Estimation system and method of power generation performance for PV system with bifacial solar modules using machine learning training

Abstract

The present invention relates to a system and method for estimating the power generation performance of a photovoltaic power generation system using a bifacial solar module through machine learning. More specifically, the present invention relates to a technology capable of improving estimation accuracy by utilizing time stamp information, specifically month data and date data, as input variables to perform inference of voltage and current values ​​of a photovoltaic power generation system using a learning model.

Description

Technology Field

[0001] The present invention relates to a system and method for estimating the power generation performance of a photovoltaic power generation system using a bifacial solar module through machine learning. More specifically, the invention relates to a system and method for estimating the power generation performance of a photovoltaic power generation system using a bifacial solar module through machine learning, which can estimate power generation performance data according to external conditions by collecting inverter output data from a photovoltaic power generation system using a bifacial solar module for a certain period, collecting environmental sensing data during the said period, and applying them to machine learning. Background Technology

[0002] Photovoltaic (PV) power generation is a method of producing electricity by converting sunlight into direct current. It utilizes solar panels composed of multiple solar cells, and as the demand for renewable and sustainable energy increases, the demand for photovoltaic power generation is also on the rise.

[0003] Such photovoltaic power generation is typically configured by connecting small-sized solar cells in series and parallel and connecting solar cell module units, which are combined with frames, in series and parallel. Solar cell modules connected in series are called solar cell strings, and solar cell strings connected in parallel are called solar cell arrays.

[0004] Recently, bifacial solar modules are being installed that can generate additional power by utilizing reflected light from the rear, in addition to solar power generation from the front.

[0005] Measuring the power generation of these bifacial solar modules is difficult because the rear gain cannot be accurately determined.

[0006] Furthermore, while the voltage characteristics through rear gain are similar to those of single-sided solar modules, the current characteristics vary depending on the structural shape of the solar power generation system using double-sided solar modules and the difference in the albedo value of the rear side. Consequently, there is a problem in that it is difficult to measure or estimate power generation due to the difference in rear gain.

[0007] Here, the albedo value refers to the ratio of frontal solar radiation (or power generation) to rearal solar radiation (or power generation) measured by a pyranometer.

[0008] In this regard, Korean registered patent No. 10-2401086 ("Solar power generation prediction system using a bifacial solar module model") discloses a technology that measures the optimal reflector installation area to obtain maximum power generation efficiency by considering the reflectivity actually acting on a bifacial solar module and economic feasibility, and predicts power generation efficiency based on the measurement results. Prior art literature

[0009] Korean Registered Patent No. 10-2401086 (Registration Date: May 18, 2022) The problem to be solved

[0010] Accordingly, the present invention has been devised to solve the problems of the prior art as described above. The objective of the present invention is to provide a system and method for estimating the power generation performance of a photovoltaic power generation system using a bifacial solar module through machine learning, which can estimate the current power generation performance by utilizing current external condition information (solar irradiance, solar cell module temperature, month, time, etc.) of the photovoltaic power generation system using a bifacial solar module. means of solving the problem

[0011] The power generation performance estimation system of a solar power generation system applying a bifacial solar module through machine learning according to the present invention for achieving the above-mentioned purpose preferably comprises: a bifacial basic input unit that generates a plurality of basic data sets including at least one of solar irradiance sensing data for each time stamp data during a predetermined period, temperature data of a solar cell module, measured output current value, and measured output voltage value from a solar power generation system applying a bifacial solar module; a cross-sectional calculation unit that generates a plurality of additional data sets including calculated current value and calculated voltage value corresponding to a solar power generation system applying a single-sided solar module by applying a stored algorithm and using the solar irradiance sensing data and temperature data included in each basic data set; a learning data generation unit that generates a learning data set for machine learning using the plurality of basic data sets and the plurality of additional data sets; a learning processing unit that performs learning processing of the input learning data set using a stored machine learning algorithm; and an estimation execution unit that estimates power generation performance data for a solar power generation system applying a bifacial solar module using a learning model by the learning processing unit.

[0012] Furthermore, the above-described cross-section calculation unit applies a stored algorithm to the above-described basic data set and uses the solar irradiance sensing data and temperature data included in each basic data set to calculate the maximum current value and maximum voltage value at the Maximum Power Point Tracking (MPPT) point to generate the above-described additional data set, wherein it is preferable that the solar power generation system applying the above-described single-sided solar module be assumed to be a power generation system applying a single-sided solar module corresponding to the same conditions as the solar power generation system applying the above-described double-sided solar module.

[0013] Furthermore, it is preferable that the learning data generation unit uses the plurality of basic data sets and the plurality of additional data sets to generate a plurality of learning data sets including at least one of solar radiation sensing data, temperature data, output current value, output voltage value, calculated current value, and calculated voltage value based on each time stamp data.

[0014] Furthermore, it is preferable that the above-mentioned training data generation unit further includes a data analysis unit that analyzes the generated training data set and extracts a valid training data set.

[0015] Furthermore, the data analysis unit further includes a first preprocessing unit that extracts a learning data set in which solar radiation sensing data for each learning data set satisfies a preset valid condition range by utilizing each of the plurality of learning data sets; a second preprocessing unit that extracts a learning data set in which the output current value satisfies a preset valid condition range by utilizing the learning data set extracted by the first preprocessing unit to configure an output current value measured based on solar radiation sensing data for each learning data set; and a third preprocessing unit that extracts a learning data set in which the output voltage value satisfies a preset valid condition range by utilizing the learning data set extracted by the second preprocessing unit to configure an output voltage value measured based on solar radiation sensing data for each basic data set. It is preferable that the learning processing unit performs learning processing using the learning data set extracted by the third preprocessing unit.

[0016] Furthermore, the above-mentioned estimation performing unit includes a learning output unit that receives time stamp information, solar irradiance sensing information, and temperature sensing information corresponding to a time when the estimation of power generation performance of a photovoltaic power generation system applying a bifacial photovoltaic module is desired, inputs these information into a learning model by the above-mentioned learning processing unit to receive estimated voltage and current values, and a power generation calculation unit that calculates the power generation amount using the estimated voltage and current values, and it is preferable that the above-mentioned power generation performance data includes at least one of the estimated voltage value, the estimated current value, and the calculated power generation amount.

[0017] A method for estimating the power generation performance of a photovoltaic power generation system applying a bifacial solar module through machine learning, wherein each step is performed by a computational processing means of the present invention for achieving the above-mentioned purpose, comprising: a basic set generation step in which a plurality of basic data sets are generated from a photovoltaic power generation system applying a bifacial solar module, including at least one of solar irradiance sensing data for each time stamp data during a preset period, temperature data of a solar cell module, a measured output current value, and a measured output voltage value, wherein in a bifacial basic input unit, a basic set generation step is generated from a photovoltaic power generation system applying a bifacial solar module, the plurality of additional data sets including calculated current values ​​and calculated voltage values ​​corresponding to a photovoltaic power generation system applying a bifacial solar module, by applying a stored algorithm in a cross-sectional calculation unit and using the solar irradiance sensing data and temperature data included in each basic data set by the basic set generation step; and in a learning data generation unit, using the plurality of basic data sets by the basic set generation step and the plurality of additional data sets by the additional set generation step, It is preferable to include a training set generation step for generating a plurality of training data sets for machine learning, a training processing step in which training processing of an input training data set is performed using a pre-stored machine learning algorithm in a training processing unit, and an estimation execution step in which power generation performance data for a photovoltaic power generation system with a bifacial solar module is estimated using a training model from the training processing step in an estimation execution unit.

[0018] Furthermore, the additional set generation step generates the additional data set by applying a stored algorithm and using the basic data set from the basic set generation step, and using the solar irradiance sensing data and temperature data included in each basic data set to calculate the maximum current value and maximum voltage value at the Maximum Power Point Tracking (MPPT) point, wherein it is preferable that the solar power generation system using the single-sided solar module is assumed to be a power generation system using a single-sided solar module corresponding to the same conditions as the solar power generation system using the double-sided solar module.

[0019] Furthermore, the above learning set generation step preferably generates a plurality of learning data sets including at least one of solar radiation sensing data, temperature data, output current value, output voltage value, output current value, and output voltage value based on each time stamp data, by utilizing a plurality of basic data sets generated by the above basic set generation step and a plurality of additional data sets generated by the above additional set generation step.

[0020] Furthermore, the method for estimating the power generation performance of a photovoltaic power generation system applying a bifacial solar module through the machine learning described above further includes a learning set extraction step, which analyzes a plurality of learning data sets generated by the learning set generation step to extract a valid learning data set prior to performing the learning processing step described above; the learning set extraction step comprises a first preprocessing step that uses each of the plurality of learning data sets generated by the learning set generation step to extract a learning data set in which the solar irradiance sensing data for each learning data set satisfies a preset valid condition range; a second preprocessing step that uses the learning data set extracted by the first preprocessing step to construct output current values ​​measured based on solar irradiance sensing data for each learning data set, thereby extracting a learning data set in which the output current values ​​satisfy a preset valid condition range; and using the learning data set extracted by the second preprocessing step to construct output voltage values ​​measured based on solar irradiance sensing data for each learning data set, thereby extracting a learning data set in which the output voltage values ​​satisfy a preset valid condition range. It includes a third preprocessing step, and it is preferable that the learning processing step performs learning processing using the learning data set extracted by the third preprocessing step.

[0021] Furthermore, the above-mentioned estimation execution step includes a learning output step that receives time stamp information, solar irradiance sensing information, and temperature sensing information corresponding to the time when estimation of power generation performance is desired for a photovoltaic power generation system applying a bifacial photovoltaic module, inputs the estimated voltage and current values ​​to the learning model, and outputs the estimated voltage and current values, and a power generation calculation step that calculates the power generation amount using the estimated voltage and current values ​​from the learning output step, and it is preferable that the power generation performance data includes at least one of the estimated voltage value, the estimated current value, and the calculated power generation amount. Effects of the invention

[0022] According to the present invention, the system and method for estimating the power generation performance of a photovoltaic power generation system using a bifacial solar module through machine learning has the advantage of improving estimation accuracy by utilizing month data and date data, which are time stamp information, as input variables to perform inference of voltage and current values ​​of the photovoltaic power generation system using a learning model.

[0023] In addition, there is an advantage in that the accuracy and reliability of the learning model itself can be improved by analyzing the generated training data set once again to extract only valid training data sets and using them for training processing. Brief explanation of the drawing

[0024] FIG. 1 is an example diagram showing the configuration of a system for estimating the power generation performance of a solar power generation system using a double-sided solar module through machine learning according to an embodiment of the present invention. FIG. 2 is a flowchart illustrating a method for estimating the power generation performance of a photovoltaic power generation system using a double-sided solar module through machine learning according to an embodiment of the present invention. Specific details for implementing the invention

[0025] Hereinafter, a fault diagnosis system and method for a photovoltaic power generation system using machine learning according to the present invention, having the configuration as described above, will be explained in detail with reference to the attached drawings. The drawings presented below are provided as examples to ensure that the concept of the present invention is sufficiently conveyed to those skilled in the art. Accordingly, the present invention is not limited to the drawings presented below and may be embodied in other forms. In addition, throughout the specification, the same reference numerals indicate the same components.

[0026] Unless otherwise defined, technical and scientific terms used herein have the meaning commonly understood by those skilled in the art to which this invention pertains, and descriptions of known functions and configurations that could unnecessarily obscure the essence of the invention are omitted in the following description and accompanying drawings.

[0027] Furthermore, a system refers to a set of components, including devices, mechanisms, and means, that are organized and interact regularly to perform necessary functions.

[0028] A system and method for estimating the power generation performance of a photovoltaic power generation system using a bifacial solar module through machine learning according to an embodiment of the present invention relates to a technology capable of inferring and estimating the current and voltage values ​​of a photovoltaic power generation system using a bifacial solar module based on input solar irradiance and solar module temperature by applying solar irradiance and the temperature of the solar module to a pre-stored physical model equation to calculate current and voltage values ​​for a power generation system using a single-sided solar module that corresponds to the same conditions as a photovoltaic power generation system using a bifacial solar module, i.e., a front-side solar module constituting the bifacial solar module, and a power generation system using a single-sided solar module corresponding to the same conditions as a photovoltaic power generation system using a bifacial solar module, and by performing learning processing of the measured current and voltage values ​​of the photovoltaic power generation system using the bifacial solar module and the calculated current and voltage values ​​using a machine learning algorithm. Furthermore, the invention relates to a technology capable of calculating the estimated power generation amount using the inferred current and voltage values.

[0029] Through this, there is an advantage in being able to accurately infer / estimate the power generation of a solar power generation system using bifacial solar modules with various structural forms.

[0030] A system for estimating the power generation performance of a solar power generation system using a double-sided solar module through machine learning according to one embodiment of the present invention is preferably configured to include a double-sided basic input unit (100), a cross-sectional calculation unit (200), a learning data generation unit (300), a learning processing unit (400), and an estimation execution unit (400), as illustrated in FIG. 1. The basic input unit (100), the cross-sectional calculation unit (200), the learning data generation unit (300), and the learning processing unit (400) are linked to a solar power generation system using a double-sided solar module for which power generation performance estimation is desired, and perform operations by collecting inverter output data of the solar power generation system in advance.

[0031] The above-mentioned estimation execution unit (500) uses a learning model and is linked to a solar power generation system that applies a double-sided solar module for which power generation performance is desired, and performs the operation of estimating and outputting power generation performance data according to input current conditions or power generation performance data according to input predetermined conditions.

[0032] To perform such operations, it is preferable that the above components be configured individually or integratedly in a computational processing means including an MCU to perform the operations.

[0033] Let's take a closer look at each component.

[0034] The above-described double-sided basic input unit (100) is linked to a solar power generation system that applies a double-sided solar module for which power generation performance is desired, and it is preferable to generate a plurality of basic data sets including at least one of solar irradiance sensing data for each time stamp data during a predetermined period, temperature data of a solar cell module, measured output current value, and measured output voltage value from the solar power generation system that applies the double-sided solar module.

[0035] At this time, in the present invention, it is preferable to collect time stamp data, solar irradiance sensing data, temperature data, measured output current value, and measured output voltage value from a photovoltaic power generation system with a double-sided solar module applied for at least 6 months, and to generate a basic data set based on each time stamp data.

[0036] Here, time stamp data refers to a time displacement parameter indicating a starting point in time. In this invention, since solar irradiance has the greatest influence on solar power generation and the month and time are the major points of change for this irradiance, the time stamp data is constructed by including month data and time data.

[0037] It is preferable that the cross-section calculation unit (200) applies a pre-stored algorithm and generates a plurality of additional data sets including calculated current values ​​and calculated voltage values ​​corresponding to a photovoltaic power generation system with a cross-section photovoltaic module by using solar radiation sensing data and temperature data included in each basic data set generated by the double-sided basic input unit (100).

[0038] Here, the solar power generation system applying the above-mentioned single-sided solar module refers to the front module of the solar power generation system applying the above-mentioned double-sided solar module, which is assumed to be a power generation system applying a single-sided solar module corresponding to the same conditions as the solar power generation system applying the above-mentioned double-sided solar module; in other words, it refers to the front module of the solar power generation system applying the double-sided solar module for which the estimation of connected power generation performance is desired.

[0039] In addition, it is desirable that the above additional data set includes solar radiation sensing data and temperature data used together with the calculated current value and the calculated voltage value.

[0040] In detail, it is preferable that the cross-section calculation unit (200) generates an additional data set using each basic data set generated by the double-sided basic input unit (100) by using a pre-stored algorithm, that is, using a mathematical physical model equation.

[0041] As described above, the cross-section calculation unit (200) calculates power and voltage values ​​by the front module of a solar power generation system that applies a double-sided solar module for which power generation performance is to be estimated. To this end, the solar irradiance meter that generates solar irradiance sensing data constituting the basic data set must be installed at the same angle of inclination as the solar module of the solar power generation system is installed / arranged.

[0042] Based on this, the cross-section calculation unit (200) uses solar radiation sensing data and temperature data included in each basic data set to derive a regression model equation from the change in voltage according to the amount of light at the MPPT (Maximum Power Point Tracking) point for each amount of light received in advance from the manufacturer of the solar module, and calculates the maximum voltage value.

[0043] In addition, a regression model equation is derived from the change in current according to light intensity to calculate the maximum current value. Using this, the voltage and current values ​​of the photovoltaic power generation system are inferred and calculated based on a cross-sectional view (front side).

[0044] That is, reflecting the characteristics of photovoltaic power generation in which power output changes according to solar irradiance, the solar irradiance sensing data and temperature data constituting the above basic data set are used to plot the measured voltage and measured current values ​​for each light intensity at the maximum output point tracking point, and curve fitting is performed on the maximum voltage and maximum current values ​​at the maximum output point tracking point. The physical model of the maximum voltage and maximum current values ​​is derived through the fitting of the maximum voltage value (V_mp) (calculated voltage value) and the maximum current value (I_mp) (calculated current value) as shown in Equation 1 below.

[0045]

[0046] Here, is the voltage coefficient, and

[0047] is the solar irradiance on a slope (W / m² 2 ) and,

[0048] is the voltage-temperature coefficient (% / ℃), and

[0049] is temperature data (°C), and

[0050] is the current coefficient, and

[0051] is the current temperature coefficient (% / ℃).

[0052] It is preferable that the above-described learning data generation unit (300) generates a learning data set for machine learning by using a plurality of basic data sets from the above-described double-sided basic input unit (100) and a plurality of additional data sets from the above-described single-sided calculation unit (200).

[0053] In detail, the learning data generation unit (300) uses the plurality of basic data sets and the plurality of additional data sets to generate a plurality of learning data sets including at least one of solar radiation sensing data, temperature data, output current value, output voltage value, calculated current value, and calculated voltage value based on each time stamp data.

[0054] That is, the training data set generates a single data set of current / voltage values ​​for a solar power generation system using a double-sided solar module and a solar power generation system using a single-sided solar module (front module) under the same external environmental conditions (date, time, solar irradiance, temperature). Of course, the training data generation unit (300) generates a training data set by performing normalization on each of the basic data set by the double-sided basic input unit (100) and the additional data set by the single-sided calculation unit (200), and then performing data combination based on each time stamp data.

[0055] The above learning processing unit (400) performs learning processing of the input learning data set using a machine learning algorithm stored in advance.

[0056] At this time, in order to improve the accuracy and reliability of the learning model generated through the learning processing unit (400), the power generation performance estimation system of a solar power generation system applying a double-sided solar module according to one embodiment of the present invention preferably performs a validity review of a plurality of learning data sets generated by the learning data generation unit (300).

[0057] To this end, as illustrated in FIG. 1, the learning data generation unit (300) further includes a data analysis unit (310).

[0058] It is preferable that the data analysis unit (310) analyzes the generated training data set and extracts a valid training data set.

[0059] As shown in FIG. 1, the data analysis unit (310) includes a first preprocessing unit (311), a second preprocessing unit (312), and a third preprocessing unit (313).

[0060] The data analysis unit (310) receives a plurality of generated training data sets and extracts a training data set that satisfies all conditions through the first preprocessing unit (311), the second preprocessing unit (312), and the third preprocessing unit (313). That is, each component of the data analysis unit (310) operates under an AND condition.

[0061] It is preferable that the first preprocessing unit (311) uses each of the plurality of training data sets to extract a training data set in which the daily sensing data for each training data set satisfies a preset valid condition range.

[0062] In this case, the preset effective condition range is 300 W / m 2 It is advisable to set it below. That is, when the solar irradiance sensing data is ≥ 300 W / m² 2 The training data set is extracted.

[0063] It is preferable that the second preprocessing unit (312) uses the learning data set extracted (first extraction) by the first preprocessing unit (311) to construct output current values ​​measured based on solar irradiance sensing data for each learning data set. That is, output current values ​​according to solar irradiance sensing data are graphed in the form of a linear regression curve. It is preferable to extract a learning data set in which the output current value satisfies a preset valid condition range through graph analysis.

[0064] At this time, it is desirable to set the pre-set valid condition range to within ± 10% of the linear regression curve. That is, the training data set extracted first by the first preprocessing unit (311) is analyzed, and a training data set in which the output current value is within ± 10% of the linear regression curve is extracted secondarily.

[0065] The third preprocessing unit (313) uses the training data set extracted (second extraction) by the second preprocessing unit (312) to graph the output voltage values ​​measured based on solar radiation sensing data for each training data set. It is desirable to extract training data sets in which the output voltage values ​​satisfy a preset valid condition range through graph analysis.

[0066] At this time, it is desirable to set the preset effective condition range to within ± 15% of the rated voltage of the solar module. That is, the learning data set extracted secondarily by the second preprocessing unit (312) is analyzed to finally extract a learning data set in which the output voltage value is within ± 15% of the rated voltage.

[0067] Through this, the learning processing unit (400) performs learning processing using the learning data set extracted by the third preprocessing unit (313).

[0068] The machine learning algorithm stored for learning in the above-mentioned learning processing unit (400) is preferably a regression learning algorithm, and for example, it is preferable to include a linear algorithm, a neural algorithm, an SVM algorithm, a Gauss algorithm, etc., but is not limited thereto.

[0069] It is preferable that the above estimation execution unit (500) estimates power generation performance data for a solar power generation system with a double-sided solar module applied using a learning model after the learning processing is completed by the above learning processing unit (400).

[0070] It is preferable that the above power generation performance data includes at least one of an estimated voltage value, an estimated current value, and a calculated power generation amount.

[0071] The above-mentioned estimation execution unit (500) includes a learning output unit (510) and a power generation amount calculation unit (520), as shown in FIG. 1.

[0072] The above learning output unit (510) receives time stamp information, solar irradiance sensing information, and temperature sensing information corresponding to the time when the power generation performance of a solar power generation system with a double-sided solar module is to be estimated, inputs this information into a learning model by the above learning processing unit (400), and outputs the estimated voltage value and current value.

[0073] The estimated voltage and current values ​​output by the above-mentioned learning output unit (510) have the advantage of overcoming the limitations of a solar power generation system using a double-sided solar module, which is difficult to estimate because the rear gain cannot be accurately known.

[0074] In addition, since the estimated voltage value and estimated current value output by the learning output unit (510) are the current value / voltage value expected when the photovoltaic power generation system with the corresponding double-sided photovoltaic module operates ideally, they can be used as a standard to determine the operating state (normal state or abnormal state) of the photovoltaic power generation system by comparing them with the actual measured voltage value / current value.

[0075] The above power generation calculation unit (520) calculates the power generation amount (P, kw) using the estimated voltage value and estimated current value from the above learning output unit (510).

[0076] This can be simply calculated / inferred through the formula (Voltage * Current = Power).

[0077] FIG. 2 is a sequence example diagram illustrating a method for estimating the power generation performance of a photovoltaic power generation system using a bifacial solar module through machine learning according to an embodiment of the present invention. The method for estimating the power generation performance of a photovoltaic power generation system using a bifacial solar module through machine learning according to an embodiment of the present invention may include a basic set generation step (S100), an additional set generation step (S200), a learning set generation step (S300), a learning processing step (S400), and an estimation execution step (S500). It is preferable to use a system for estimating the power generation performance of a photovoltaic power generation system using a bifacial solar module through machine learning, which is operated by a computation processing means for each step.

[0078] Let's take a closer look at each step.

[0079] The above basic set generation step (S100) is linked to a solar power generation system with a double-sided solar module for which power generation performance is desired in the above-determined double-sided basic input unit (100), and generates a plurality of basic data sets including at least one of solar irradiance sensing data for each time stamp data during a predetermined period, temperature data of a solar cell module, measured output current value, and measured output voltage value from the solar power generation system with the double-sided solar module.

[0080] At this time, in the present invention, it is preferable to collect time stamp data, solar irradiance sensing data, temperature data, measured output current value, and measured output voltage value from a photovoltaic power generation system with a double-sided solar module applied for at least 6 months, and to generate a basic data set based on each time stamp data.

[0081] Here, time stamp data refers to a time displacement parameter indicating a starting point in time. In this invention, since solar irradiance has the greatest influence on solar power generation and the month and time are the major points of change for this irradiance, the time stamp data is constructed by including month data and time data.

[0082] The above additional set generation step (S200) generates a plurality of additional data sets including calculated current values ​​and calculated voltage values ​​corresponding to a photovoltaic power generation system with a cross-sectional photovoltaic module by applying a pre-stored algorithm in the cross-sectional calculation unit (200) and using the solar radiation sensing data and temperature data included in each basic data set generated by the above basic set generation step (S100).

[0083] Here, the solar power generation system applying the above-mentioned single-sided solar module refers to the front module of the solar power generation system applying the above-mentioned double-sided solar module, which is assumed to be a power generation system applying a single-sided solar module corresponding to the same conditions as the solar power generation system applying the above-mentioned double-sided solar module; in other words, it refers to the front module of the solar power generation system applying the double-sided solar module for which the estimation of connected power generation performance is desired.

[0084] In addition, it is desirable that the above additional data set includes solar radiation sensing data and temperature data used together with the calculated current value and the calculated voltage value.

[0085] In detail, the additional set generation step (S200) generates additional data sets using each basic data set by using a pre-stored algorithm, that is, using a mathematical physical model equation.

[0086] The power and voltage values ​​are calculated by the front module of a solar power generation system using a bifacial solar module for which the connected power generation performance is to be estimated. To this end, the solar irradiance meter that generates solar irradiance sensing data constituting the above basic data set must be installed at the same angle of inclination as the solar module of the solar power generation system is installed / arranged.

[0087] Based on this, the additional set generation step (S200) uses solar irradiance sensing data and temperature data included in each basic data set to derive a regression model equation from the change in voltage according to light intensity at the MPPT (Maximum Power Point Tracking) point for each light intensity received in advance from the manufacturer of the solar module, and calculates the maximum voltage value.

[0088] In addition, a regression model equation is derived from the change in current according to light intensity to calculate the maximum current value. Using this, the voltage and current values ​​of the photovoltaic power generation system are inferred and calculated based on a cross-sectional view (front side).

[0089] That is, reflecting the characteristics of photovoltaic power generation in which power output changes depending on solar irradiance, the solar irradiance sensing data and temperature data constituting the above basic data set are used to plot the measured voltage and measured current values ​​for each light intensity at the maximum output point tracking point, and curve fitting is performed on the maximum voltage and maximum current values ​​at the maximum output point tracking point. The physical model of the maximum voltage and maximum current values ​​is derived as shown in Equation 1 above through the fitting of the maximum voltage value (V_mp) (calculated voltage value) and the maximum current value (I_mp) (calculated current value).

[0090] The above learning set generation step (S300) generates a learning data set for machine learning using a plurality of basic data sets from the above basic set generation step (S100) and a plurality of additional data sets from the above additional set generation step (S200) in the above learning data generation unit (300).

[0091] In detail, the learning set generation step (S300) generates a plurality of learning data sets including at least one of solar radiation sensing data, temperature data, output current value, output voltage value, calculated current value, and calculated voltage value based on each time stamp data, using the plurality of basic data sets and a plurality of additional data sets.

[0092] In other words, the training dataset is created by combining the current / voltage values ​​of a solar power generation system using bifacial solar modules and a solar power generation system using monofacial solar modules (front modules) under the same external environmental conditions (date, time, solar irradiance, temperature) into a single dataset. Of course, after performing normalization on each of the input basic dataset and additional dataset, the training dataset is created by combining the data based on each time stamp data.

[0093] The above learning processing step (S400) performs learning processing of the input learning data set using a machine learning algorithm stored in advance in the above learning processing unit (400).

[0094] At this time, the method for estimating the power generation performance of a solar power generation system using a double-sided solar module through machine learning according to one embodiment of the present invention further includes a learning set extraction step (S10) for performing a validity review of a plurality of learning data sets in order to improve the accuracy and reliability of the generated learning model prior to performing the learning processing step (S400), as illustrated in FIG. 2.

[0095] The above learning set extraction step (S10) involves the data analysis unit (310) analyzing a plurality of learning data sets generated by the learning set generation step (S300) to extract a valid learning data set.

[0096] To this end, the learning set extraction step (S10) includes a first preprocessing step (S11), a second preprocessing step (S12), and a third preprocessing step (S13).

[0097] The first preprocessing step (S11) uses each of the multiple training data sets generated by the training set generation step (S300) to extract a training data set in which the daily sensing data for each training data set satisfies a preset valid condition range.

[0098] In this case, the preset effective condition range is 300 W / m 2 It is advisable to set it below. That is, when the solar irradiance sensing data is ≥ 300 W / m² 2 The training data set is extracted.

[0099] The second preprocessing step (S12) uses the training data set extracted (first extraction) by the first preprocessing step (S11) to construct output current values ​​measured based on solar irradiance sensing data for each training data set. That is, output current values ​​according to solar irradiance sensing data are graphed in the form of a linear regression curve. Through graph analysis, training data sets in which output current values ​​satisfy a preset valid condition range are extracted.

[0100] At this time, it is desirable to set the pre-set valid condition range to within ± 10% of the linear regression curve. That is, the training data set extracted first by the first preprocessing step (S11) is analyzed, and a training data set in which the output current value is within ± 10% of the linear regression curve is extracted secondarily.

[0101] The third preprocessing step (S13) uses the training data set extracted (second extraction) by the second preprocessing step (S12) to graph the output voltage values ​​measured based on solar irradiance sensing data for each training data set. Through graph analysis, training data sets in which the output voltage values ​​satisfy a preset valid condition range are extracted.

[0102] At this time, it is desirable to set the pre-set valid condition range to within ± 15% of the rated voltage of the solar module. That is, the learning data set extracted secondarily by the second preprocessing step (S12) is analyzed to finally extract a learning data set in which the output voltage value is within ± 15% of the rated voltage.

[0103] In this way, the first preprocessing step (S11), the second preprocessing step (S12), and the third preprocessing step (S13) operate under an AND condition, and through this, a training data set satisfying all valid condition ranges is extracted.

[0104] Through this, the learning processing step (S400) performs learning processing using the learning data set extracted by the third preprocessing step (S13).

[0105] The above learning processing step (S400) preferably uses a machine learning algorithm stored for learning that is a regression learning algorithm, and preferably includes, for example, a linear algorithm, a neural algorithm, an SVM algorithm, a Gauss algorithm, etc., but is not limited thereto.

[0106] The above estimation execution step (S500) estimates power generation performance data for a photovoltaic power generation system with a bifacial photovoltaic module applied using a learning model after the learning processing is completed by the above estimation execution unit (500).

[0107] It is preferable that the above power generation performance data includes at least one of an estimated voltage value, an estimated current value, and a calculated power generation amount.

[0108] The above estimation execution step (S500) operates as a learning output step (S510) and a power generation calculation step (S520), as shown in FIG. 2.

[0109] The above learning output step (S510) receives time stamp information, solar irradiance sensing information, and temperature sensing information corresponding to the time when the power generation performance of a solar power generation system using a double-sided solar module is desired, inputs them into a learning model according to the above learning processing step (S400), and receives the estimated voltage and current values.

[0110] The output estimated voltage and current values ​​have the advantage of overcoming the limitations of solar power generation systems using bifacial solar modules, where estimation is difficult because the rear gain cannot be accurately known.

[0111] In addition, since the output estimated voltage and current values ​​are the current and voltage values ​​expected when a photovoltaic power generation system equipped with the corresponding bifacial solar module operates ideally, they can be used as a standard to determine the operating state (normal or abnormal state) of the photovoltaic power generation system by comparing them with the actual measured voltage and current values.

[0112] The above power generation calculation step (S520) calculates the power generation amount (P, kW) using the estimated voltage value and estimated current value obtained from the above learning output step (S510). This can be simply calculated / inferred through the formula (voltage * current = power).

[0113] As such, the power generation performance estimation system of a photovoltaic power generation system applying a bifacial photovoltaic module through machine learning according to one embodiment of the present invention is characterized by estimating the power generation performance data by using information including time stamp information corresponding to the time point for which estimation is desired as input variables.

[0114] Through various experiments, verification was performed on the estimation accuracy of the learning model based on whether or not timestamp information (month information, time information) was included.

[0115] The results of calculating the RMSE (Root Mean Square Error) using training and validation data in which solar radiation sensing information, temperature information, and voltage / current values ​​were set as input variables are shown in Table 1 below.

[0116] V_RMSE I_RMSE P_RMSE training data 4.6227 0.2970 0.0677 Validation data 4.5852 0.2980 0.0678

[0117] In addition, as in the present invention, the results of calculating the RMSE using training and verification data in which time stamp information (month information, hour information), solar radiation sensing information, temperature information, and voltage / current values ​​are set as input variables are as shown in Table 2 below.

[0118] V_RMSE I_RMSE P_RMSE training data 2.6859 0.1631 0.0396 Validation data 4.0164 0.2370 0.0566

[0119] As such, it was found that including time stamp information lowered the RMSE by more than 20%.

[0120] Meanwhile, in one embodiment of the present invention, a method for estimating the power generation performance of a photovoltaic power generation system applying a bifacial photovoltaic module through machine learning may be implemented in the form of program instructions that can be executed through various means of electronically processing information and recorded on a storage medium. The storage medium may include program instructions, data files, data structures, etc., either individually or in combination.

[0121] Program instructions recorded on a storage medium may be those specifically designed and configured for the present invention, or they may be those known and available to those skilled in the art of software. Examples of storage media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a device that processes information electronically using an interpreter, such as a computer.

[0122] As described above, the present invention has been explained with specific details such as specific constituent elements and limited exemplary drawings; however, this is provided merely to aid in a more comprehensive understanding of the invention, and the invention is not limited to the above-mentioned exemplary embodiment. Those skilled in the art can make various modifications and variations from this description.

[0123] Accordingly, the scope of the present invention should not be limited to the described embodiments, and all things equivalent to or having equivalent variations to the claims set forth below, as well as the claims set forth below, shall be considered to fall within the scope of the concept of the present invention. Explanation of the symbols

[0124] 100 : Double-sided basic input section 200 : Section Calculation Unit 300 : Training data generation section 310 : Data Analysis Department 311: 1st Preprocessing Unit 312: 2nd Preprocessing Unit 313 : 3rd Preprocessing Unit 400 : Learning processing unit 500 : Estimation Execution Unit 510: Learning Output Unit 520: Power Generation Calculation Unit

Claims

Claim 1 A double-sided basic input unit that generates a plurality of basic data sets including at least one of solar irradiance sensing data for each time stamp data during a preset period, temperature data of a solar cell module, a measured output current value, and a measured output voltage value for a solar power generation system using a double-sided solar module; a single-sided calculation unit that generates a plurality of additional data sets including calculated current values ​​and calculated voltage values ​​corresponding to a solar power generation system using a single-sided solar module by applying a stored algorithm and using the solar irradiance sensing data and temperature data included in each basic data set; a learning data generation unit that generates a learning data set for machine learning using the plurality of basic data sets and the plurality of additional data sets; and a learning processing unit that performs learning processing of the input learning data set using a stored machine learning algorithm. The method comprises: an estimation execution unit that estimates power generation performance data including at least one of an estimated voltage value, an estimated current value, and a calculated power generation amount for a photovoltaic power generation system with a bifacial solar module, using a learning model by the learning processing unit; wherein the cross-section calculation unit generates the additional data set by applying a stored algorithm, using the basic data set, and using solar irradiance sensing data and temperature data included in each basic data set to calculate the maximum current value and the maximum voltage value at the Maximum Power Point Tracking (MPPT) point; and the estimation execution unit inputs time stamp information, solar irradiance sensing information, and temperature sensing information corresponding to the time when estimation of power generation performance is desired for a photovoltaic power generation system with a bifacial solar module to the learning model by the learning processing unit, and receives the estimated voltage value and the estimated current value as outputs; and a power generation amount calculation unit that calculates the power generation amount using the estimated voltage value and the estimated current value.A system for estimating the power generation performance of a photovoltaic power generation system using bifacial solar modules through machine learning, further including Claim 2 In claim 1, the solar power generation system applying the single-sided solar module is assumed to be a power generation system applying a single-sided solar module corresponding to the same conditions as the solar power generation system applying the double-sided solar module, a system for estimating the power generation performance of a solar power generation system applying a double-sided solar module through machine learning. Claim 3 In claim 2, the learning data generation unit generates a plurality of learning data sets including at least one of solar irradiance sensing data, temperature data, output current value, output voltage value, calculated current value, and calculated voltage value based on each time stamp data, using the plurality of basic data sets and a plurality of additional data sets, a system for estimating the power generation performance of a photovoltaic power generation system applying a bifacial photovoltaic module through machine learning. Claim 4 In claim 3, the system for estimating the power generation performance of a photovoltaic power generation system applying a bifacial photovoltaic module through machine learning further comprises: a data analysis unit that analyzes the generated training data set and extracts a valid training data set. Claim 5 In claim 4, the data analysis unit further comprises: a first preprocessing unit that uses each of the plurality of learning data sets to extract a learning data set in which solar irradiance sensing data for each learning data set satisfies a preset valid condition range; a second preprocessing unit that uses the learning data set extracted by the first preprocessing unit to configure output current values ​​measured based on solar irradiance sensing data for each learning data set, and extracts a learning data set in which output current values ​​satisfy a preset valid condition range; and a third preprocessing unit that uses the learning data set extracted by the second preprocessing unit to configure output voltage values ​​measured based on solar irradiance sensing data for each basic data set, and extracts a learning data set in which output voltage values ​​satisfy a preset valid condition range; wherein the learning processing unit performs learning processing using the learning data set extracted by the third preprocessing unit, a system for estimating the power generation performance of a photovoltaic power generation system applying a bifacial photovoltaic module through machine learning. Claim 6 delete Claim 7 A method for estimating the power generation performance of a photovoltaic power generation system applying a bifacial solar module through machine learning, wherein each step is performed by a computational processing means, comprising: a basic set generation step in which, at a bifacial basic input unit, a plurality of basic data sets are generated from a photovoltaic power generation system applying a bifacial solar module, including at least one of solar irradiance sensing data for each time stamp data during a preset period, temperature data of a solar cell module, a measured output current value, and a measured output voltage value; an additional set generation step in which, at a single-sided calculation unit, a plurality of additional data sets are generated by applying a stored algorithm and using the solar irradiance sensing data and temperature data included in each basic data set by the basic set generation step, and including a calculated current value and a calculated voltage value corresponding to a photovoltaic power generation system applying a single-sided solar module; and, at a learning data generation unit, a plurality of learning data for machine learning are generated by using the plurality of basic data sets by the basic set generation step and the plurality of additional data sets by the additional set generation step. A learning set generation step for generating a set; a learning processing step in which, in a learning processing unit, learning processing of an input learning data set is performed using a pre-stored machine learning algorithm; and an estimation execution step in which, in an estimation execution unit, at least one of an estimated voltage value, an estimated current value, and a calculated power generation amount for a photovoltaic power generation system applying a bifacial photovoltaic module is estimated using a learning model from the learning processing step;A method for estimating the power generation performance of a photovoltaic power generation system applying a bifacial solar module through machine learning, comprising: an additional set generation step that generates the additional data set by applying a stored algorithm, using the basic data set from the basic set generation step, using solar irradiance sensing data and temperature data included in each basic data set, and calculating the maximum current value and maximum voltage value at the Maximum Power Point Tracking (MPPT) point; and an estimation execution step that inputs time stamp information, solar irradiance sensing information, and temperature sensing information corresponding to the time when estimation of power generation performance is desired for a photovoltaic power generation system applying a bifacial solar module into a learning model from the learning processing step, and receives an estimated voltage value and an estimated current value as output; and a power generation amount calculation step that calculates the power generation amount using the estimated voltage value and the estimated current value from the learning output step. Claim 8 A method for estimating the power generation performance of a solar power generation system using a bifacial solar module through machine learning, wherein, in claim 7, the solar power generation system using the single-sided solar module is assumed to be a power generation system using a single-sided solar module corresponding to the same conditions as the solar power generation system using the bifacial solar module. Claim 9 In claim 8, the method for estimating the power generation performance of a photovoltaic power generation system applying a bifacial photovoltaic module through machine learning, wherein the learning set generation step generates a plurality of learning data sets including at least one of solar irradiance sensing data, temperature data, output current value, output voltage value, output current value, and output voltage value based on each time stamp data, using a plurality of basic data sets generated by the basic set generation step and a plurality of additional data sets generated by the additional set generation step. Claim 10 In claim 9, the method for estimating the power generation performance of a photovoltaic power generation system applying a bifacial photovoltaic module through machine learning further comprises, prior to performing the learning processing step, a learning set extraction step of analyzing a plurality of learning data sets generated by the learning set generation step to extract a valid learning data set; wherein the learning set extraction step comprises: a first preprocessing step of using each of the plurality of learning data sets generated by the learning set generation step to extract a learning data set in which solar irradiance sensing data for each learning data set satisfies a preset valid condition range; and a second preprocessing step of using the learning data set extracted by the first preprocessing step to construct output current values ​​measured based on solar irradiance sensing data for each learning data set, and extracting a learning data set in which the output current values ​​satisfy a preset valid condition range. A method for estimating the power generation performance of a photovoltaic power generation system applying a bifacial photovoltaic module through machine learning, comprising: a third preprocessing step of using the learning data set extracted by the second preprocessing step to configure output voltage values ​​measured based on solar irradiance sensing data for each learning data set, and extracting a learning data set in which the output voltage value satisfies a preset valid condition range; wherein the learning processing step performs learning processing using the learning data set extracted by the third preprocessing step. Claim 11 delete

Images