Photovoltaic power station assessment and early warning method based on machine learning

Abstract

The invention belongs to the technical field of photovoltaic power station assessment, and particularly relates to a photovoltaic power station assessment and early warning method based on machine learning. The invention relates to a photovoltaic power station assessment and early warning method based on machine learning, and the method comprises the steps: carrying out the normalization processing of single-machine information, booster station information and meteorological environment information of a photovoltaic power station, obtaining a data set, inputting a training set in the data set into an LSTM neural network for training, inputting a test set into the trained LSTM neural network, and carrying out the early warning of the photovoltaic power station. And acquiring a short-term power prediction result of the photovoltaic power station, calculating values of various assessment indexes extracted from the power grid access assessment standard by using the prediction result, and comparing the assessment indexes with a set early warning threshold value until the assessment indexes are smaller than or equal to the set early warning threshold value. According to the network access assessment of photovoltaic power generation, an index calculation method and parameters can be flexibly adjusted, and clear assessment indexes and methods are provided.

Description

Technical Field

[0001] The present invention belongs to the technical field of photovoltaic power station evaluation, and specifically relates to a photovoltaic power station assessment and early warning method based on machine learning. Background Art

[0002] In order to promote the construction of a new power system with new energy as the main body, ensure the safe, high-quality and economical operation of the power system, and effectively regulate the grid-connected operation management and auxiliary service management of the power system, power regulatory agencies in various regions have strengthened the assessment and management of photovoltaic power stations. However, the existing assessment methods for photovoltaic power stations lack scientific design of assessment indicators and parameters, and there is no unified standard.

[0003] Traditionally, new energy grid connection is assessed through self-inspection, but the assessment scheme is too general and lacks detailed parameter assessment, which cannot meet the assessment needs of the existing power sector for photovoltaic power generation grid connection requirements. Summary of the invention

[0004] The purpose of the present invention is to provide a photovoltaic power station assessment and early warning method based on machine learning in view of the above problems, and to formulate an assessment index system including the photovoltaic power generation power prediction accuracy, power generation control and regulation qualification rate, active power change rate and photovoltaic power station operation data report qualification rate on the basis of the photovoltaic power station access assessment standard; adopt a machine learning model to predict and calculate the short-term power of the photovoltaic power station, and calculate the various assessment index values ​​of the photovoltaic power station according to the short-term power prediction result, so as to improve the scientificity and accuracy of the assessment calculation; use an integrated monitoring platform to realize the calculation and analysis of the assessment index at the station end, and in case the assessment index value is greater than the assessment early warning threshold, promptly issue an early warning to the photovoltaic power station to improve the stability of power grid operation.

[0005] In order to achieve the above object, the technical solution provided by the present invention is: A photovoltaic power station assessment and early warning method based on machine learning includes the following steps: S1: Obtain the single-machine information, booster station information and meteorological environment information of the photovoltaic power station in the order of historical time, and construct the data set D1; S2: Normalize the data set D1 to obtain the data set D2, and divide the data set D2 into a training set and a test set according to the proportion; S3: Input the training set into the LSTM neural network for training, and input the test set into the trained LSTM neural network to obtain the short-term, ultra-short-term and medium-term power prediction results of the photovoltaic power station; S4: extracting various assessment indicators from the grid access assessment standards of the photovoltaic power station, and calculating the values ​​of various assessment indicators through the short-term, ultra-short-term and medium-term power prediction results obtained in step S2; S5: Compare the values ​​of the various assessment indicators obtained in step S3 with the set assessment warning thresholds to conduct grid access assessment of the photovoltaic power station.

[0006] Furthermore, in step S1, data set D1 includes single machine information, booster station information and meteorological environment information; single machine information includes available power, actual power and meteorological environment information; environmental information includes light radiation intensity, atmospheric temperature, atmospheric humidity, cloud height, sun position, air pressure, wind speed and direction at a specified height and light time information.

[0007] Furthermore, in step S2, the data set D1 is normalized, including using the normalization method to clean up the single-machine information, booster station information, and meteorological environment information of the photovoltaic power station. The calculation formula is as follows:

[0008] In the formula, is the normalized data of the assessment indicators. Indicates the assessment indicator data. Indicates the minimum value of the assessment indicator data. Indicates the maximum value of the assessment indicator data.

[0009] Preferably, in step S4, the assessment indicators include power prediction accuracy, power prediction upload rate, automatic power generation control operation rate, automatic voltage control operation rate, automatic power generation control regulation pass rate, automatic voltage control regulation pass rate and active power change rate.

[0010] Preferably, the assessment indicators also include an E-text upload rate and an E-text pass rate, and the E-text is used to report the operation data of the photovoltaic power station.

[0011] Furthermore, the power prediction accuracy includes short-term power prediction accuracy, ultra-short-term power prediction accuracy and medium-term power prediction accuracy; the short-term power prediction accuracy is calculated by obtaining the prediction data within 24 hours from the E file, the ultra-short-term power prediction accuracy is calculated by obtaining the prediction data of the first 96 points from the E file, and the medium-term prediction accuracy is calculated by obtaining the prediction data 4 days ago from the E file, taking the 16th point of each ultra-short-term prediction file, and taking 96 files every day for calculation.

[0012] Preferably, the power prediction accuracy includes: 1) Short-term power prediction accuracy, calculated as: ; In the formula, represents the short-term power prediction accuracy result, Indicates the available power at time i in the power-limited state within 24 hours obtained from the E file, or the actual power at time i in the power-unlimited state; represents the predicted power at time i within 24 hours obtained from the E file, is the total installed capacity of the photovoltaic power station in that month; 2) Ultra-short-term accuracy, calculated as: ; in, represents the ultra-short-term accuracy result, Indicates the available power at time i in the power-limited state or the actual power at time i in the unlimited state obtained from the E file for the first 96 points; represents the predicted power of the first 96 points obtained from the E file for time i; 3) Medium-term forecast accuracy, calculated as follows: ; in, represents the mid-term forecast accuracy result, Indicates the available power at time i in the power-limited state or the actual power at time i in the unlimited state obtained from the E file 4 days ago; It represents the predicted power at time i 4 days ago obtained from the E file.

[0013] Furthermore, the calculation formula for the E-text pass rate is:

[0014] In the formula, k is the scoring item of data scoring, represents the actual score of the kth item at time T, and K represents the maximum value of the scoring item; Preferably, the photovoltaic power station assessment indicators also include: (1) Power prediction upload rate, calculated as:

[0015] In the formula, The upload rate M represents the number of files successfully uploaded, and N represents the number of files scheduled to be uploaded; (2) Automatic power generation control operation rate, calculated as follows:

[0016] Where TF represents the automatic generation control operation rate, Indicates the automatic generation control substation exit time, Indicates the total stop time of the automatic generation control unit. Indicates free examination time. Indicates the total statistical time of the month; (3) Automatic voltage control operation rate: used to determine the substation automatic voltage control operation rate. The calculation formula is:

[0017] In the formula, TD represents the operation rate of automatic voltage control, Indicates the automatic voltage control substation exit time, Indicates the total stop time of the automatic voltage control unit. Indicates free examination time. Indicates the total statistical time of the month; (4) Automatic generation control and regulation qualification rate is used to determine the qualification of the automatic generation control and regulation of the substation. The calculation formula is:

[0018] In the formula, HF represents the automatic generation control regulation qualification rate, Indicates the number of qualified points of automatic power generation control substation. Indicates the number of times the master station sends adjustment instructions; (5) Automatic voltage control regulation pass rate is used to determine the pass rate of automatic voltage control regulation of the substation. The calculation formula is:

[0019] In the formula, HD represents the automatic voltage control regulation pass rate, Indicates the number of qualified points of automatic voltage control substation execution; (6) Active power change rate: used to determine the active power change rate of the substation. The average value of the power change difference per minute within 10 minutes at a certain moment is calculated on a rolling basis. The calculation formula is:

[0020] In the formula, They represent the real-time common power of the two grid-connected devices at time i+j-1 and i+j respectively, and i and j represent constants.

[0021] Compared with the prior art, the beneficial effects of the present invention include: 1) The present invention uses a machine learning algorithm to accurately predict the power of a photovoltaic power station, track and monitor parameter changes of photovoltaic power generation grid connection, assess items and indicators, and use an integrated monitoring platform to calculate and analyze assessment indicators at the station end, and flexibly adjust indicator calculation methods and parameters.

[0022] 2) The photovoltaic power station assessment and early warning method proposed in the present invention adjusts the power generation strategy of the photovoltaic power plant to avoid the situation where the photovoltaic power station is fined for failing to meet the assessment indicators. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0024] Figure 1 The present invention is a flowchart of a photovoltaic power station assessment and early warning method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] like Figure 1 As shown, a photovoltaic power station assessment and early warning method based on machine learning includes the following steps: S1: Obtain the single-machine information, booster station information and meteorological environment information of the photovoltaic power station in the order of historical time to obtain the data set D1; S2: Normalize the data set D1 to obtain the data set D2, divide the data set D2 into a training set and a test set according to the ratio, and input the training set into LSTM The neural network is trained and the test set is input into the trained LSTM In the neural network, the short-term, ultra-short-term and medium-term power prediction results of the photovoltaic power station are obtained; S3: extracting various assessment indicators from the grid access assessment standards of photovoltaic power stations, and calculating the values ​​of various assessment indicators through the short-term, ultra-short-term and medium-term power prediction results obtained in step S2; S4: Compare the assessment index obtained in step S3 with the set warning threshold. If the assessment index is greater than the warning threshold, issue a warning for the assessment index of the photovoltaic power station. If the assessment index is not greater than the warning threshold, repeat steps S2 to S3.

[0026] Furthermore, in step S1, the data set D1 , including single machine information, booster station information and meteorological environment information, which are counted in the short term and erroneous data are deleted; the single machine information includes available power, actual power and meteorological environment information; the environmental information includes light radiation intensity, atmospheric temperature, atmospheric humidity, cloud height, sun position, air pressure, wind speed and direction at a specified height and light time information.

[0027] In step S2, the data set D1 Normalization processing is performed, including data cleaning of the single-machine information of the photovoltaic power station, the booster station information, and the meteorological environment information using the normalization method. The calculation formula is as follows:

[0028] In the formula, is the normalized data of the assessment indicators. It is the assessment indicator data. is the minimum value of the assessment indicator data, It is the maximum value of the assessment indicator data.

[0029] In step S2, the grid access assessment standards for photovoltaic power stations include power prediction accuracy, power prediction upload rate,E Text upload rate, E E-text is an electronic file generated during the operation of the photovoltaic power station and uploaded to the main station, which contains the operation data, status information and forecast data of the photovoltaic power station.

[0030] In step S3, the calculation formula of the assessment index is: (1) Power forecast accuracy, which refers to the closeness between the predicted power curve and the actual power generation curve, including short-term power forecast accuracy, ultra-short-term power forecast accuracy and medium-term power forecast accuracy; 1) Short-term power prediction accuracy: The prediction data within 24 hours is obtained from the E file and the calculation formula is: ; In the formula, represents the short-term power prediction accuracy result, It indicates the available power at time i obtained from the E file in the power-limited state within 24 hours, and the actual power at time i in the unlimited state; It represents the predicted power at time i within 24 hours obtained from the E file. is the total installed capacity of the photovoltaic power station in that month; 2) Ultra-short-term accuracy, calculated by obtaining the forecast data of the first 96 points from the E file, the calculation formula is: ; in, represents the ultra-short-term accuracy result, It indicates the available power at time i in the power-limited state and the actual power at time i in the unlimited state for the first 96 points obtained from the E file; represents the predicted power of the first 96 points at time i obtained from the E file, and Cap is the total installed capacity of the photovoltaic power station in that month; 3) Medium-term forecast accuracy: Get the forecast data 4 days ago from the E file, take the 16th point of each ultra-short-term forecast file, and take 96 files every day for calculation. The calculation formula is: ; in, represents the mid-term forecast accuracy result, It indicates the available power at time i obtained from the E file 4 days ago under the power-limited state, and the actual power at time i under the power-unlimited state; represents the predicted power at time i 4 days ago obtained from the E file, It is the total installed capacity of photovoltaic power stations in that month.

[0031] (2) The power prediction upload rate calculation formula is: SCL=M / N×100%; In the formula, SCL represents the upload rate, M represents the number of files successfully uploaded, and N represents the number of files planned to be uploaded; (3) E-text pass rate Verify the format and data of the E text from the obtained E file. The calculation formula is: ; In the formula, k is the scoring item of data scoring, represents the actual score of the kth item at time T, and K represents the maximum value of the scoring item; (4) Automatic power generation control operation rate, calculated as follows: ; Where TF represents the automatic generation control operation rate, Indicates the automatic generation control substation exit time, Indicates the total stop time of the automatic generation control unit. TM Indicates free examination time. TDY Indicates the total statistical time for the month.

[0032] (5) Automatic voltage control operation rate: used to determine the substation automatic voltage control operation rate. The calculation formula is: ; Where TD represents the automatic voltage control operation rate, Indicates the automatic voltage control substation exit time, Indicates the total stop time of the automatic voltage control unit. TM Indicates free examination time. TDY Indicates the total statistical time for the month.

[0033] (6) Automatic generation control and regulation qualification rate is used to determine the qualification of the automatic generation control and regulation of the substation. The calculation formula is: ; In the formula, HF represents the automatic generation control regulation qualification rate, Indicates the number of qualified points of automatic power generation control substation. ZZ Indicates the number of times the master station sends adjustment instructions; (7) Automatic voltage control regulation pass rate is used to determine the pass rate of automatic voltage control regulation of the substation. The calculation formula is: ; In the formula, HD represents the automatic voltage control regulation pass rate, Indicates the number of qualified points of automatic voltage control substation execution. ZZ Indicates the number of times the master station sends adjustment instructions; (8) Active power change rate: used to determine the active power change rate of the substation. The average value of the power change difference per minute within 10 minutes at a certain moment is calculated on a rolling basis. The calculation formula is: ; In the formula, They represent the real-time public power of the two grid-connected nodes at time i+j-1 and time i+j respectively.

[0034] In step S5, the warning threshold is manually set, and the value is in the range of 105% to 110% of the preset assessment value.

[0035] The photovoltaic power station assessment and early warning method based on machine learning in Example 1 of the present invention has been verified to have an assessment early warning rate of more than 97% for photovoltaic power stations, and the prediction results are close to actual needs. The method digitizes the assessment indicators, calculates and accurately predicts the assessment content, and meets the assessment requirements of the power department for the grid connection requirements of photovoltaic power generation.

Claims

1. A photovoltaic power station assessment and early warning method based on machine learning, characterized in that: The following steps are involved: S1, obtain historical data of photovoltaic power stations, booster stations and meteorological environment in chronological order to construct data set D1; S2, normalize the data set D1 to obtain the data set D2, and divide the data set D2 into a training set and a test set according to the proportion; S3. Use LSTM neural network to build a power prediction model for photovoltaic power plants, and use the training set to train the LSTM neural network; input the test set into the trained LSTM neural network to obtain short-term, ultra-short-term and medium-term power prediction results of photovoltaic power plants; S4, determining the assessment indicators of the photovoltaic power station, and calculating the values ​​of various assessment indicators according to the short-term, ultra-short-term and medium-term power prediction results obtained in step S3; S5. Compare the values ​​of each assessment indicator obtained in step S4 with the set assessment warning threshold. If the value of a single assessment indicator is greater than the warning threshold of the assessment indicator, a warning is issued to the photovoltaic power station.

2. The photovoltaic power station assessment and early warning method based on machine learning according to claim 1 is characterized in that: In step S1, the data set D1 includes the single machine information, booster station information and meteorological environment information of the photovoltaic power station; The single-machine information of a photovoltaic power station includes available power and actual power; The meteorological environment information includes light radiation intensity, atmospheric temperature, atmospheric humidity, cloud height, sun position, air pressure, wind speed and direction at a specified height, and light exposure time information.

3. The photovoltaic power station assessment and early warning method based on machine learning according to claim 1 or 2 is characterized in that: In step S2, the formula for normalizing the data set D1 is: In the formula, is the normalized data, represents the data item before normalization, Represents the minimum value of a data item, Indicates the maximum value of a data item.

4. The photovoltaic power station assessment and early warning method based on machine learning according to claim 3 is characterized in that: In step S4, the assessment indicators of the photovoltaic power station include power prediction accuracy, power prediction upload rate, automatic power generation control operation rate, automatic voltage control operation rate, automatic power generation control regulation qualification rate, automatic voltage control regulation qualification rate and active power change rate.

5. The photovoltaic power station assessment and early warning method based on machine learning according to claim 4 is characterized in that: The assessment indicators also include an E-text upload rate and an E-text pass rate. The E-text is a photovoltaic power station operation data reporting text used to report the photovoltaic power station operation data.

6. The photovoltaic power station assessment and early warning method based on machine learning according to claim 5 is characterized in that: In step S4, the power prediction accuracy includes short-term power prediction accuracy, ultra-short-term power prediction accuracy and medium-term power prediction accuracy; The calculation formula for the short-term power prediction accuracy is: ； In the formula, represents the short-term power prediction accuracy, Indicates the available power at time i in the power-limited state within 24 hours obtained from the E file, or the actual power at time i in the power-unlimited state; represents the predicted power at time i within 24 hours obtained from the E file, is the total installed capacity of the photovoltaic power station in that month; The calculation formula for ultra-short-term accuracy is: ； in, represents the ultra-short-term accuracy, Indicates the available power at time i in the power-limited state or the actual power at time i in the unlimited state obtained from the E file for the first 96 points; represents the predicted power of the first 96 points obtained from the E file for time i; The calculation formula for the accuracy of medium-term forecast is: ； in, represents the mid-term forecast accuracy result, Indicates the available power at time i in the power-limited state or the actual power at time i in the unlimited state obtained from the E file 4 days ago; It represents the predicted power at time i 4 days ago obtained from the E file.

7. The photovoltaic power station assessment and early warning method based on machine learning according to claim 6 is characterized in that: In step S4, the calculation formula of the E text pass rate is: ； In the formula, k is the scoring item of data scoring, represents the actual score of the kth item at time T, and K represents the maximum value of the scoring item.

8. The photovoltaic power station assessment and early warning method based on machine learning according to claim 4, 5, 6 or 7, characterized in that: In step S4, the calculation formula of the power prediction upload rate is: ； In the formula, The upload rate M represents the number of files successfully uploaded, and N represents the number of files scheduled to be uploaded; The calculation formula for the automatic generation control commissioning rate is: ； Where TF represents the automatic generation control operation rate, Indicates the automatic generation control substation exit time, Indicates the total stop time of the automatic generation control unit. Indicates free examination time. Indicates the total statistical time of the month; The calculation formula for the automatic voltage control operation rate is: ； In the formula, TD represents the operation rate of automatic voltage control, Indicates the automatic voltage control substation exit time, Indicates the total stop time of the automatic voltage control unit. Indicates free examination time. Indicates the total statistical time of the month; The calculation formula for the automatic generation control regulation pass rate is: ； In the formula, HF represents the automatic generation control regulation qualification rate, Indicates the number of qualified points of automatic power generation control substation. Indicates the number of times the master station sends adjustment instructions; The calculation formula for the automatic voltage control regulation pass rate is: ； In the formula, HD represents the automatic voltage control regulation pass rate, Indicates the number of qualified points of automatic voltage control substation execution; The calculation formula for the active power change rate is: ； In the formula, They represent the real-time common power of the two grid-connected devices at time i+j-1 and time i+j respectively, and i and j are both constants.

Images