maximum power point voltage reference value of the photovoltaic cell; ΔT = TT ref Indicates the difference between the actual temperature and the reference temperature, T ref Indicates the temperature under standard working conditions, which is 25 degrees Celsius; ΔS = SS ref Represents the difference between actual illumination and reference illumination, S ref Indicates the light temperature under standard working conditions, which is 1000W / m 2 ; The maximum power of photovoltaic cells is expressed as: P m =V m I m , change V in the mathematical model of photovoltaic cells m ,I m Substitute into the formula P m =V m I m get: Among them, P m (S,T) represents the maximum power of the photovoltaic cell as a function of light intensity and temperature.

3. The accuracy improvement method for photovoltaic power generation system clustering according to claim 2, characterized in that: In step S2, the temperature T is set as a constant, and the nonlinear characteristic P of the maximum output power of the photovoltaic cell and the light intensity is obtained. m (S) is as follows: Similarly, when the light intensity is stable, the nonlinear characteristic P of the maximum output power and temperature is obtained. m (T) is as follows:

4. The accuracy improvement method for photovoltaic power generation system clustering according to claim 3 is characterized in that: The error loss in step S3 is: m (S), and its second-order derivative is: Among them, in the interval of light intensity S>0, P″ m (S)>0; get function P m (S) is a strictly convex function. For a strictly convex function, if for any x1, x2 and λ∈[0,1], the following inequality is satisfied: f(λx1+(1-λ)x2)<λf(x1)+(1-λ)f(x2); when When: Here, f(x) represents any strictly convex function.

5. The accuracy improvement method for photovoltaic power generation system clustering according to claim 4, characterized in that: According to the general form of probability theory of the Jensen inequality: The expected value of a random variable. represents the expectation of the value of a function of a random variable; Prove that the following holds, Among them, x n Represents the nth variable.

6. The accuracy improvement method for photovoltaic power generation system clustering according to claim 5, characterized in that: The modified weight function w(S) in step S3 is: w(S)=1+α·Sln(e+0.0005*(S-1000)); Among them, α represents the weight function coefficient.

Images