KR102309077B1 - Solar power generation system and method applying sensor-based safety diagnosis technology - Google Patents

Abstract

A photovoltaic power generation system applied with a sensor-based safety diagnosis technology according to an embodiment of the present invention includes: a plurality of sensors which are to collect sensing information including voltage, current, insolation, temperature, humidity, atmospheric information, inclination information, and wind speed/wind direction, to identify the state of photovoltaic power generation and the state of a photovoltaic power generation structure; a connection board which receives the sensing information from the plurality of sensors to provide the same to a management server, performs structural safety diagnosis by using the inclination information on an inclination angle of the photovoltaic power generation structure, and receives a structure safety diagnosis result from the management server to control the inclination angle of the photovoltaic power generation structure by using a controller; and the management server which performs safety diagnosis of the photovoltaic power generation structure by using the inclination information collected through the connection board.

Description

Solar power generation system and method applying sensor-based safety diagnosis technology

The present invention relates to a photovoltaic power generation system and method, and more particularly, a sensor-based safety diagnosis technology capable of predicting the lifespan of a photovoltaic facility structure based on a sensor and a neural network algorithm and systematically performing safety diagnosis. It relates to an applied solar power generation system and method.

There is always the risk of accidents with solar cells and photovoltaic facilities exposed to harsh external environments. In addition, due to the nature of the solar power plant, it is necessary to secure a large site for the construction of the power plant, but it is very difficult to secure a photovoltaic site in Korea, which has a relatively small area.

For this reason, photovoltaic power plants are often built in landfills or mountainous areas, and in the case of landfills or mountainous areas, the ground condition (soft ground) is unstable, leading to accidents with structures and facilities of solar cells and photovoltaic power plants due to long-term subsidence. many.

In addition, accidents occur frequently due to natural disasters and environmental factors such as the rainy season, heavy snow, and strong winds. This causes loss of power generation efficiency and financial damage such as tilting, overturning, and destruction due to structural deformation.

Accidents in such solar power generation facilities may lead to electric shock due to leakage current, fire accidents due to short circuits/electrical short circuits, and may cause enormous human life/property damage.

Moreover, as the frequency of earthquakes has increased rapidly in recent years, it can be seen that Korea is no longer an earthquake-safe zone. When an earthquake occurs, various dangerous situations such as overturning and destruction of facilities, subsidence of the ground, and liquefaction occur, especially in soft ground. In the case of landfills or mountainous areas, which are the ground, there is a high possibility of ground subsidence, liquefaction, and uplift caused by earthquakes, and there was a problem that could lead to a major accident.

Therefore, in order to solve the above problems, the solar power generation system and method to which the sensor-based safety diagnosis technology that can systematically perform safety diagnosis and predict the lifespan of the solar power generation facility structure based on the sensor and neural network algorithm is applied. There is a need for research on it.

Republic of Korea Patent No. 10-1856320 (Registered on May 2, 2018)

It is an object of the present invention to predict the lifespan of a solar power generation facility structure based on a sensor and a neural network algorithm, and obtain inclination information and thermal image information using an inclination sensor and a thermal imaging camera so that safety diagnosis can be systematically performed. And based on the collected information, it performs structural safety diagnosis using a neural network algorithm, predicts the safety state of the structure according to the nystagmus diagnosis result, and continuously calibrates and controls the optimal inclination conditions for safety. To provide a solar power generation system and method to which safety diagnosis technology is applied.

The photovoltaic power generation system to which the sensor-based safety diagnosis technology according to an embodiment of the present invention is applied is voltage, current, insolation, temperature, humidity, atmospheric information, inclination information, and wind speed for determining the photovoltaic power generation state and the power generation structure state. / A plurality of sensors for collecting sensing information including wind direction information; It receives sensing information from the plurality of sensors and provides it to the management server, performs structural safety diagnosis using the inclination information on the inclination angle of the photovoltaic structure, and receives the structure safety diagnosis result of the management server to control the controller a connection panel for controlling the inclination angle of the photovoltaic structure using; Based on the voltage and current collected from the sensor to determine the photovoltaic power generation state, the period, generation amount, accumulated generation amount, power generation amount for each power plant location, and power consumption are calculated and checked through the monitoring screen, and the slope collected through the connection panel It includes a management server that performs a safety diagnosis of the photovoltaic structure using the information.

In the above, comprising a thermal imaging camera for obtaining thermal image information photographing the surface of the photovoltaic structure, wherein the management server receives the thermal image information from the thermal imaging camera, from the temperature difference of the surface of the photovoltaic structure It is characterized in that the structure safety diagnosis result is calculated by grasping the properties and internal conditions of the structure through the thermal state.

In the above, the structural safety diagnosis of the management server is made based on preset diagnosis criteria, and as a result of the safety diagnosis, if the inclination angle of the structure or the thermal deformation state value due to thermal image information is out of the diagnosis criteria range, the diagnosis criteria range It is retransmitted to the controller through the connection panel to correct the inclination angle deviating from , and correcting the structure to have an inclination angle stably supported.

In the above, the management server is a collection unit for receiving the sensing information; Based on a prediction model using a neural network algorithm using the sensing information as an input variable, a prediction model for error or malfunction of the sensor can be generated, and the tolerance range of the sensor is determined using the prediction result of the prediction model, and allowed a prediction unit that determines whether the sensor operates normally within an error range; a safety diagnosis unit for predicting and diagnosing a structural safety state by using the sensing information, slope information, and thermal image information collected from the collection unit as input variables based on a diagnosis model using a neural network algorithm; Sensing information collected from the collecting unit, inclination information, thermal image information, prediction result of the prediction unit, control information of the controller, and data on structure safety diagnosis result are collected, and sensing/control history, prediction/ It includes a statistics learning unit that generates statistical data on the diagnosis result list, events, and management details.

In the above, the management server calculates the prediction accuracy of the prediction result by using the evaluation model generated in advance for prediction verification based on the prediction result of the prediction model, and performs learning to be sensitive to the prediction error according to the prediction accuracy. By allowing the weights or variables of the predictive model to be adjusted, it further includes an evaluation evolution unit that upgrades the predictive model.

The photovoltaic power generation method to which the sensor-based safety diagnosis technology is applied according to an embodiment of the present invention is connected to a communication network with a connection panel that collects sensing information from a plurality of sensors, and a management server that receives sensing information and monitors the photovoltaic power generation status In the photovoltaic power generation method to which a sensor-based safety diagnosis technology is applied using collecting the sensing information including inclination information for ; The connecting panel may include: collecting thermal image information obtained by photographing the surface of a photovoltaic structure from a thermal imaging camera; The connecting group transmits the collected sensing information, inclination information, and thermal image information to the management server: The management server provides a diagnostic model for learning a neural network algorithm using sensing information, inclination information, and thermal image information as input variables. Based on the prediction of the structural safety state, calculating a safety diagnosis result; includes.

In the above, the management server determines whether the sensing error is within the tolerance range by comparing whether the respective sensing values of the collected sensing information operate within the tolerance range as a reference, and determines whether the sensor operates normally. The step of determining; further includes.

In the above, the management server calculates the structure safety diagnosis result based on a preset diagnosis standard, and when the inclination angle of the structure or the thermal deformation state value due to thermal image information using the safety diagnosis result is out of the diagnosis standard range, the corresponding The method further includes; controlling to correct the inclination angle that is out of the diagnostic reference range.

In the above, the management server generates evaluation information using an evaluation model that evaluates the diagnosis model using the structure diagnosis result, and upgrades the diagnosis model by using the evaluation information.

In the above, the management server calculates the prediction accuracy of the diagnosis result by using the evaluation model based on the evaluation information, and performs learning to be sensitive to the prediction error according to the prediction accuracy to adjust the weights or variables of the diagnosis model. It is characterized by upgrading so that

The photovoltaic power generation system to which the sensor-based safety diagnosis technology of the present invention is applied acquires inclination information and thermal image information, and performs structural safety diagnosis using a neural network algorithm based on the collected information. It has the advantage of predicting the condition and continuously calibrating and controlling the optimal inclination conditions for safety.

Provides the manager with integrated statistical data including various sensing information such as solar power generation and temperature and humidity When the set risk standard is exceeded, it has the advantage of real-time monitoring by providing an alarm notification to the administrator.

1 is a block diagram showing the configuration of a photovoltaic power generation system to which a sensor-based safety diagnosis technology is applied according to an embodiment of the present invention.
FIG. 2 is a detailed block diagram showing the internal configuration of a solar power generation system to which the sensor-based safety diagnosis technology of FIG. 1 is applied.
3 is a flowchart of a photovoltaic power generation method to which a sensor-based safety diagnosis technology is applied according to an embodiment of the present invention.
4 is a view showing an example of the configuration of a solar power generation system to which the sensor-based safety diagnosis technology of FIG. 1 is applied.
5 is a view showing an example of collecting sensing information of a solar power generation system to which the sensor-based safety diagnosis technology of FIG. 1 is applied.
FIG. 6 is a view showing an example of a power generation state in a management monitoring screen of a solar power generation system to which the sensor-based safety diagnosis technology of FIG. 1 is applied.
7 is a diagram illustrating an example of a management monitoring setting screen of a solar power generation system to which the sensor-based safety diagnosis technology of FIG. 1 is applied.
FIG. 8 is a view showing an example of displaying the environment and structure status among the management monitoring screens of the solar power generation system to which the sensor-based safety diagnosis technology of FIG. 1 is applied.

Hereinafter, specific embodiments of the present invention will be described in detail with reference to the drawings. However, the spirit of the present invention is not limited to the embodiments presented, and those skilled in the art who understand the spirit of the present invention may add, change, delete, etc. other components within the scope of the same spirit, through addition, change, deletion, etc. Other embodiments included within the scope of the invention may be easily proposed, but this will also be included within the scope of the invention. In addition, components having the same function within the scope of the same idea shown in the drawings of each embodiment will be described using the same reference numerals.

1 is a block diagram showing the configuration of a solar power generation system to which a sensor-based safety diagnosis technology is applied according to an embodiment of the present invention, and FIG. 2 is an internal configuration of a solar power generation system to which the sensor-based safety diagnosis technology of FIG. 1 is applied. It is a block diagram showing in detail.

The photovoltaic power generation system to which the sensor-based safety diagnosis technology of the present invention is applied is connected to the photovoltaic string 200 constituting the photovoltaic panel, and a plurality of sensors 100 and photovoltaic string 200 for collecting various kinds of sensing information. A connection panel 300 that measures and monitors the voltage and current of It includes a management server 500 and a management terminal 600 connected to the connection panel 300 through the communication network 400 .

The plurality of sensors 100 are for collecting sensing information to check various states affecting solar power generation, and specifically, for example, a voltage/current sensor, a solar radiation sensor, a temperature sensor, a humidity sensor, and an atmospheric sensor (oxygen , including sensors that measure CO2, etc.), inclination sensors, wind speed/vane meters, etc. Accordingly, the sensing information collected from the sensor 100 may include voltage, current, insolation, temperature, humidity, atmospheric information (oxygen, CO2), wind speed information, wind direction information, slope information, and the like.

Here, the inclination sensor may be, for example, an acceleration sensor, a gyro sensor, an inertial measurement sensor (IMU sensor), etc., by installing at least one inclination sensor to obtain inclination information measuring the inclination of a structure necessary for diagnosing a photovoltaic structure. It can be provided by the connection panel 300 and the management server 500 .

Here, the photovoltaic structure may be a structure for supporting the photovoltaic string, increase the amount of insolation, and the photovoltaic string is disposed to be inclined at a predetermined angle for optimal photovoltaic efficiency. Since the structure cannot stably support the solar string and the power generation efficiency may also decrease, this is why it is important to pay attention to the safety diagnosis of the structure.

The wind speed / wind vane may be installed in a site around a power plant equipped with a photovoltaic power plant, and a temperature sensor and a humidity sensor are installed inside/outside the connecting panel 300 as well as the solar string 200 to provide the connecting panel 300 ) can also be monitored.

In addition, the thermal imaging camera 310 is photographed for use in structural safety diagnosis by grasping the properties and internal conditions of the structure from the temperature difference on the surface of the solar power generation structure through the thermal state. A thermal change may be detected by photographing a column of the string structure, and the thermal image information may be transmitted to the management server 500 and utilized for safety diagnosis.

The controller 320 collects optimal sunlight by tilting the inclination of the string structure in which the solar string 200 is installed, left and right or up and down, or to control the angle adjustment to a position where the structure can be stably supported. can

In particular, the structure safety diagnosis result by the neural network algorithm of the management server 500 by the controller 320 and the inclination angle of the structure are corrected according to weather conditions, seasons or time zones, thereby stably performing photovoltaic power generation. In addition, the inclination angle corrected according to the weather condition, season or time of day may be an inclination angle suitable for collecting the direct rays of the sun.

The connection panel 300 collects sensing information from a plurality of sensors 100 to monitor the voltage and current of the solar string 200 and check the state, as well as the inclination information of the inclination sensor and the heat of the thermal imaging camera. By collecting image information and providing it to the management server 500, the structure safety diagnosis can be performed, and the structure safety diagnosis result of the management server 500 is provided and transmitted to the controller 320 to control the inclination angle of the structure. to perform solar power generation under optimal conditions.

In addition, it is preferable that the connection panel 300 collects the sensing information collected from the various sensors 100 in a separate line. Referring to FIG. 5 , a sensor board for this may be provided. That is, the insolation sensor, the temperature sensor, the CO2 concentration sensor, and the gyro sensor are connected to the sensor board, respectively, and the corresponding sensing information can be collected.

In addition, the sensor board is provided in the photovoltaic power generation (connection panel 300) provided at each power plant location based on FIG. 4, is connected to a plurality of sensors 100, collects sensing information, and through the RTU, the management server (SERVER) ), sensing information such as voltage, current, temperature, gradient (slope), and the like, and thermal image information can be transmitted. In addition, it is preferable that a switching unit is provided between the photovoltaic power generation (connection panel) and the RTU, so that each sensor 100 to be collected is switched through a separate communication line to transmit a signal, and a monitor signal and a control signal. may also be provided as a separate line.

In addition, the management server 500 calculates the collected voltage and current for each period (day/month/year), the amount of power generation, the accumulated amount of power generation, the amount of power generated by each power plant location, and the power consumption, and it can be checked through the monitoring screen.

In addition, the connection panel 300 may transmit the DC voltage generated from the solar string 200 to supply power to the power system 340 through the inverter 330 .

The management server 500 performs safety diagnosis of the solar power generation structure by using the sensing information collected through the connection panel 300, in particular, the inclination information of the inclination sensor and the thermal image information of the thermal imaging camera in a complex manner, resulting in a safety diagnosis. can be calculated. Here, the thermal imaging camera 310 may be a camera that detects a temperature change in an infrared method and takes pictures.

The structural safety diagnosis of the management server 500 is made based on preset diagnostic criteria, and the safety diagnosis result, which is the result of the safety diagnosis, is retransmitted to the controller 320 through the connection panel 300 to provide an optimal structure that is stably supported. Make it possible to correct it to have an inclination angle.

For example, if the inclination angle of the structure or the thermal deformation state value due to the thermal image information is out of the diagnostic reference range, distortion or deformation of the structure may occur. For this reason, the inclination angle outside the diagnostic reference range may be corrected.

The corrected inclination angle may vary depending on the numerical degree outside the diagnostic reference range, and for this purpose, it is preferable that a numerical value is set in a table form for a plurality of inclination angles to be corrected corresponding to the diagnostic reference range, and correction is made using the inclination angle of the corresponding numerical value. .

In addition, depending on the weather condition, season or time zone, the optimal inclination value is predicted through neural network learning based on the collected inclination information, and the inclination angle is adjusted to improve the solar power generation efficiency by providing the optimal inclination angle suitable for collecting the direct sunlight of the sun. In addition to the inclination information, sensing information on insolation, temperature, humidity, and wind direction/wind speed can also be used as input variables when learning the neural network. In other words, since the structure is affected by the wind, when the wind direction/wind speed exceeds the standard, deformation, distortion, or departure of the structure may occur. .

In addition, when the structure safety diagnosis result of the management server 500 exceeds the set risk standard, the safety diagnosis result is transmitted to the monitoring screen of the management terminal 600 and the connection panel 300 and a warning alarm signal is provided. Therefore, it may be provided so that the status of the structure can be checked in real time in local and external areas.

Furthermore, the structural safety diagnosis using the neural network learning algorithm of the management server 500 described above can also be performed in a control unit (not shown) such as an MCU inside the connection panel 300, and through this, the connection panel 300 is directly connected to the structure. The tilt angle tilting control by the controller 320 may be performed according to the safety diagnosis result.

In addition, as shown in FIG. 2 in detail, the management server 500 includes a collection unit 510, a prediction unit 520, a safety diagnosis unit 530, a statistics learning unit 540, an evaluation evolution unit 550 and It further includes a database 560 .

The collection unit 510 performs communication to collect sensing information, tilt information, and thermal image information through the connection panel 300 and the wired/wireless communication network 400 , and may include one or more communication protocols for this purpose.

The prediction unit 520 compares the sensing information, the inclination information, and the thermal image information collected from the collecting unit 510 whether the corresponding sensing values operate within the tolerance range as a reference, to determine whether the sensing error is within the tolerance range. can be determined, and through this, it can be determined whether the sensor 100 operates normally.

Furthermore, the prediction unit 520 may generate a prediction model for an error or malfunction of the sensor 100 based on a prediction model using at least one neural network algorithm using the sensing information as an input variable, and the prediction result of the prediction model. can be used to determine an allowable error range of the sensor 100 and determine whether the sensor 100 operates normally within the allowable error range. When it is determined that the operation is not normal according to the prediction result of the prediction unit 520, the collection of sensing information may be stopped, and a warning alarm may be transmitted to the management terminal 600, and the type of warning alarm may be changed in stages (eg, normal, caution, abnormal) can be set.

In addition, since the tolerance as a standard does not exist in the initial learning period, the tolerance determined in the manual may be reflected according to the sensor design at the time of shipment of the initial sensor 100 .

The safety diagnosis unit 530 may predict and diagnose the safety state of a structure by using the sensing information, inclination information, and thermal image information collected from the collection unit 510 as input variables based on a diagnosis model using a neural network algorithm. . In addition, it is preferable that the safety diagnosis unit 530 determines the safety state of the structure by adding sensing information about the wind direction/wind speed collected from the wind vane/anemometer in addition to the input variables, because the environment such as rainy season, heavy snow, and strong wind This is because the probability of deformation or distortion of the structure increases due to the negative factors. Through this, it is possible to further perform structural safety diagnosis from multiple angles and to improve the accuracy of diagnosis results.

In addition, the safety diagnosis unit 530 calculates a diagnosis result for the safety state of the structure, retransmits the diagnosis result to the controller 320 through the connection panel 300 to correct the inclination angle of the structure according to the diagnosis result, and if When the risk criterion set in the diagnosis result is exceeded, a warning alarm may be provided to the management terminal 600 and the connection panel 300 so that the state of the structure can be checked in real time in local and external areas. The type of warning alarm for the diagnosis result may be set in stages (eg, structure safety, caution, danger, etc.).

Furthermore, the safety diagnosis unit 530 may predict the current state of the structure using one or more neural network algorithms to calculate a diagnosis result for the safety state of the structure, and an example of a specifically used neural network algorithm is a Recurrent Neural Network (RNN) , a support vector machine (hereinafter referred to as SVM), a convolution neural network (CNN), an attention algorithm, or the like.

Specifically, in the case of SVM, pattern information is generated in the SVM by using the displacement of the acceleration value, and by applying the SVM algorithm, it is possible to recognize various structure situations and determine the state of the structure according to the pattern.

For example, by using the Support Vector Machine (SVM) algorithm, pattern information is generated according to the displacement of the acceleration value obtained from the inclination sensor, and the pattern can be classified using the characteristics and acceleration values of the neural network as parameters. The inclination, deformation, and distortion of structures due to the influence of environmental factors (wind, earthquake, temperature, humidity, etc.) , it is possible to determine the degree of damage to the structure or the degree of external impact, and to calculate a safety diagnosis result.

Here, SVM is one of the supervised learning machine learning methods mainly used for classification, regression, outliers detection, etc. For example, the two groups of datasets are divided Among the various methods to do this, it can be said that the best way to increase the classification accuracy is to be able to accurately classify the midpoint at the maximum distance of each group according to the pattern. In particular, SVM is known as an optimized method for finding an optimal decision boundary capable of distinguishing a plurality of dimensions according to a pattern for data having a plurality of dimensions.

Furthermore, the safety diagnosis unit 530 uses the characteristic data of the sensing information for the inclination information, the thermal image information and the wind direction/wind speed, and in addition to the above-described SVM, CNN, LSTM Long Short Term Memory RNN (RNN), attention algorithm It is configured to perform deep learning through a combination of

Here, the attention algorithm consists of an encoder and a decoder, and the encoder makes the hidden states of the LSTM into a matrix, and the decoder can highlight important states among the hidden states. This method makes it possible to focus on important output vectors over a certain distance, It can show great performance in neural machine translation (NMT).

Therefore, the safety diagnosis unit 530 may use a model (CLA model) in which CNN, LSTM RNN, and attention algorithm are combined as a diagnosis model for calculating the safety diagnosis result.

The CLA model uses the advantages of CNN, which is advantageous for local feature identification, LSTM RNN, which is advantageous for sequential data processing, and the attention mechanism, which allows you to focus on important information, to induce better performance compared to models using simple SVM or CNN. be able to

For example, the CLA model is converted into an 80-dimensional mel spectrogram during the feature extraction process of inclination information and thermal image information, and 300 framesets of 3 seconds are input to the model. The model extracts the spatial characteristics of the spectrum with the CNN layer, the temporal information with the LSTM layer, and then collects the information by concentrating the local characteristics with the attention layer.

The CNN layer is a layer that repeats convolution operation and max pooling twice.

For example, the CNN layer can be used to analyze the input spectrogram and work to extract features more suitable for structural safety diagnosis. This is done through convolution operation and max pooling, and the convolution operation is a process of extracting necessary features through operation from a local part of the input data. In addition, the Max pooling operation reduces the dimension of the data while leaving only the necessary features, and plays a role in lowering the computational complexity of the LSTM and attention layer that will be performed later.

In the LSTM layer, LSTM is a type of Recurrent Neural Network (RNN) architecture, and in general, when referring to RNN, it is a structure that is used so much that it consists of an LSTM structure.

In the CLA model, it is used to obtain temporal information from the feature vectors obtained through the CNN layer. This is to determine an abnormal state exceeding the set risk level during safety diagnosis of a photovoltaic structure that can be known only by processing long-distance information using the sequence information processing capability of the RNN.

In addition, the attention layer is a layer made by utilizing the attention mechanism and is used to create a context vector for making a final judgment from the information vector obtained through CNN and LSTM.

The input of the attention mechanism is the output of the LSTM layer H=h ₁ , h ₂ , h ₃ , … , hT, hi=(p ₁ , p ₂ , …, p _C ) and the weight parameter q=(q ₁ , q ₂ , q ₃ , …, q _C ) to calculate the desired output (safety diagnosis result). can As a result of such a safety diagnosis, it can be used to determine whether the solar power structure is in a stable state, the remaining life span, or the maintenance and reinforcement period, and to directly control the inclination angle according to the stability of the structure.

The statistical learning unit 540 receives the sensing information collected from the collecting unit 510, the inclination information, the thermal image information, the prediction result of the prediction unit 520, the control information of the controller 320, and data on the structure safety diagnosis result. As the amount of collected data increases, it is possible to use big data to generate comprehensive statistical data on sensing/control history, prediction/diagnosis result list, event, and management history.

In particular, the statistics learning unit 540 may systematically show information collected through a monitoring program to utilize statistical data as monitoring data, and may provide, for example, a screen as shown in FIGS. 6 to 8 .

6 is a program screen showing the current status of photovoltaic power generation, and daily/monthly power generation, cumulative power generation, solar power generation power, carbon reduction, power consumption, etc. are displayed, and an administrator's setting interface for systematic management and setting when collecting data. can provide

7 exemplarily shows a screen when setting the administrator interface, through which communication serial port setting, server address, port assignment, inverter 330 setting, sensor setting, solar string 200 setting, device (solar light) Power plant devices) ID confirmation and selection, device list, power plant installation location, power plant ID, power plant capacity, etc. may be set with administrator authority.

In addition, FIG. 8 shows that insolation information for photovoltaic power generation efficiency, temperature information, inclination information for a slope that can determine the structure state, and CO2 concentration are displayed, and additionally, image information of the thermal imaging camera 310 is provided. could be

Furthermore, the statistical learning unit 540 performs learning using a separate prediction model generated in advance based on the statistical data collected in the form of big data, and predicts the state and structure of the connection panel 300 in the future predicted as a result of the learning. It is possible to calculate the prediction result for the future prediction state for maintenance management according to the life state, the state of the connected sensor 100 and the controller 320, and the accuracy of the prediction result for learning will improve as the amount of data increases. can

In addition, by providing the data on the prediction result to the management terminal 600 possessed by the manager, it is possible to grasp the lifespan of the structure according to the future predicted state, and use it for maintenance management.

The evaluation evolution unit 550 calculates the prediction accuracy of the prediction result and the diagnosis result by using the evaluation model generated in advance for prediction verification based on the prediction result of the above-described prediction model and the structural safety diagnosis result of the diagnosis model, and the prediction accuracy By performing training to be sensitive to the prediction error according to It can adapt and ultimately achieve self-evolution that allows the weights or variables of the prediction model to be adjusted and corrected by itself even if there is no prediction result.

Specifically, when the prediction values of the prediction model and the diagnostic model show a certain pattern, the model goes through an adaptation step in which the upgrade is continuously performed by adjusting the variables of the corresponding model according to the pattern, and the corresponding models that have undergone the adaptation step are artificial intelligence algorithms. Ultimately, self-evolution is possible.

Here, the predictive model, the diagnostic model, and the evaluation model may be a predictive model and an evaluation model based on an artificial neural network, a radial basis function (RBF) neural network, an SVM algorithm, and the like.

In addition, since sensing information, prediction results, diagnosis results, and evaluation information of predictive models and diagnostic models become big data, and big data is periodically updated, it is necessary to evaluate these models as well as predictive models and diagnostic models using big data. By periodically upgrading the evaluation model for big data to be able to adapt according to the changed pattern of big data, it is possible to ensure the reliability of the model.

That is, if the predicted values show a certain pattern, the model undergoes an adaptation step in which the upgrade is continuously performed by adjusting the variables of the corresponding model according to the pattern. In addition, the models that have undergone the adaptation stage are ultimately able to self-evolution by the artificial intelligence algorithm.

Therefore, the evaluation evolution unit 550 receives the re-evaluation information re-evaluated for the evaluation model that has performed the evaluation of the predictive model and the diagnostic model, and learns and adapts the problems of the evaluation model with the received re-evaluation information based on the neural network learning algorithm, The above-described evaluation model may be upgraded.

The database 560 classifies sensing information, inclination information, thermal image information, sensing error prediction result, structure safety diagnosis result, etc., and is systematically stored and managed through a plurality of databases 560 subdivided by category, and provided upon request. can be

3 is a flowchart of a photovoltaic power generation method to which a sensor-based safety diagnosis technology is applied according to an embodiment of the present invention.

Photovoltaic applied with sensor-based safety diagnosis technology using a management server 500 that is connected to a connection panel 300 that collects sensing information from a plurality of sensors 100 and a communication network to receive sensing information and monitor the state of photovoltaic power generation In the power generation method, the connection panel 300 receives the sensing information including voltage, current, insolation, temperature, humidity, atmospheric information (oxygen, CO2), wind speed information, wind direction information, and inclination information for the inclination angle of the solar power generation structure. Collect (S401).

The connection panel 300 collects thermal image information obtained by photographing the surface of the photovoltaic power generation structure from the thermal imaging camera 310 (S402).

In addition, the connection panel 300 transmits the collected sensing information, inclination information, and thermal image information to the management server 500 (S403).

Thereafter, the management server 500 determines whether the sensing error is within the tolerance range by comparing whether the corresponding sensing values of each of the collected sensing information operate within the tolerance range as a reference, and whether the sensor 100 operates normally. It is determined whether there is (S404).

In addition, the management server 500 may calculate the safety diagnosis result by predicting the safety state of the structure based on the diagnosis model for learning the neural network algorithm using sensing information, inclination information, and thermal image information as input variables (S405) .

In addition, the management server 500 calculates a structure safety diagnosis result based on a preset diagnosis standard, and using the safety diagnosis result, when the inclination angle of the structure or the thermal deformation state value due to thermal image information is out of the diagnosis criterion range, the corresponding diagnosis It is possible to control to correct the inclination angle out of the reference range (S406).

Also, the management server 500 may generate evaluation information using an evaluation model that evaluates a diagnosis model using the structure diagnosis result, and upgrade the diagnosis model using the evaluation information (S407).

As used herein, a 'terminal' may be a wireless communication device with guaranteed portability and mobility, for example, any type of handheld-based wireless communication device such as a smartphone, tablet PC, or notebook computer. have. In addition, the 'terminal' may be a wired communication device such as a PC that can be connected to another terminal or server through a communication network.

In addition, the wired/wireless communication network 400 means a connection structure capable of exchanging information between each node, such as terminals and servers, and includes a local area network (LAN), a wide area network (WAN). , the Internet (WWW: World Wide Web), wired and wireless data networks, telephone networks, wired and wireless television networks, and the like.

Examples of wireless data communication networks include 3G, 4G, 5G, 3rd Generation Partnership Project (3GPP), Long Term Evolution (LTE), World Interoperability for Microwave Access (WIMAX), Wi-Fi, Bluetooth communication, infrared communication, ultrasound Communication, Visible Light Communication (VLC), LiFi, etc. are included, but are not limited thereto.

100 ; sensor
200 ; solar string
300 ; junction panel
310; Thermal imaging camera
320 ; controller
330; inverter
340; power system
400 ; wired and wireless communication network
500 ; management server
510; collector
520 ; predictor
530; Safety Diagnosis Department
540 ; Department of Statistics
550 ; Evaluation and Evolution Department
560; database
600 ; management terminal

Claims (10)

a plurality of sensors for collecting sensing information including voltage, current, insolation, temperature, humidity, atmospheric information, inclination information, and wind speed/direction information of the solar string for determining the state of photovoltaic power generation and the state of the power generation structure;
a thermal imaging camera that acquires thermal image information photographing the surface of a photovoltaic power generation structure;
It receives sensing information from the plurality of sensors and provides it to the management server, performs structural safety diagnosis using the inclination information on the inclination angle of the photovoltaic structure, and receives the structure safety diagnosis result from the management server and uses the controller a connection panel to control the inclination angle of the photovoltaic structure;
Based on the voltage and current collected from the sensor to determine the photovoltaic power generation state, the generation amount by period, cumulative generation amount, generation amount by power plant location, and power consumption are calculated and checked through the monitoring screen, and the slope collected through the connection panel Safety diagnosis of photovoltaic structures is performed using the information, and thermal image information is transmitted from the thermal imaging camera, and the properties and internal conditions of structures are identified through thermal conditions from the temperature difference on the surface of the photovoltaic structure to ensure structural safety. Includes a management server that calculates diagnostic results,
The management server
a collection unit receiving the sensing information and the thermal image information;
Based on a prediction model using a neural network algorithm using the sensing information as an input variable, a prediction model for error or malfunction of the sensor can be generated, and the tolerance range of the sensor is determined using the prediction result of the prediction model, and allowed a prediction unit that determines whether the sensor operates normally within an error range;
a safety diagnosis unit for predicting and diagnosing a structural safety state by using the sensing information, slope information, and thermal image information collected from the collection unit as input variables based on a diagnosis model using a neural network algorithm;
Sensing information collected from the collecting unit, inclination information, thermal image information, prediction result of the prediction unit, control information of the controller, and data on structure safety diagnosis result are collected, and sensing/control history, prediction/ It further includes a statistics learning unit that generates statistical data on the diagnosis result list, event, and management history,
The safety diagnosis unit
To predict the current state of the structure using one or more neural network algorithms to produce a diagnostic result for the safety state of the structure,
Using the Support Vector Machine (SVM) algorithm, pattern information is generated according to the displacement of the acceleration value obtained from the inclination sensor, and the pattern can be classified using the characteristics and acceleration values of the neural network as parameters. The inclination, deformation, distortion of the structure due to the influence of wind, earthquake, temperature, and humidity, which are environmental factors that affect the photovoltaic structure for a situation where the pattern of the pattern information generated based on the generated pattern is similar It is possible to calculate the safety diagnosis result by judging the degree of damage or external impact.
Using the characteristic data of the sensing information for the slope information, thermal image information, and wind direction/wind speed, structure safety diagnosis is made through the CLA model that combines CNN, LSTM RNN (Long Short Term Memory RNN), and Attention mechanism. to perform deep learning for
The CLA model extracts the spatial characteristics of the spectrum with the CNN layer during the feature extraction process of the gradient information and the thermal image information, and extracts the temporal information with the LSTM layer. collect information by focusing on characteristics,
The safety diagnosis result value, which is the desired output, can be calculated using the output of the LSTM layer of the attention mechanism and the weight parameter. A solar power generation system with sensor-based safety diagnosis technology that can be used to control the angle of inclination according to the stable state. delete According to claim 1,
The structural safety diagnosis of the management server is made based on preset diagnostic criteria. A photovoltaic power generation system to which sensor-based safety diagnosis technology is applied, characterized in that it is retransmitted to the controller through the connection panel to correct and corrects the structure to have an inclination angle at which it is stably supported. delete According to claim 1,
The management server
Based on the prediction results of the prediction model, the prediction accuracy of the prediction results is calculated using the evaluation model created in advance for prediction verification, and the weights or variables of the prediction model are adjusted by learning to be sensitive to the prediction error according to the prediction accuracy. By making it possible, the evaluation evolution unit that performs the upgrade of the predictive model
A solar power generation system with sensor-based safety diagnosis technology that further includes. In the photovoltaic power generation method to which a sensor-based safety diagnosis technology is applied using a management server that is connected to a connection panel that collects sensing information from a plurality of sensors and a communication network and receives sensing information to monitor the state of photovoltaic power generation,
The connection panel collects the sensing information including voltage, current, insolation, temperature, humidity, atmospheric information (oxygen, CO2), wind speed information, wind direction information, and inclination information on the inclination angle of the photovoltaic structure;
The connecting panel may include: collecting thermal image information obtained by photographing the surface of a photovoltaic structure from a thermal imaging camera;
Transmitting, by the connection panel, the collected sensing information, inclination information, and thermal image information to the management server:
The management server determines whether the sensing error is within the tolerance range by comparing whether the respective sensing values of the collected sensing information operate within the tolerance range as a reference, and determining whether the sensor operates normally;
The management server is based on a diagnostic model for learning a neural network algorithm using sensing information, inclination information, and thermal image information as input variables, predicts the safety state of the structure and calculates a safety diagnosis result from the temperature difference on the surface of the photovoltaic structure. Comprising the step of calculating the structural safety diagnosis result by grasping the properties and internal conditions of the structure through the thermal state,
The management server
a collecting unit receiving the sensing information and the thermal image information;
A prediction model for error or malfunction of a sensor can be generated based on a prediction model using a neural network algorithm using the sensing information as an input variable, and the tolerance range of the sensor is determined using the prediction result of the prediction model, a prediction unit that determines whether the sensor operates normally within an allowable error range;
a safety diagnosis unit for predicting and diagnosing a structural safety state using the sensing information, inclination information, and thermal image information collected from the collection unit as input variables based on the diagnosis model using the neural network algorithm;
Sensing information collected from the collecting unit, inclination information, thermal image information, prediction result of the prediction unit, control information of the controller, and data on structure safety diagnosis result are collected, and sensing/control history, prediction/ It further includes a statistics learning unit that generates statistical data on the diagnosis result list, event, and management history,
The safety diagnosis unit
To predict the current state of the structure using one or more neural network algorithms to produce a diagnostic result for the safety state of the structure,
Using the Support Vector Machine (SVM) algorithm, pattern information is generated according to the displacement of the acceleration value obtained from the inclination sensor, and the pattern can be classified using the characteristics and acceleration values of the neural network as parameters. The inclination, deformation, distortion of the structure due to the influence of wind, earthquake, temperature, and humidity, which are environmental factors that affect the photovoltaic structure for a situation where the pattern of the pattern information generated based on the generated pattern is similar It is possible to calculate the safety diagnosis result by judging the degree of damage or external impact.
Using the characteristic data of the sensing information for the slope information, thermal image information, and wind direction/wind speed, structure safety diagnosis is made through the CLA model that combines CNN, LSTM RNN (Long Short Term Memory RNN), and Attention mechanism. to perform deep learning for
The CLA model extracts the spatial characteristics of the spectrum with the CNN layer during the feature extraction process of the gradient information and the thermal image information, and extracts the temporal information with the LSTM layer. collect information by focusing on characteristics,
The safety diagnosis result value, which is the desired output, can be calculated using the output of the LSTM layer of the attention mechanism and the weight parameter. Photovoltaic power generation method with sensor-based safety diagnosis technology, characterized in that it can be used to control the inclination angle according to the stable state. delete 7. The method of claim 6,
The management server calculates the structure safety diagnosis result based on a preset diagnosis standard, and when the inclination angle of the structure or the thermal deformation state value due to thermal image information is out of the diagnosis criterion range using the safety diagnosis result, the corresponding diagnosis criterion range is determined. controlling the deviated inclination angle to be corrected;
A photovoltaic power generation method using sensor-based safety diagnosis technology that further includes. 7. The method of claim 6,
generating, by the management server, evaluation information using an evaluation model that evaluates a diagnosis model using a structure diagnosis result, and upgrading the diagnosis model using the evaluation information;
A photovoltaic power generation method using sensor-based safety diagnosis technology that further includes. 10. The method of claim 9,
The management server
A sensor, characterized in that it is upgraded to adjust the weights or variables of the diagnostic model by calculating the prediction accuracy of the diagnosis result by using the evaluation model based on the evaluation information, and performing learning to be sensitive to the prediction error according to the prediction accuracy Photovoltaic power generation method applied with based safety diagnosis technology.

Images