Real-time calculation method for solar irradiation intensity of photovoltaic power station

Abstract

The invention discloses a real-time calculation method for solar irradiation intensity of a photovoltaic power station. The method comprises the following steps: classifying data quality levels of a distributed photovoltaic power station according to operation data; carrying out modeling description on the geographic position information of the distributed photovoltaic power station; carrying out data cleaning on the distributed photovoltaic power stations with relatively high data quality levels; performing data cleaning on the distributed photovoltaic power stations with relatively low data quality levels; and designing a solar radiation intensity calculation method for the distributed photovoltaic system after data cleaning and calculating. According to the method, abnormal data cleaning is carried out on original data by considering that the quality of operation data of the distributed photovoltaic power station is relatively poor and the influence on a result is relatively large; meanwhile, in order to reduce errors of calculation results, a joint model is established for irradiance calculation, a high-cost-performance and high-precision solution is provided for solar irradiance monitoring of the distributed photovoltaic power station, and the accuracy of output power prediction of the distributed photovoltaic power station can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation, and particularly to a method for real-time calculation of solar irradiance intensity in a photovoltaic power station. Background Art

[0002] In recent years, under the background of the greenhouse effect and energy crisis, photovoltaic power generation, as a sustainable and green new energy, has developed very rapidly. Since distributed photovoltaic power stations generally lack solar radiation monitoring devices, and the measurement level of power generation is low, the problem of missing operation data is serious. The lack of real-time solar irradiance information seriously affects the realization of core key functions such as power generation calculation, power prediction, fault diagnosis, and scheduling plans. Therefore, after cleaning the data of the photovoltaic power station, calculating the solar irradiance intensity data through the operation data of the photovoltaic power station has certain guiding significance for studying the operation characteristics of the photovoltaic power station, reasonably planning the layout location, and monitoring the operation status.

[0003] In view of the above problems, the present invention provides a method for real-time calculation of solar irradiance intensity in a photovoltaic power station, which reduces the installation, operation, and instrument costs of measuring related data. Summary of the Invention

[0004] The present invention provides a method for real-time calculation of solar irradiance intensity in a photovoltaic power station, which reduces the installation, operation, and instrument costs of measuring related data.

[0005] The object of the present invention is to provide a method for real-time calculation of solar irradiance intensity in a photovoltaic power station. The specific steps of the method are as follows:

[0006] Step 1: Classify the data quality level of the distributed photovoltaic power station according to the operation data;

[0007] Step 2: Model and describe the geographical location information of the distributed photovoltaic power station;

[0008] Step 3: Clean the data of the distributed photovoltaic power station with a higher data quality level;

[0009] Step 4: Clean the data of the distributed photovoltaic power station with a lower data quality level;

[0010] Step 5: Design a solar radiation intensity calculation method for the distributed photovoltaic power station after data cleaning and calculate.

[0011] Further, the specific steps of Step 1 are as follows:

[0012] Step 11: Obtain the operating power data of the distributed photovoltaic power station and the irradiance data of the nearby meteorological station, calculate the correlation coefficient between the power and irradiance data of each power station, and select the five power stations with the highest correlation to calculate the average normalized power as the "standard value";

[0013]

[0014] Among them, D norm，t represents the standard value, and P norm_max1，t , P norm_max2，t , P norm_max3，t , P norm_max4，t , P norm_max5，t respectively represent the correlation coefficients of the operating power data and irradiance data of five power stations;

[0015] Step 12: Compare the normalized power value of each power station with the "standard value". If it exceeds 20% of the "standard value" range at a certain moment, it is determined as abnormal data, and count the number K of abnormal data values of each power station in that year i，T ;

[0016] IF P norm_i，t <0.8×D norm,t or P norm_i，t >1.2×D norm,t

[0017] K i，T =K i，T +1

[0018] Step 13: Introduce the data efficiency q i to quantify the quality of distributed photovoltaic data, and classify distributed photovoltaic power stations according to the data efficiency results. Power stations with a data efficiency greater than 0.7 are reliable power stations; the rest are non-reliable power stations. The data efficiency calculation formula is as follows:

[0019]

[0020] Among them, q i is the power data efficiency of the i-th distributed photovoltaic power station, K i is the number of samples of abnormal data in the operating data of the i-th photovoltaic power station, T d is the data collection time length, and N i is the unit standard point number.

[0021] Furthermore, the specific steps for modeling the geographical distribution of regional distributed photovoltaic power stations in Step 2 are as follows:

[0022] Step 21: Regard the longitude and latitude of each power station as two-dimensional data, calculate the data variances on the two dimensions respectively, and select the dimension with the larger variance as the splitting axis;

[0023] Step 22: Sort the data according to the splitting axis dimension, determine the median, and use its position as the splitting node; in the splitting axis direction, divide the left and right branches with the splitting node as the standard; reselect the splitting axis and splitting node on the left and right branches;

[0024] Step 23: Repeat the operation until each space contains only one point, and complete the modeling of the geographical information of the power station.

[0025] Further, the specific steps of Step 3 are as follows:

[0026] Step 31: Divide the original operation data of reliable power stations into 120 subsets according to the irradiance interval T = 10W / m 2 Arrange the power data in each subset in descending order of irradiance, and then calculate the moving average and moving standard deviation. The calculation formulas are as follows:

[0027]

[0028] where MA i is the i-th moving average in a certain subset of a reliable power station, and SD i is the i-th moving standard deviation in a certain subset of a reliable power station, X j is the j-th data point in the sorted data sequence of a certain subset of a reliable power station, is the average value of the data within the window, and k is the window size;

[0029] Step 32: Set the moving standard deviation threshold to 0.2, calculate the mean and standard deviation of each subset of the operation data, and perform abnormal data discrimination and cleaning within the subset;

[0030]

[0031] where n is the number of data in this subset of this reliable power station, is the j-th data of the z-th subset of the i-th reliable power station, μ i，z is the mean of the z-th subset of the i-th reliable power station, and σ i，z is the standard deviation value of the z-th subset of the i-th reliable power station.

[0032] Further, the specific steps of Step 4 are as follows:

[0033] Step 41: For each unreliable power station, use the geographical information model constructed in Step 2 to search for the 5 nearest reliable power stations through the index path of the model, and extract the cleaning reference indicators μ and σ of the 5 reliable power stations;

[0034] Step 42: Calculate the weights of the cleaning indicators of the 5 reliable power stations in Step 41, obtain the weighted average and weighted standard deviation as the cleaning reference values of the unreliable power station, and perform abnormal data discrimination and cleaning:

[0035]

[0036]

[0037] Among them, d i is the distance from each i-th reliable power station to the unreliable power station, x and y are the location information of the unreliable power station, x i and y i are the location information of the i-th neighboring reliable power station, w i is the weight of the μ and σ values of the i-th reliable power station in the index, μ o，j is the cleaning index value μ of the j-th classification subset at the point to be cleaned, μ i，j is the cleaning index value μ of the j-th subset in the i-th reliable power station; σ o，j is the cleaning index value σ of the j-th classification subset at the point to be cleaned, σ i，j is the cleaning index value of the j-th subset in the i-th reliable power station.

[0038] Furthermore, the specific steps of step 5 are as follows:

[0039] Step 51: Utilize the geographical information model constructed in step 2, search for the 5 nearest power stations through the index path of the model, and extract the power data of the 5 power stations after cleaning;

[0040] Step 52: Take the power of the site itself, the power data of the 5 power stations obtained in step 51, and the irradiance data of the meteorological stations in the region as the input of the CNN-LSTM joint model, and calculate the irradiance data of the point to be measured.

[0041] The present invention has the following advantages: The present invention quantifies the distributed photovoltaic data quality by introducing the data efficiency, and performs abnormal data cleaning on the power station classification according to the data quality situation.

[0042] The present invention reduces the installation, operation, and instrument costs of measuring relevant data by using the known information in the region to obtain unknown meteorological data.

[0043] The solar irradiance data of the photovoltaic power station calculated by the present invention helps to improve the accuracy of the output power prediction of the distributed photovoltaic power station, and has certain guiding significance for studying the operating characteristics of the photovoltaic power station, reasonably planning the layout location, and monitoring the operating status. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is the flow chart of the present invention;

[0045] Figure 2 is the result diagram of the regional geographical location information modeling in the present invention;

[0046] Figure 3 is the comparison diagram of the data efficiency values before and after abnormal data cleaning of each power station in the present invention;

[0047] Figure 4 It is the comparison curve graph of the calculated irradiance and the actual measured value at the weather station in spring in the present invention;

[0048] Figure 5 It is the comparison curve graph of the calculated irradiance and the actual measured value at the weather station in summer in the present invention;

[0049] Figure 6 It is the comparison curve graph of the calculated irradiance and the actual measured value at the weather station in autumn in the present invention;

[0050] Figure 7 It is the comparison curve graph of the calculated irradiance and the actual measured value at the weather station in winter in the present invention;

[0051] Figure 8 It is the comparison graph of the calculated irradiance description coefficients of a typical power station in the present invention. Detailed implementation manners

[0052] The present invention provides a method for real-time calculation of solar irradiance intensity of a photovoltaic power station. The specific steps of the method are as follows:

[0053] Step 1: Classify the data quality level of the distributed photovoltaic power station according to the operation data;

[0054] Step 2: Model and describe the geographical location information of the distributed photovoltaic power station;

[0055] Step 3: Clean the data of the distributed photovoltaic power station with a higher data quality level;

[0056] Step 4: Clean the data of the distributed photovoltaic power station with a lower data quality level;

[0057] Step 5: Calculate the solar radiation intensity calculation method for the distributed photovoltaic design after data cleaning and calculate.

[0058] In this embodiment, the specific steps of Step 1 are as follows:

[0059] Step 11: Obtain the operation power data of the distributed photovoltaic power station and the irradiance data of the nearby weather station, calculate the correlation coefficients of the power and irradiance data of each power station, and select the five power stations with the highest correlation to calculate the average normalized power as the "standard value";

[0060]

[0061] Among them, D norm，t represents the standard value, and P norm_max1，t , P norm_max2，t , P norm_max3，t , P norm_max4，t , P norm_max5，t respectively represent the correlation coefficients of the operation power data and the irradiance data of the five power stations;

[0062] Step 12: Compare the normalized power values of each power station with the "standard value". If the value at a certain moment exceeds 20% of the "standard value" range, it is determined as abnormal data, and count the number K of abnormal data values of each power station in that year. i，T ;

[0063] IF P norm_i，t <0.8×D norm，t or P normi，t >1.2×D norm，t

[0064] K i，T =K i，T +1

[0065] Step 13: Introduce the data efficiency q i Quantify the data quality of distributed photovoltaic data, and classify the distributed photovoltaic power stations according to the data efficiency results. Power stations with a data efficiency greater than 0.7 are reliable power stations; the rest are non-reliable power stations. The data efficiency calculation formula is as follows:

[0066]

[0067] where q i is the power data efficiency of the i-th distributed photovoltaic power station, K i is the number of samples of abnormal data in the operation data of the i-th photovoltaic power station, T d is the acquisition data time length, and N i is the unit standard point number.

[0068] In this embodiment, the specific steps for modeling the geographical distribution of the regional distributed photovoltaic power stations in step 2 are as follows:

[0069] Step 21: Regard the longitude and latitude of each power station as two-dimensional data, calculate the data variances on the two dimensions respectively, and select the dimension with the larger variance as the splitting axis;

[0070] Step 22: Sort the data according to the splitting axis dimension, determine the median, and use its position as the splitting node; in the splitting axis direction, divide the left and right branches with the splitting node as the standard; re-select the splitting axis and splitting node on the left and right branches;

[0071] Step 23: Repeat the operation until each space contains only one point, and complete the modeling of the geographical information of the power stations.

[0072] In this embodiment, the specific steps of step 3 are as follows:

[0073] Step 31: Arrange the original operation data of the reliable power stations according to the irradiance interval T = 10W / m 2It is divided into 120 subsets. The power data within each subset is arranged in descending order of irradiance, and then the moving mean and moving standard deviation are calculated. The calculation formulas are as follows:

[0074]

[0075] Where, MA i is the i-th moving average within a subset of a certain reliable power station, and SD i is the i-th moving standard deviation within a subset of a certain reliable power station. X j is the j-th data point in the sorted data sequence within a subset of a certain reliable power station. is the average value of the data within the window, and k is the window size.

[0076] Step 32: Set the moving standard deviation threshold to 0.2, calculate the mean and standard deviation of each subset of the operation data, and perform abnormal data discrimination and cleaning within the subset;

[0077]

[0078] Where, n is the number of data within this subset of this reliable power station. is the j-th data of the z-th subset of the i-th reliable power station, μ i，z is the mean value of the z-th subset of the i-th reliable power station, and σ i，z is the standard deviation value of the z-th subset of the i-th reliable power station.

[0079] In this embodiment, the specific steps of Step 4 are as follows:

[0080] Step 41: For each unreliable power station, use the geographic information model constructed in Step 2 to search for the 5 nearest reliable power stations through the index path of the model, and extract the cleaning reference indicators μ and σ of the 5 reliable power stations;

[0081] Step 42: Calculate the weights of the cleaning indicators of the 5 reliable power stations in Step 41, obtain the weighted average value and weighted standard deviation as the cleaning reference values of the unreliable power station, and perform abnormal data discrimination and cleaning:

[0082]

[0083] Where, d i is the distance from each i-th reliable power station to the unreliable power station, x and y are the location information of the unreliable power station, x i and y i are the location information of the i-th neighboring reliable power station, w i is the weight of the μ and σ values of the i-th reliable power station in the indicator, μ o，j is the cleaning indicator value μ of the j-th classification subset at the point to be cleaned, μi，j The cleaning index values μ; σ of the j-th subset in the i-th reliable power station o，j For the cleaning index value σ of the j-th classification subset at the point to be cleaned, σ, σ i，j The cleaning index value of the j-th subset in the i-th reliable power station.

[0084] In this embodiment, the specific steps of step 5 are as follows:

[0085] Step 51: Utilize the geographic information model constructed in step 2 to search for the 5 nearest power stations through the index path of the model, and extract the power data after cleaning of the 5 power stations;

[0086] Step 52: Use the power of the site itself, the power data of the 5 power stations obtained in step 51, and the irradiance data of the meteorological stations in the region as the input of the CNN-LSTM joint model to calculate the irradiance data of the point to be measured.

[0087] Embodiment 1:

[0088] Taking distributed photovoltaic power stations in a certain region of a country as the research object, there are 162 distributed power stations and one meteorological station built in this region. The collected data includes the power data of distributed photovoltaic power stations and the solar irradiance, temperature, and wind speed data collected by the meteorological station, and the data sampling interval is 15 minutes. Figure 2 Shows the result of the geographical location information modeling of this region. Figure 3 Compares the data availability before and after cleaning abnormal data of the power stations, showing that the data quality of the power stations has been significantly improved after cleaning. Figures 4 - 7 Shows the comparison between the calculated irradiance and the actual measured value at the meteorological station in different seasons, indicating that the model results are roughly consistent with the actual irradiance data. However, the figure shows that due to the influence of data quality, there is a phenomenon that the calculated irradiance data suddenly becomes zero.

[0089] At the same time, four stations equipped with irradiance monitoring devices are selected, and one month's data is extracted from each station to verify the calculated irradiance. Considering that the data availability after cleaning of the four power stations is 100%, 73.53%, 66.35%, and 84.73% respectively, the verification results are representative. We compared the irradiance results calculated by the CNN-LSTM model before and after data cleaning and the irradiance results calculated by using the CNN model alone after data cleaning. Through indicators such as root mean square error RMSE, mean absolute error MAE, and coefficient of determination R2, the relationship between the calculated irradiance of the power stations and the actual value is evaluated. The specific comparison results are shown in Table 1. Figure 8 , shows the descriptive coefficients of the calculated irradiance of the four power stations. The results show that the calculation performance of the CNN-LSTM model is better than that of the single CNN model, and the data cleaning step has a significant impact on improving the calculation accuracy.

[0090] Table 1 Comparison of Calculated Irradiance Description Coefficients

[0091]

[0092]

[0093] Considering the low quality of the collected operation data of the photovoltaic power station, abnormal data cleaning is carried out on the original operation data; at the same time, in order to reduce the error of the calculation results, a joint model is established for irradiance calculation, providing a cost-effective and high-precision solution for the solar irradiance monitoring of distributed photovoltaic power stations.

[0094] Although the specific embodiments of the present invention have been described in detail with reference to the accompanying drawings, it should not be construed as a limitation on the protection scope of the present invention. Within the scope described in the claims, various modifications and deformations that can be made by those skilled in the art without creative work still fall within the protection scope of the present invention.

Claims

1. A real-time calculation method for solar radiation intensity of a photovoltaic power station, characterized by: The specific steps of the method are as follows: Step 1: Classify the data quality level of distributed photovoltaic power stations according to the operation data; Step 2: Model and describe the geographical location information of the distributed photovoltaic power station; Step 3: Clean the data of distributed photovoltaic power stations with high data quality level; Step 4: Clean the data of distributed photovoltaic power stations with low data quality level; Step 5: Design and calculate the solar radiation intensity calculation method for distributed photovoltaic after data cleaning; The specific steps of step 5 are as follows: Step 51: using the geographic information model constructed in step 2, searching for neighboring power stations through the index path of the model, and extracting the power data of the power stations after cleaning; Step 52: The power of the site itself, the power data of the power station obtained in step 51, and the irradiance data of the meteorological station in the area are used as inputs of the CNN-LSTM joint model to calculate the irradiance data of the point to be measured.

2. A method for real-time calculation of solar radiation intensity in a photovoltaic power station according to claim 1, characterized in that: The specific steps of step 1 are as follows: Step 11: Obtain the operating power data of the distributed photovoltaic power station and the irradiance data of the nearby meteorological station, calculate the correlation coefficient between the power and irradiance data of each power station, and select the five power stations with the highest correlation to calculate the average normalized power as the standard value; Among them, D norm，t Indicates standard value, P norm_max1，t , P norm_max2，t , P norm_max3，t , P norm_max4，t , P norm_max5，t Respectively represent the correlation coefficients between the operating power data and irradiance data of the five power stations; Step 12: Compare the normalized power value of each power station with the standard value. If it exceeds the standard value range by 20% at a certain moment, it is determined to be abnormal data. The number of abnormal data values ​​K of each power station in that year is counted. i，T ; IF P norm_i，t <0.8×D norm，t or P norm_i，t >1.2×D norm，t K i，T =K i，T +1; Step 13: Introducing data efficiency q i Quantify the quality of distributed photovoltaic data and classify distributed photovoltaic power stations according to the data efficiency results. Power stations with data efficiency greater than 0.7 are reliable power stations, and the rest are unreliable power stations. The data efficiency calculation formula is as follows: Among them, q i is the efficiency of the power data of the i-th distributed photovoltaic power station, K i is the number of samples of abnormal data in the operation data of the i-th photovoltaic power station, T d N is the length of time for collecting data, i The unit is standard points.

3. A method for real-time calculation of solar radiation intensity in a photovoltaic power station according to claim 1, characterized in that: The specific steps for modeling the geographical distribution of regional distributed photovoltaic power stations in step 2 are: Step 21: Consider the longitude and latitude of each power station as two-dimensional data, calculate the data variance in two dimensions respectively, and select the dimension with the largest variance as the segmentation axis; Step 22, sort the data according to the split axis dimension, determine the median, and use its position as the split node; in the direction of the split axis, divide the left and right branches according to the split node; reselect the split axis and split node on the left and right branches; Step 23: Repeat the operation until each space contains only one point, thus completing the modeling of the geographic information of the power station.

4. A method for real-time calculation of solar radiation intensity in a photovoltaic power station according to claim 1, characterized in that: The specific steps of step 3 are as follows: Step 31: The original operating data of the reliable power station is calculated according to the irradiance interval T = 10W / m 2 Divide into 120 subsets, and arrange the power data in each subset in descending order according to irradiance. Then calculate the sliding mean and sliding standard deviation. The calculation formula is as follows: Among them, MA i Is the first in a subset of a reliable power station i Sliding mean, SD i is the i-th sliding standard deviation of a subset of reliable power stations, X j is the jth data point in the sorted data sequence of a subset of a reliable power station, is the average value of the data in the window, and k is the window size; Step 32: Set the sliding standard deviation threshold to 0.2, calculate the mean and standard deviation of each subset of the running data, and identify and clean abnormal data in the subset; if SD i ≤0.2 Where n is the number of data in the subset of the reliable power station, is the jth data of the zth subset of the i-th reliable power station, μ i，z is the mean of the zth subset of the ith reliable power station, σ i，z is the standard deviation of the zth subset of the ith reliable power station.

5. A method for real-time calculation of solar radiation intensity in a photovoltaic power station according to claim 1, characterized in that: The specific steps of step 4 are as follows: Step 41: for each unreliable power station, using the geographic information model constructed in step 2, search for the five nearest reliable power stations through the index path of the model, and extract the cleaning reference indicators μ and σ of the five reliable power stations; Step 42: Calculate the weights of the cleaning indicators of the five reliable power stations in step 41, obtain the weighted average value and weighted standard deviation as the cleaning reference value of the unreliable power station, and perform the identification and cleaning of abnormal data: Among them, d i is the distance from each i-th reliable power station to the unreliable power station, x and y are the location information of the unreliable power station, x i and i is the location information of the i-th neighboring reliable power station, w i is the weight of μ and σ of the i-th reliable power station in the index, μ o，j is the cleaning index value μ of the jth classification subset at the cleaned point, μ i，j The cleaning index value μ of the jth subset in the i-th reliable power station; σ o，j is the cleaning index value σ of the jth classification subset at the cleaned point, σ i，j The cleaning index value of the jth subset in the i-th reliable power station.

Images