KR102600462B1 - Photovoltaic Management Method Using Artificial Intelligence Algorithm - Google Patents

Abstract

The present invention relates to a solar power management system using an artificial intelligence algorithm.
The solar power management system according to the present invention is a sensor unit that is fixedly installed on each of the plurality of solar panels that make up the solar power generation system and acquires images taken of the surface of the solar panel and sensing data on temperature/humidity and illuminance. , an integrated sensing module that is installed away from the solar panel and collects voltage/current information applied to the solar panel and sensing data measured by the sensor unit, 3D imaging all or part of the solar power generation system An imaging device, an inclinometer that measures the inclination of the solar panel, a weather station that collects external environmental data where the solar power generation system is installed, and data acquired from the integrated sensing module, 3D imaging device, inclinometer, and weather station are already learned. A prediction device that predicts failure of the corresponding solar panel by inputting it into the completed prediction model, and collects monitoring result data, failure history, and previous analysis and prediction result data of the solar power generation system, and records the failure history. It includes a data collection device that builds a plurality of scenario-specific datasets using .

Description

Photovoltaic Management Method Using Artificial Intelligence Algorithm}

The present invention relates to a solar power management method using an artificial intelligence algorithm. More specifically, the present invention relates to a solar power management method using an artificial intelligence algorithm. More specifically, the data collected from the sensor unit installed in the solar panel and the sensing device installed around the solar panel are input into the artificial intelligence algorithm to manage the solar power panel. This relates to a solar power management method that predicts the condition of and performs maintenance.

Since solar power generation facilities produce electricity by directly utilizing sunlight outside the building, the amount of electrical energy generated varies depending on the management status.

In general, contamination of solar modules includes cracks, use of clamps other than standard, lightning, whitening/yellowing, and module damage due to hot spots. However, since large areas such as building rooftops and mountainous areas are often used, it is realistically impossible for people to check directly.

In addition, because it is operated without a clear understanding of the decline in power generation efficiency, it is operated with potential problems embedded in it, causing problems in performance/profit as the years go by.

Therefore, there is a need for technology that can check the amount of power generation and the cause of its decline from a remote location.

However, compared to the increasing number of power generation facilities, there are no companies that specialize in managing them, and most are managed directly by business owners and inspected on average twice a year. In addition, since the existing O&M service itself consists of periodic management rather than regular management, it takes a lot of time to identify the cause of the decrease in efficiency, making immediate response impossible.

Republic of Korea Patent Publication No. 10-2084784 (announced on May 26, 2020)

The technical task to be achieved by the present invention is to input data collected from sensor units installed on solar panels and sensing devices installed around solar panels into an artificial intelligence algorithm to predict the status of solar panels and perform maintenance. It is intended to provide a light management method.

To achieve this technical task, the solar power management method using an artificial intelligence algorithm according to an embodiment of the present invention is fixed to each of the plurality of solar panels constituting the solar power generation system, and includes images taken of the surface of the solar panels, Obtaining sensing data on temperature/humidity and illuminance;

Installed away from the solar panel, collecting voltage/current information applied to the solar panel and sensing data measured in the step;

Taking 3D photographs of all or part of the solar power generation system;

Measuring the inclination of the solar panel;

A weather observation device that collects external environmental data where the solar power generation system is installed,

Inputting data obtained from the integrated sensing module, 3D imaging device, inclinometer, and weather station into a pre-trained prediction model to predict whether the corresponding solar panel will fail;

And, collecting data as a result of monitoring the solar power generation system, failure history, and previous analysis and prediction result data, and constructing a plurality of data sets for each scenario using the failure history.

The step of obtaining images taken of the surface of the solar panel and sensing data on temperature/humidity and illuminance by being fixedly installed for each of the plurality of solar panels constituting the solar power generation system,

It is formed in the shape of an "ㄹ" and is fastened by inserting it from the side of the solar panel, and is equipped with a camera, an illuminance sensor, and a temperature and humidity sensor.

It is installed away from the solar panel, and the step of collecting voltage/current information applied to the solar panel and sensing data measured by the sensor unit includes,

Matched with one solar panel or multiple solar panels, obtaining information about the voltage and current applied to the solar panel,

An image of the surface of the solar panel, temperature, humidity, and illuminance values are obtained from a sensor unit installed on one side of the matched solar panel.

An imaging device for filming all or part of the solar power generation system,

An inclinometer that measures the inclination of the solar panel,

A weather observation device that collects external environmental data where the solar power generation system is installed,

Predicting whether the corresponding solar panel will fail by inputting data obtained from the integrated sensing module, 3D imaging device, inclinometer, and weather station into a pre-trained prediction model;

Generating a plurality of datasets for each scenario using the sensing data and environmental data received from the data collection device,

Input the received data sets for each scenario into a prediction model for which prior learning has been completed, and train the prediction model to predict the state of the solar panel,

Receive measured data from integrated sensing modules, imaging devices, inclinometers, and weather stations at the current time,

The received data is input into the trained prediction model to predict the state of the corresponding solar panel as at least one of a damaged state, a clean state, and a normal state.

Inputting the received data sets for each scenario into a prediction model for which prior learning has been completed and predicting the state of the solar panel by the prediction model,

Data augmentation techniques are applied to images included in multiple scenario-specific datasets to increase the amount of data to be learned.

Using the increased dataset, a prediction model built based on a deep learning model is supervised.

Creating datasets for multiple scenarios is,

There is at least one scenario among a state in which the solar panel is damaged, a state in which whitening or yellowing occurs in the solar panel, and a normal state.

By inputting the received sensing data into a trained prediction model to predict the state of the corresponding solar panel as at least one of a damaged state, a clean state, and a normal state, the cleanliness state is classified into a plurality of levels.

As such, according to the present invention, the cleanliness of the solar surface is confirmed using artificial intelligence technology such as deep learning, and the time required for cause analysis is reduced because a method is used to enable module-level confirmation of the entire array area. Although it is small, the accuracy of the method is high, and because continuous management is possible, it has the effect of responding to efficiency/profitability problems in the solar energy industry.

1 is a configuration diagram illustrating a solar energy management method according to an embodiment of the present invention.
FIG. 2 is an example diagram for explaining the sensor unit shown in FIG. 1.
Figure 3 is an example diagram for explaining a data collection system.
FIG. 4 is a configuration diagram for explaining the prediction device shown in FIG. 1.
Figure 5 is a flowchart for explaining a solar energy management method using a solar energy management system according to an embodiment of the present invention.
Figure 6 is an example diagram for explaining the prediction model built in step S520 shown in Figure 5.
Figure 7 is an example diagram to explain a solar panel that is damaged, whitened, or yellowed.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the attached drawings. In this process, the thickness of lines or sizes of components shown in the drawing may be exaggerated for clarity and convenience of explanation.

In addition, the terms described below are terms defined in consideration of functions in the present invention, and may vary depending on the measurement subject, operator's intention, or custom. Therefore, definitions of these terms should be made based on the content throughout this specification.

Hereinafter, a solar power management system according to an embodiment of the present invention will be described in more detail using FIGS. 1 to 4.

1 is a configuration diagram for explaining a solar power management system according to an embodiment of the present invention.

As shown in FIG. 1, the solar power management system according to an embodiment of the present invention includes a sensor unit 100, an integrated sensing module 200, a 3D imaging device 300, an inclinometer 400, and a weather station 500. , includes a prediction device 600 and a data collection device 700.

First, the sensor unit 100 is installed on one side of the solar panel.

FIG. 2 is an example diagram for explaining the sensor unit shown in FIG. 1.

As shown in FIG. 2, the sensor unit 100 includes a square-shaped main body 110, a support part 120 protruding downward from the bottom of the main body 110, and a support part 120 that protrudes in a direction perpendicular to the support part 120. It includes a plurality of fastening portions 130 formed at regular intervals.

The solar panel is interpolated between the plurality of coupling portions 130 so that the sensor unit 100 can be installed on the solar panel.

In addition, a camera is installed on the front of the main body 110 to capture the upper surface of the solar panel, and the captured image is transmitted to the integrated sensing module 200, which will be described later.

Additionally, an illuminance sensor is installed on the upper surface of the main body 110 to measure the illuminance of the solar panel, and a temperature and humidity sensor is mounted on the bottom to measure the surrounding temperature and humidity of the solar panel.

The sensor unit 100 acquires sensing data for image, illuminance, temperature and humidity, and transmits the acquired sensing data to the integrated sensing module 200 using RS485 communication.

The integrated sensing module 200 is matched one-to-one with one solar panel or N:1 with multiple solar panels and collects the voltage and current applied to the solar panels. Additionally, the integrated sensing module 200 collects sensing data received from the sensor unit 100.

Figure 3 is an example diagram for explaining a data collection system.

As shown in FIG. 3, the integrated sensing module 200 uses RS485 communication with sensor units matched at a 1:1 or N:1 ratio, and Ethernet, Wi-Fi, and RS485 communication with the prediction device 600, which will be described later. Data is transmitted and received using wired and wireless communications, etc.

Next, the 3D imaging device 300 is installed on one side of the solar power generation system and photographs part or the entire solar power generation system. At this time, the 3D imaging device 300 is for collecting environmental data and performs a different function from the camera installed in the sensor unit 100.

The inclinometer 400 is a sensor that collects the inclination of the solar panel, and allows the angle of the solar panel to be determined depending on the terrain where the solar panel is installed and the amount of sunlight.

The weather observer 500 is used to collect external environmental data of the solar power generation system, and collects at least one data from wind direction, wind speed, temperature, humidity, rainfall, atmospheric pressure, and solar radiation.

The prediction device 600 inputs the sensing data and environmental data obtained from the integrated sensing module 200, 3D imaging device 300, inclinometer 400, and weather observation device 500 into the learned prediction model to predict the solar panel. Predict whether or not there will be a failure.

Finally, the data collection device 700 collects a dataset for training the prediction model built by the prediction device 600. To elaborate, the data collection device 700 collects result data obtained by monitoring the solar power generation system. And, the data collection device 700 collects previous failure history. Additionally, the data collection device 700 collects analysis and prediction result data obtained from the previous prediction device 600.

FIG. 4 is a configuration diagram for explaining the prediction device shown in FIG. 1.

As shown in FIG. 4, the prediction device 600 according to an embodiment of the present invention includes a dataset generator 610, a learning unit 620, a data reception unit 630, and a prediction unit 640.

First, the dataset generator 610 creates a dataset using data received from the data collection device 700.

To elaborate, the dataset generator 610 receives data as a result of monitoring the solar power generation system, failure history, and previous analysis and prediction result data, and creates a dataset for training a prediction model using the received data. creates .

The learning unit 620 builds a prediction model on which pre-training has been completed, and inputs the generated dataset into the built prediction model to train it.

The data receiver 630 receives sensing data and environmental data measured at the current time.

The prediction unit 640 inputs the received sensing data and environmental data into a trained prediction model to predict the state of the solar panel.

Hereinafter, a solar power management method using a solar power management system according to an embodiment of the present invention will be described in more detail using FIGS. 5 to 7.

Figure 5 is a flowchart for explaining a solar energy management method using a solar energy management system according to an embodiment of the present invention.

As shown in FIG. 5, the prediction device 600 according to an embodiment of the present invention receives data from the data collection device 700 (S510).

To elaborate, the data collection device 700 collects monitoring data and failure history data input by managers, developers, and operators. Additionally, the data collection device 700 collects prediction data from the prediction device 600.

And the data collection device 700 transmits the collected monitoring data, failure history data, and prediction data to the prediction device 600.

The data set generator 610 creates a data set using the received monitoring data, failure history data, and prediction data. However, the data set generator 610 generates a plurality of data sets for each scenario using the failure history.

Here, the plurality of scenario-specific datasets have at least one scenario among a state in which the solar panel is damaged, a state in which the solar panel is bleached or yellowed, and a state in which the solar panel is normal.

Next, the learning unit 620 trains a prediction model using the generated data sets for each scenario (S520).

Figure 6 is an example diagram for explaining the prediction model built in step S520 shown in Figure 5.

As shown in FIG. 6, the learning unit 620 builds a prediction model based on a CNN model.

Then, the learning unit 620 trains a prediction model using a data set on damage and a data set on whitening or yellowing.

Figure 7 is an example diagram to explain a solar panel that is damaged, whitened, or yellowed.

Meanwhile, as shown in FIG. 7, when damage occurs to a solar panel, it is difficult to secure a large amount of data due to its nature, which may cause over-fitting of the prediction model. Therefore, the learning unit 620 increases the amount of data to be learned by applying a data augmentation technique to the images included in the dataset for each scenario,

Using the increased dataset, a prediction model built based on a deep learning model is supervised.

With learning completed, the data receiver 630 receives sensing data and environmental data from the integrated sensing module 200, the imaging device 300, the inclinometer 400, and the weather observer 500 (S530).

As described above, the data receiver 630 receives at least one sensing data from the integrated sensing module 200, including a photo containing information on voltage, current, and contamination of the solar module unit panel, temperature and humidity, and illuminance.

Additionally, the data receiver 630 receives environmental data including at least one of fine dust, weather information, solar radiation, slope, and image information captured by an IP network camera.

Next, the prediction unit 640 inputs the received sensing data and environmental data into the learned prediction model to predict the state of the solar panel (S540).

The prediction model predicts the state of the solar panel as at least one of damaged state, clean state, and normal state.

Meanwhile, the prediction unit 640 may classify the cleanliness status into multiple levels.

As such, the solar power management system according to the present invention utilizes artificial intelligence technology such as deep learning to check the cleanliness of the solar surface, and uses a system to enable module-by-module verification of the entire array area, allowing for cause analysis. It has the characteristics of reducing the time required and having high system accuracy, and because continuous management is possible, it has the effect of responding to efficiency/profitability issues in the solar energy industry.

The present invention has been described with reference to the embodiments shown in the drawings, but these are merely illustrative, and those skilled in the art will understand that various modifications and other equivalent embodiments are possible therefrom. will be. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the patent claims below.

100: sensor unit
200: Integrated sensing module
300: 3D imaging device
400: Inclinometer
500: weather station
600: prediction device
610: Dataset creation unit
620: Learning Department
630: Data receiving unit
640: prediction unit
700: data collection device

Claims (3)

In a solar power management method using an artificial intelligence algorithm,
Each of the plurality of solar panels constituting the solar power generation system is fixedly installed to obtain an image of the surface of the solar panel and sensing data on temperature/humidity and illuminance;
Installed away from the solar panel, collecting voltage/current information applied to the solar panel and sensing data measured in the step of acquiring the sensing data;
Taking 3D photographs of all or part of the solar power generation system;
Measuring the inclination of the solar panel;
Data acquired from weather observation devices, integrated sensing modules, 3D imaging devices, and inclinometers that collect external environmental data where the solar power generation system is installed are input into the prediction model learned by the learning unit to determine whether the corresponding solar panel is broken. predicting;
In addition, data on the results of monitoring the solar power generation system by managers, developers, and operators, failure history, and analysis and prediction result data from prediction devices are collected, and multiple scenario-specific datasets are constructed using the failure history. Including the step of doing;
The step of each of the plurality of solar panels constituting the solar power generation system being fixedly installed to obtain an image taken of the surface of the solar panel and sensing data on temperature/humidity and illuminance,
It is formed in the shape of an "ㄹ" and is fixed by inserting from the side of the solar panel, and includes a sensor unit equipped with a camera, an illuminance sensor, and a temperature and humidity sensor,
The data obtained from the weather station, integrated sensing module, 3D imaging device, and inclinometer that collect external environmental data where the solar power generation system is installed are input into the prediction model learned by the learning unit to prevent failure of the corresponding solar panel. The step of predicting whether
Generate multiple scenario-specific datasets using sensing data and environmental data received from the data collection device,
Input the plurality of scenario-specific data sets into a prediction model for which prior learning has been completed, and train the prediction model to predict the state of the solar panel,
Receive data obtained from the integrated sensing module, imaging device, inclinometer, and weather station at the current point in time,
By inputting the received data into a trained prediction model to predict the state of the corresponding solar panel as at least one of damaged state, clean state, and normal state,
The learning department,
An artificial intelligence algorithm that applies data augmentation techniques to images included in multiple scenario-specific datasets to increase the amount of data to be learned, and uses the increased dataset to supervise and learn a prediction model built based on a deep learning model. Solar energy management method using . According to paragraph 1,
It is installed away from the solar panel, and the step of collecting the sensing data measured in the step of acquiring the voltage/current information applied to the solar panel and the sensing data includes,
Matched with one solar panel or multiple solar panels, obtaining information about the voltage and current applied to the solar panel,
A solar power management method using an artificial intelligence algorithm to obtain images of the surface of the solar panel, temperature, humidity, and illuminance values from the sensor unit installed on one side of the matched solar panel. delete