KR20220128737A - Method and system for operating photovoltaic inverter using multi-layer neural network fault diagnosis model - Google Patents

Abstract

According to the present invention, a solar inverter drive system using a multilayer neural network failure diagnosis model comprises: a solar panel converting solar energy into electrical energy to output electric power; a master module and a slave module each including an inverter module, converting the power output from the solar panel into alternating current, and transmitting the power to a power system; and a monitoring system collecting monitoring data of the master module and slave module and determining failure states of the master module and slave module using a predefined multilayer neural network failure diagnosis model. Here, the solar inverter drive system is configured so that the master module preferentially transmits the power to the power system. When the master module is determined to be in the failure state by the monitoring system, the master module is shut off and the slave module operates.

Description

Solar inverter driving method and system using multi-layer neural network failure diagnosis model

The present invention relates to a fault diagnosis technology of a solar inverter that can be applied to a solar power generation system and an inverter driving technology using the same.

In the photovoltaic power generation system, the inverter is a key component that converts the direct current generated by the solar cell module into alternating current and connects it to the power system. is a device that The lifespan of a solar inverter is about 10 years, and in the event of a failure, it directly affects the entire solar power generation, causing great economic loss, as well as fire accidents and short circuit accidents, causing great problems in the safety of managers. Therefore, it is essential to manage normal continuous operation abnormalities and failures.

Recently, the trend of applying hundreds to thousands of inverters in photovoltaic power generation complexes is that if an inverter fails in such a large-capacity photovoltaic complex, photovoltaic modules, junction boxes, meters, various circuit breakers, and electrical parts are complicated. Since they are connected, it is very difficult to directly determine whether the inverter has failed.

In this regard, in order to diagnose a failure by monitoring a plurality of inverters, a method in which an administrator checks whether there is an abnormality through a monitoring system supplied by a manufacturer is generally used. Here, the monitoring system monitors the amount and efficiency of the solar power plant, and cannot provide detailed information about the failure. have to figure out Such an operator-centered on-site inspection method has limitations in terms of diagnosis and management of inverters.

Currently, research on fault diagnosis of photovoltaic systems is mostly focused on diagnosing whether a photovoltaic array module is abnormal, a short or open connection line, and the like, and research on the fault and diagnosis of an inverter is very scarce. This is because there is a limit in accurately diagnosing whether an inverter is faulty because the solar cell module and various electrical components are intricately connected.

Currently, there is a need to develop a technique for solving the limitations and problems of the prior art.

Korean Patent Publication No. 10-2017-0120954

The present invention was derived to solve the above-described problems, and more accurately and quickly diagnose whether the solar inverter module has a failure, and the solar power generation system can be stably driven even when a failure occurs in the solar inverter module. To provide a solar inverter driving method and system that supports

In accordance with one aspect of the present invention, there is provided a solar inverter driving system using a multilayer neural network failure diagnosis model, comprising: a solar panel for converting solar energy into electrical energy and outputting power; a master module and a slave module each including an inverter module, converting the power output from the solar panel into AC and transmitting it to the power system; and a monitoring system that collects monitoring data for the master module and the slave module, and determines a failure state for the master module and the slave module using a predefined multi-layer neural network failure diagnosis model. Here, the solar inverter driving system is configured such that the master module preferentially transmits the power to the power system, and when the master module is determined to be in a fault state by the monitoring system, the master module is cut off and The slave module is configured to operate.

In one embodiment, the solar inverter driving system, when the master module is determined to be in a fault state by the monitoring system, cuts off the master module from the solar panel, and the slave module and the solar panel It may further include a power conversion switching device that is connected to switch to operate.

In one embodiment, the monitoring system includes a storage device that collects and stores monitoring data including inverter generation amount and watt-hour meter generation amount, and a failure state for the master module and slave module using the multi-layer neural network failure diagnosis model. A fault diagnosis server may be included.

In one embodiment, the multi-layer neural network failure diagnosis model is defined as an input in which the converted power generation of a predefined value, the watt-hour meter check power generation amount, the inverter check power generation amount and error (watt-hour meter check power generation amount-inverter check power generation amount) are defined as inputs, and the normal state, inverter failure State and watt-hour meter failure state are defined as outputs, and it can correspond to a multi-layer perceptron neural network model that has been trained using pre-collected basic data.

A solar inverter driving method using a multilayer neural network failure diagnosis model according to another aspect of the present invention is performed in a solar inverter driving system. Here, the solar inverter driving system may include: a solar panel that converts solar energy into electrical energy and outputs power; a master module and a slave module each including an inverter module, converting the power output from the solar panel into AC and transmitting it to the power system; and a monitoring system that collects monitoring data for the master module and the slave module, and determines a failure state for the master module and the slave module using a predefined multi-layer neural network failure diagnosis model.

Here, the method for driving a solar inverter includes: collecting, by the monitoring system, monitoring data including an inverter power generation amount and an watt-hour meter power generation amount as data associated with the master module; determining, by the monitoring system, a failure state for the master module using the multi-layer neural network failure diagnosis model; and when the master module is determined to be in a failure state, the master module is cut off from the solar panel through a power conversion switching device, and the slave module is connected to the solar panel to operate.

In one embodiment, the multi-layer neural network failure diagnosis model is defined as an input in which the converted power generation of a predefined value, the watt-hour meter check power generation amount, the inverter check power generation amount and error (watt-hour meter check power generation amount-inverter check power generation amount) are defined as inputs, and the normal state, inverter failure State and watt-hour meter failure state are defined as outputs, and it can correspond to a multi-layer perceptron neural network model that has been trained using pre-collected basic data.

According to the present invention, the failure of the solar inverter module can be diagnosed more accurately and quickly using the multi-layer neural network failure diagnosis model, and a failure occurs in any one inverter module by applying the master-slave driving method inverter module. However, there is an advantage that it can be operated stably.

1 to 14 are reference views for explaining the configuration and operation of a solar inverter driving system according to the present invention.
2 is a block diagram for explaining the configuration of a solar inverter driving system according to the present invention.

Since the present invention can apply various transformations and can have various embodiments, specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. In describing the present invention, if it is determined that a detailed description of a related known technology may obscure the gist of the present invention, the detailed description thereof will be omitted.

Terms such as first, second, etc. may be used to describe various elements, but the elements should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another.

The terms used in the present application are only used to describe specific embodiments, and are not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly dictates otherwise. In the present application, terms such as “comprise” or “have” are intended to designate that a feature, number, step, operation, component, part, or combination thereof described in the specification exists, but one or more other features It is to be understood that this does not preclude the possibility of the presence or addition of numbers, steps, operations, components, parts, or combinations thereof.

1. Introduction

The present invention has been devised to solve the above-described limitations and problems of the prior art, and proposes an inverter fault diagnosis technology that can more accurately and quickly diagnose whether an inverter is faulty using an artificial intelligence algorithm, and also We propose an inverter drive system that can stably drive an inverter using

More specifically, the present invention proposes a multi-layer neural network failure diagnosis model capable of diagnosing failures using inverter failure data of a solar power plant. The most important factor in designing a multi-layer neural network model for fault diagnosis of inverters is learning data. Here, in the present invention, data for one year (April 2018 to March 2019) acquired from 125 photovoltaic power plants currently in operation was used as learning data, and the acquired data is the power plant capacity, power generation in the previous month, and the previous month. It contains more than 1,600 items such as meter reading amount, previous month meter reading amount, meter displacement, current month inverter meter reading amount, previous month inverter meter reading amount, data in case of inverter failure, and watt-hour meter reading value.

In addition, the present invention proposes an inverter driving system of a master-slave driving method for realizing a stable operation of the inverter system. In the inverter driving system according to the present invention, the above-described multi-layer neural network failure diagnosis model may be applied.

Hereinafter, an inverter fault diagnosis technology and an inverter driving system according to the present invention will be described in more detail with reference to the drawings.

2. Inverter Failure Types and Conventional Diagnosis Methods

Referring to FIG. 1, the solar power generation system includes a solar cell module that produces electricity, a junction box that connects various modules in series and parallel, an inverter that converts DC power made in solar cells into AC power, and measures and manages the amount of power. It consists of a power meter. In addition, the photovoltaic power generation system further includes a linkage facility for operating in conjunction with commercial power, a lightning protection and grounding facility to protect the photovoltaic power plant from lightning and external abnormal voltage, and a storage battery facility that can be installed as needed. can be configured.

An inverter configured in a photovoltaic power generation system is a device that converts power, has a very complex configuration, and frequently breaks down. Here, the failure types related to the inverter of the solar power generation system are as follows.

go. overvoltage, overcurrent

Due to the initial start of the inverter or sudden change of the MPPT, overvoltage and overcurrent occur depending on characteristics such as control. If a lot of overvoltage or overcurrent warnings appear on the inverter, it puts a lot of strain on the inverter, so the lifespan will be drastically reduced and damage will cause the operation to stop.

me. overheating

The inverter delivers power through the repetitive On/Off operation of several to tens [KHz] of semiconductor devices. Here, very high heat is generated in the semiconductor device, and this heat must be well discharged. At this time, if dust accumulates in the ventilation hole or foreign matter accumulates on the heat sink, and natural convection is not good, a malfunction due to heat may occur.

All. overload

When the inverter load is greater than the rated output, an overload occurs when the inverter output is short-circuited or a load greater than the power supply capacity is connected.

la. Other inverter internal faults

As internal parts of the inverter, the parts that may fail include a cooling fan, a condenser, a switching element, and a filter.

Currently, various techniques for diagnosing inverter failures as described above are being used. Maximum power point tracking method (MPPT) using solar cell I-V characteristic curve, TDR (Time Domain Reflectometry) measurement method, SSTDR (Spread Spectrum Time Domain Reflectometry) measurement method, infrared thermal imaging method, monitoring and diagnosis through CCTV and monitoring method and the like. However, these existing diagnostic methods are a method of diagnosing faults directly in the field using measuring equipment, and the fault detection performance at a long distance is low or impossible, and it is necessary to open a wire to measure the fault point. And there are limitations in terms of speed.

The present invention proposes an inverter fault diagnosis technique and an inverter driving system described below as technical means for overcoming the limitations of the conventional diagnosis technique.

3. Multilayer Neural Network Inverter Failure Diagnosis Model

The multi-layer neural network inverter fault diagnosis model (hereinafter, multi-layer neural network model) according to the present invention may be implemented as a multi-layer perceptron neural network. As shown in FIG. 2 , the multilayer neural network model according to the present invention may be defined as four inputs (x1 to x4), three hidden layers, and three outputs y (y1 to y3). Here, as shown in Table 1 below, in the input variable x, x1 can be defined as 99[kW] converted power generation, x2 is inverter check power generation, x3 is watt-hour meter check power generation amount, and x4 is error, and there are 22 neurons in the hidden layer. It may be composed of three layers, and in the output y, y1 may be defined as a normal state, y2 may be defined as an inverter failure state, and y3 may be defined as a watt-hour meter failure state. In addition, the multilayer neural network model according to the present invention may be implemented as shown in Table 2 below.

[Table 1]

[Table 2]

On the other hand, the details of the above-described multilayer neural network inverter failure diagnosis model are for more clearly explaining the technical idea according to the present invention, and are not intended to limit the scope of the present invention, and may be defined differently if necessary. of course there is

Hereinafter, an inverter fault diagnosis method using a multilayer neural network model according to the present invention will be described.

4. Inverter fault diagnosis method using multi-layer neural network model

Inverter fault diagnosis method according to the present invention, as shown in FIG. 3, a data collection step of collecting data related to a solar inverter, a data mining step of arranging items necessary for fault diagnosis from the collected data, It can proceed to the learning step of learning the multilayer neural network inverter fault diagnosis model, which is a predefined deep learning model using data, and the diagnosis and prediction step of diagnosing the fault of the inverter using the learned multilayer neural network inverter fault diagnosis model.

As basic data for the multilayer neural network model according to the present invention, actual data of 125 solar power plants currently in operation were collected and utilized. Meanwhile, since the power generation capacity of the solar power plant varied from 30kW to 500kW, the data converted to 99kW of the power generation capacity was used for the failure diagnosis prediction simulation. There are a total of 1600 data used for failure diagnosis prediction, from April 2018 to March 2019, and it is data obtained from the field for one year. Here, the basic data (source data obtained on site) that can be applied to the multilayer neural network model according to the present invention is current month meter reading amount, previous month meter reading amount, inverter meter reading amount, previous month power generation {(previous month meter reading - current month power generation) * meter multiple} , watt-hour meter incremental generation {(current month meter reading * meter multiplier}-(previous month meter reading * meter multiplier)}, inverter inspection power generation (inverter meter reading amount - previous month inverter meter reading amount), error amount (watt-hour meter check generation - inverter check generation), 99 per month [kW] Conversion power {(99 * previous month's)/capacity}, capacity, meter drainage, meter failure (check the safety manager), inverter failure (check the safety manager), and the name of the power plant can be included.

After data collection, data mining for application to inverter fault detection is performed to derive data necessary for inverter fault diagnosis. Here, the items required for data mining are capacity (power plant capacity), power generation in the previous month {(meter reading in the previous month - meter reading in the previous month) * meter multiple}, meter reading in the previous month, meter reading in the previous month, meter multiple, inverter meter reading in the current month, inverter in the previous month It can include meter reading amount, inverter meter reading amount, previous month inverter meter reading amount, and whether there is a failure (inverter, watt-hour meter).

As data required for inverter fault diagnosis derived through data mining, the items necessary for data learning and prediction are 99[kW] converted power generation, inverter check power generation, watt-hour meter check power generation, error, and whether there is a fault condition (normal, inverter failure, power meter failure) ) may be included.

x1 : 99[kW] converted power = (99 * power generation)/capacity

x2 : Electricity meter monthly power generation = (inverter meter reading in current month * meter multiple) - (inverter meter reading in previous month * meter multiple)

x3: Inverter monthly power generation (inverter meter reading for the current month - inverter meter reading for the previous month)

x4: error (x2-x3) (watt-hour meter check power generation - inverter check power generation)

y1 : Normal

y2 : Inverter fault status (INV fault)

y3: watt-hour meter fault condition (MET fault)

On the other hand, the data items for learning and diagnosing the multi-layer neural network inverter failure diagnosis model described above are for more clearly explaining the technical idea according to the present invention, and are not intended to limit the scope of the present invention. Of course, it may be defined differently.

In order to select an activation function for multilayer neural network diagnosis, simulations for accuracy prediction were performed for Sigmoid-Softmax, TanH-Softmax, and ReLU-Softmax among the three types of activation functions that are currently most used. In FIG. 4, (A) is a simulation result of Sigmoid-Softmax, (B) is TanH-Softmax, (C) is a simulation result of ReLU-Softmax, and Table 3 below shows the results for accuracy and loss for each function.

[Table 3]

Referring to FIG. 4 and Table 3, when ReLU as input and Softmax as output are applied, accuracy of 97% and loss of 0.17% are shown, which is the most suitable.

Next, in order to select an optimization function for diagnosing a multilayer neural network, simulations for accuracy prediction were performed for SGD, Adagrad, and Adam among the three types of optimization functions that are currently most used. In FIG. 5, (A) is SGD, (B) is Adagrad, (C) is Adam's simulation results, and Table 4 below shows the results for accuracy and loss for each function.

[Table 4]

Referring to FIG. 5 and Table 4, when ReLU and Softmax are applied to the input and output and Adam is applied to the optimization function, 97% accuracy and 0.11% loss are shown, which is the most suitable.

Next, in order to select a batch size for diagnosing a multilayer neural network, the batch size was divided into four, and a simulation was performed to predict the accuracy according to the batch size size and training time. In FIG. 6, (A) is Batch size 5, (B) is Batch size 30, (C) is Batch size 100, (D) is a simulation result of Batch size 500, and Table 5 below shows batch size size and time The accuracy analysis result is shown.

[Table 5]

6 and Table 5, the training accuracy was the highest at 96.7% when the batch size was 30, and the training speed was the fastest when the batch size was 500. That is, as the batch size increases, the training speed increases, while the accuracy decreases.

Next, in order to select the number of epochs for diagnosing the multilayer neural network, simulations applying various epochs were performed to analyze accuracy and time. 7, (A) shows simulation results of Epoch 1000, (B) Epoch 2000, and (C) Epoch 3000, and Table 6 below shows the accuracy and time analysis results for each number of epochs.

[Table 6]

Referring to FIG. 7 and Table 6, it showed the highest accuracy of 96.7% when trained 1000 times. Although similar accuracy was shown in 2000 and 3000 training, it can be confirmed that overfitting occurs when training more than 1000 times.

After learning the multilayer neural network model according to the present invention, a prediction experiment was conducted to diagnose the failure of the solar inverter and the watt-hour meter. To simulate diagnostic prediction, Tensor flow-based Python provided by Anaconda was used. 1600 data sets were trained using the multilayer neural network model according to the present invention. The configuration of the training data (Table 7), the parameters of the multilayer neural network model (Table 8), and the summary of the multilayer neural network model (Table 9) are as follows.

[Table 7]

[Table 8]

[Table 9]

Hereinafter, simulation results of the multilayer neural network model according to the present invention will be described.

8 and 9 show the accuracy and loss rate of the failure prediction rate of the inverter, and FIGS. 10 and 11 show the failure prediction results for the normal state and the inverter failure state, respectively.

The loss rate and accuracy, which are the most important indicators of the multilayer neural network model according to the present invention, showed an accuracy of 97% and a loss rate of 0.11%, respectively.

[Table 10]

[Table 11]

Referring to Table 10 (the result of judging the field data that the multilayer neural network model is in a normal state) and Table 11 (the result of determining the field data that the multilayer neural network model is in the inverter failure state) below, the multilayer neural network model when the inverter is in a failure state can be predicted with an accuracy of 97%. On the other hand, the fault state means that the inverter does not operate normally due to incomplete operation, a sharp decrease in efficiency, or malfunction of internal and external components of the inverter as well as when the inverter completely stops operating. When the multilayer neural network model according to the present invention determines that the inverter operation is abnormal, the 99 [kW] power generation amount, the indicated value displayed on the inverter, the indicated value of the watt-hour meter, and the error values cause abnormal changes, which It can be confirmed that the change amount is larger than that during normal operation.

5. Master-Slave drive type inverter drive system

Hereinafter, as an inverter driving system according to the present invention, an inverter driving system of a Master-Slave driving method to which the above-described multilayer neural network model is applied will be described. The inverter driving system according to the present invention can stably operate the inverter module used in the field.

Referring to FIG. 12 , the inverter driving system according to the present invention may include a solar panel, a master module and a slave module connected thereto, a monitoring system, and a power conversion switching device. Here, the master module is connected to the solar panel and operates preferentially, and the slave module is configured to operate auxiliary when the master module is determined to be in a fault state. Here, the master module and the slave module may be configured to include a junction box module and an inverter module, and of course, other configurations (such as a watt-hour meter) not shown in FIG. 12 may be further included.

In the inverter driving system according to the present invention, the monitoring system continuously monitors the amount of power produced by the inverter of the master module, the data generated in the surrounding meters, and the connection box, and uses a multilayer neural network model to control the components such as the inverter and the watt-hour meter of the master module. Predict the steady state (or the probability of failure) for Here, when it is determined (or predicted) as a failure state through the multilayer neural network model, the power conversion switch device operates to drive the slave module. In this case, the manager can remove the fault factor by checking the faulty part of the master module while the slave module is operating.

After that, the monitoring system uses the multilayer neural network model in the same way to continuously monitor the amount of power produced by the inverter of the slave module and the data generated in the surrounding meter and junction box, and configure the inverter, watt-hour meter, etc. of the slave module using the multilayer neural network model. Predict the steady state (or probability of failure) for an element. Here, when a failure state is determined (or predicted) through the multi-layer neural network model, the power conversion switch device operates to drive the master module.

The monitoring system included in the inverter drive system according to the present invention uses a storage device for storing monitoring data (basic data for fault diagnosis such as the amount of power produced by the inverter, data generated by peripheral meters and access boxes), and a multilayer neural network model. and a failure diagnosis server for diagnosing whether there is a failure. Here, the storage device may be implemented in a cloud environment.

The fault diagnosis server transmits the diagnosis result to the manager device and outputs the diagnosis result through a predefined GUI, so that the manager can check the status of the inverter module.

The inverter driving system according to the present invention can be applied to an existing inverter operating system. The solar inverter can be broadly divided into a string method and a central method. The string method uses an inverter per series group of modules, and MPPT control for each string is possible, and effective energy harvesting is good for partial shade. However, when applied to a large-capacity power plant, the maintenance cost increases because the number of inverters is too large, and the central control of the inverter is not performed, which is somewhat inappropriate in terms of system protection such as prevention of independent operation. The central method is a combination of all modules in series and parallel, and has the disadvantage that the energy harvest is somewhat low. The central method has advantages in that the system protection is advantageous by using a single inverter and the maintenance cost is low.

Recently, as a measure to compensate for such shortcomings, a multi-central method in which a large-capacity central inverter is connected in parallel to implement a single large-capacity inverter system has been proposed. The multi-central inverter is a structure in which central inverters are connected in parallel. When configuring a power generation system, it is composed of several inverters instead of one inverter. By securing the efficiency, it is possible to improve the efficiency of the solar power generation facility. However, in order to maximize the advantages of the multi-central method and to operate the inverter module stably, it is a very important technical task to perform failure prediction when a problem occurs in one inverter in the system. The inverter driving system according to the present invention can be applied to a multi-central inverter system, as shown in FIG. 13 , thereby solving the above-described technical problem.

Also, recently, a solar micro-inverter with improved stability by connecting one inverter with a very small capacity to one solar cell module has been proposed. This solar micro-inverter has a disadvantage in that the amount of power generated from the entire connected module is separated from the system when an abnormality occurs in the inverter. As shown in FIG. 14 , the inverter driving system according to the present invention can be applied to a solar micro-inverter system, thereby solving the above-described problem.

6. Construction of the present invention

Hereinafter, a solar inverter driving system (hereinafter referred to as a solar inverter driving system) using a multilayer neural network failure diagnosis model according to the present invention and a driving method performed in the solar inverter driving system will be described, and the multilayer neural network diagnosis model will be described below. The detailed content described above, such as the definition and the learning process, will be omitted.

Referring to FIG. 5 , the solar inverter driving system 100 includes a solar panel 110 , a master module 121 and a slave module 122 , a monitoring system 130 , and a power conversion switching device 140 . It may be composed of

The solar panel 110 is a device that converts solar energy into electrical energy and outputs power. Here, the solar panel 110 is not limited to the type, shape, output amount, etc., and if it is a device that generates power through solar energy, it should be interpreted as corresponding to the solar panel 110 according to the present invention.

The master module 121 and the slave module 122 each include an inverter module, and converts power output from the solar panel 110 into alternating current and transmits the converted power to the power system. Here, as shown in FIG. 12 , the master module 121 and the slave module 122 may further include additional parts for performing intended functions, such as a junction box and a watt-hour meter, in addition to the inverter module.

The monitoring system 130 collects monitoring data for the master module 121 and the slave module 122, and uses a predefined multi-layer neural network failure diagnosis model to determine the failure state for the master module 121 and the slave module 122. to decide

In one embodiment, the monitoring system 130 collects and stores monitoring data including the inverter generation amount and the watt-hour meter generation amount, and the master module 121 and the slave module 122 using a predefined multi-layer neural network failure diagnosis model. ) may be configured to include a fault diagnosis server that determines a fault condition for

The solar inverter driving system 100 according to the present invention is configured such that the master module 121 preferentially transmits power to the power system. That is, the slave module 122 is electrically cut off from the solar panel 110 , and the master module 121 is electrically connected to the solar panel 110 , so that the master module 121 is preferentially connected to the solar panel ( 110) is configured to transmit the power produced in the power grid.

As data associated with the master module 121 , the monitoring system 130 may collect monitoring data including an inverter power generation amount and an watt-hour meter power generation amount. Thereafter, the monitoring system 130 determines a failure state (normal state, inverter failure, watt-hour meter failure, etc.) for the master module 121 using a predefined multi-layer neural network failure diagnosis model.

On the other hand, in the multilayer neural network failure diagnosis model, as described above, the converted power generation of a predefined value (eg, 99 kW), the watt-hour meter check power generation amount, the inverter check power generation amount and error (watt-hour meter check power generation amount - inverter check power generation amount) are defined as inputs, , steady state, inverter failure state, and watt-hour meter failure state are defined as outputs, and it can correspond to a multi-layer perceptron neural network model that has been trained using pre-collected basic data.

When the master module 140 is determined by the monitoring system 130 as a failure state, the monitoring system 130 transmits a power switching control signal to the power conversion switching device 140, and the power conversion switching device 140 is The master module 121 may be blocked from the solar panel 110 , and the slave module 122 may be switched to be connected to the solar panel 110 to operate the slave module 122 .

According to the present invention, the failure of the solar inverter module can be diagnosed more accurately and quickly using the multilayer neural network failure diagnosis model, and stable operation is possible by applying the master-slave drive type inverter module.

The above-described preferred embodiments of the present invention have been disclosed for the purpose of illustration, and various modifications, changes, and additions may be made by those skilled in the art with respect to the present invention within the spirit and scope of the present invention, and such modifications, changes and Additions should be considered to fall within the scope of the following claims.

100: solar inverter drive system
110: solar panel
121: master module
122: slave module
130: monitoring device
140: power conversion switching device

Claims (6)

In a solar inverter driving system using a multilayer neural network failure diagnosis model,
a solar panel that converts solar energy into electrical energy to output power;
a master module and a slave module each including an inverter module, converting the power output from the solar panel into AC and transmitting it to the power system; and
a monitoring system that collects monitoring data for the master module and the slave module, and determines a failure state for the master module and the slave module using a predefined multi-layer neural network failure diagnosis model;
includes,
The master module is configured to preferentially transmit the power to the power grid, wherein when the master module is determined to be in a fault state by the monitoring system, the master module is cut off and the slave module is configured to operate,
A solar inverter driving system using a multi-layer neural network failure diagnosis model.
According to claim 1, The solar inverter driving system,
When the master module is determined to be in a fault state by the monitoring system, the master module is cut off from the solar panel, and the slave module is connected to the solar panel and switched to operate, a power conversion switching device further characterized in that it comprises
A solar inverter driving system using a multi-layer neural network failure diagnosis model.
According to claim 1, wherein the monitoring system,
A storage device that collects and stores monitoring data including inverter power generation and watt-hour meter power generation;
It characterized in that it comprises a failure diagnosis server for determining the failure state for the master module and the slave module using the multi-layer neural network failure diagnosis model,
A solar inverter driving system using a multi-layer neural network failure diagnosis model.
According to claim 1, wherein the multi-layer neural network failure diagnosis model,
Conversion power generation of a predefined value, power meter check power generation, inverter check power generation amount and error (watt-hour meter check power generation amount - inverter check power generation amount) are defined as inputs,
Normal state, inverter failure state and watt-hour meter failure state are defined as outputs,
Characterized in that it corresponds to a multi-layer perceptron neural network model that has been trained using pre-collected basic data,
A solar inverter driving system using a multi-layer neural network failure diagnosis model.
In a solar inverter driving method using a multi-layer neural network failure diagnosis model performed in a solar inverter driving system,
The solar inverter drive system,
a solar panel that converts solar energy into electrical energy to output power;
a master module and a slave module each including an inverter module, converting the power output from the solar panel into AC and transmitting it to the power system; and
a monitoring system that collects monitoring data for the master module and the slave module, and determines a failure state for the master module and the slave module using a predefined multi-layer neural network failure diagnosis model;
It consists of
collecting, by the monitoring system, monitoring data including an inverter generation amount and an watt-hour meter generation amount as data associated with the master module;
determining, by the monitoring system, a failure state for the master module using the multi-layer neural network failure diagnosis model; and
When the master module is determined to be in a failure state, the master module is cut off from the solar panel through a power conversion switching device and the slave module is connected to the solar panel to operate; including,
A method of driving a solar inverter using a multi-layer neural network failure diagnosis model.
According to claim 5, wherein the multi-layer neural network failure diagnosis model,
Conversion power generation of a predefined value, power meter check power generation, inverter check power generation amount and error (watt-hour meter check power generation amount - inverter check power generation amount) are defined as inputs,
Normal state, inverter failure state and watt-hour meter failure state are defined as outputs,
Characterized in that it corresponds to a multi-layer perceptron neural network model that has been trained using pre-collected basic data,
A method of driving a solar inverter using a multi-layer neural network failure diagnosis model.

Images