KR102553193B1 - System and method for predicting photovoltaic power generation - Google Patents

Abstract

According to an embodiment of the present invention, a solar power generation amount prediction system comprises: a data collection unit that collects sensing data and solar power generation amount data from an environmental sensor and a solar power generation system; a data pre-processing unit that pre-processes the sensing data and the solar power generation amount data; a multi-layer perceptron model learning unit that learns a multi-layer perceptron model based on the pre-processed sensing data and the pre-processed solar power generation amount data; and a solar power generation amount prediction unit that predicts a solar power generation amount by using the multi-layer perceptron model.

Description

Solar power generation prediction system and method {System and method for predicting photovoltaic power generation}

Embodiments of the present invention relate to a system and method for predicting solar power generation using a multi-layer perceptron model.

In order to cope with environmental pollution and energy shortage, solar power is attracting attention as a future energy solution, and the installation of photovoltaic power generation systems is recommended as a policy as the new renewable energy mandatory use policy and greenhouse gas emission trading system are implemented according to the global trend. Solar power plant capacity is gradually increasing.

The photovoltaic power generation system includes solar modules and inverters as important facilities, and as the performance of the facilities deteriorates over time, the efficiency of photovoltaic power generation decreases. To prevent this, the importance of facility maintenance is emerging. It is becoming.

For facility maintenance, accurate diagnosis of facilities is required. , a method of diagnosing an abnormal state using the solar power generation data measured by an inverter is being used.

Recently, as the performance of hardware has improved and analysis methods using machine learning have become common, methods of diagnosing abnormal conditions of facilities by applying machine learning to improve prediction accuracy are increasing.

Embodiments of the present invention are to provide a more accurate and efficient solar power generation prediction system and method based on machine learning.

A solar power generation prediction system according to an embodiment of the present invention includes a data collection unit for collecting sensing data and solar power generation data from an environment sensor and a solar power generation system; a data preprocessing unit preprocessing the sensing data and the photovoltaic power generation data; a multi-layer perceptron model learning unit that learns a multi-layer perceptron model based on the pre-processed sensing data and the pre-processed solar power generation data; and a photovoltaic power generation estimation unit that predicts solar power generation using the multi-layer perceptron model.

In this embodiment, the sensing data may include insolation data and temperature data, and the solar power generation data may include DC power generation data and AC power generation data of an inverter included in the photovoltaic system.

In this embodiment, the data pre-processing unit may extract input data from the sensing data and the photovoltaic power generation data through a correlation analysis with the AC power generation data.

In this embodiment, the data pre-processing unit may extract the insolation data and the DC power generation data as input data.

In this embodiment, the multilayer perceptron model includes one input layer, two hidden layers, and one output layer, and the input layer includes a node representing the solar radiation data and a node representing the DC power generation data, , The output layer may include a node representing the AC generation amount data.

In this embodiment, the multilayer perceptron model includes one input layer, two hidden layers, and one output layer, and the input layer is composed of a node representing the solar radiation data and a node representing the DC power generation data, , One of the two hidden layers may consist of 4 nodes, the other of the two hidden layers may consist of 2 nodes, and the output layer may consist of nodes representing the AC generation amount data.

In this embodiment, the multilayer perceptron model includes 1 input layer, 2 hidden layers, and 1 output layer, the input layer has 2 nodes, and one of the 2 hidden layers has 4 nodes , the other of the two hidden layers may consist of two nodes, and the output layer may consist of one node.

In this embodiment, the multilayer perceptron model learning unit may learn a predetermined amount of data from among data preprocessed by the data preprocessing unit, and verify the multilayer perceptron model with remaining data excluding the learned data.

In this embodiment, the solar power generation prediction unit may predict new AC power generation data based on a result obtained by inputting new insolation data from the environment sensor and new DC power generation data from the inverter into the multilayer perceptron model.

A solar power generation prediction method according to an embodiment of the present invention includes the steps of collecting sensing data and solar power generation data from an environment sensor and a photovoltaic power generation system by a data collection unit; pre-processing the sensing data and the photovoltaic generation amount data by a data pre-processing unit; learning a multi-layer perceptron model based on the pre-processed sensing data and the pre-processed solar power generation data by a multi-layer perceptron model learning unit; and predicting the amount of solar power generation using the multi-layer perceptron model by a solar power generation predictor.

According to the embodiments of the present invention, by learning and verifying the multi-layer perceptron model based on data highly related to the AC power generation data produced by the solar power system, the accuracy of the solar power generation prediction system using the multi-layer perceptron model is improved. can be improved

In addition, since the multilayer perceptron model of the embodiments of the present invention can make more accurate prediction based on relatively few resources, an efficient solar power generation amount prediction system can be provided.

In addition, since the multilayer perceptron model of the embodiments of the present invention can be learned and verified using two or more training data sets, a solar power generation prediction system with improved accuracy can be provided.

1 is a block diagram showing the structure of a system for predicting solar power generation amount according to an embodiment of the present invention.
2 is a graph showing a correlation between input data and output data of a multilayer perceptron model according to an embodiment of the present invention.
3 is a diagram for explaining a multi-layer perceptron model according to an embodiment of the present invention.
4 is a flowchart illustrating a method for predicting solar power generation amount according to an embodiment of the present invention.
5 is a table showing linear regression progress results for checking efficiency reduction of an inverter according to an embodiment of the present invention.
6 is a graph showing the operation result of the solar power prediction system according to an embodiment of the present invention.

Since the present invention can apply various transformations and have various embodiments, specific embodiments will be illustrated in the drawings and described in detail in the detailed description. However, it should be understood that this is not intended to limit the present invention to specific embodiments, and includes all transformations, 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 related known technologies may obscure the gist of the present invention, the detailed description will be omitted.

In the following embodiments, terms such as first and second may be used to describe various components, but the components should not be limited by the terms. Terms are only used to distinguish one component from another.

Terms used in the following examples are only used to describe specific examples, and are not intended to limit the present invention. Expressions in the singular number include plural expressions unless the context clearly dictates otherwise. In the following embodiments, the terms "include" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, but one or more It should be understood that the presence or addition of other features, numbers, steps, operations, components, parts, or combinations thereof is not precluded.

Embodiments of the invention may be presented as functional block structures and various processing steps. These functional blocks may be implemented with any number of hardware or/and software components that perform specific functions. For example, embodiments of the present invention may be directed to memory, processing, logic, look-up tables, etc., which may perform various functions by means of the control of one or more microprocessors or other control devices. Circuit configurations may be employed. Similar to the components of an embodiment of the invention may be implemented as software programming or software elements, an embodiment of the invention may include various algorithms implemented as data structures, processes, routines, or combinations of other programming constructs. , C, C++, Java (Java), can be implemented in a programming or scripting language such as assembler (assembler). Functional aspects may be implemented in an algorithm running on one or more processors. In addition, embodiments of the present invention may employ conventional techniques for electronic environment setting, signal processing, and/or data processing. Terms such as mechanism, element, means, and configuration may be used broadly and are not limited to mechanical and physical configurations. The term may include a meaning of a series of software routines in association with a processor or the like.

Hereinafter, with reference to FIGS. 1 to 3, the operation of the system for predicting the amount of solar power generation according to an embodiment of the present invention will be described in detail.

1 is a block diagram showing the structure of a system for predicting solar power generation amount according to an embodiment of the present invention.

Referring to FIG. 1 , the solar power generation prediction system 100 includes a data collection unit 110, a data pre-processing unit 130, a multi-layer perceptron model learning unit 150, and a solar power generation estimation unit 170. .

The data collection unit 110 collects sensing data and solar power generation data from the environment sensor 10 and the photovoltaic system 20 .

The environmental sensor 10 installed in the photovoltaic system 20 may obtain insolation data and temperature data.

The photovoltaic power generation system 20 may include a photovoltaic module 21 and an inverter 23 . The solar module 21 may generate electricity, and the inverter 23 converts direct current (DC) power generation data generated by the solar module 21 into alternating current (AC) power generation data. can do.

The solar power generation amount prediction system 100 may be implemented as a Remote Maintenance Controller (RMC) gateway.

The data collection unit 110 may collect temperature data and solar radiation data from the environmental sensor 10 and DC power generation data and AC power generation data from the solar power generation system 20 .

The data collection unit 110 may collect temperature data, solar radiation data, DC power generation data, and AC power generation data from one or more environmental sensors 10 and the photovoltaic system 20 interlocked with each other.

For example, the data collection unit 100 is the temperature from the environment sensor 10 and the photovoltaic system 20 installed in region A, and the environment sensor 10 and photovoltaic system 20 installed in region B, respectively. data, insolation data, DC power generation data, and AC power generation data can be collected.

The data preprocessing unit 130 preprocesses the collected sensing data and the collected photovoltaic power generation data.

The data pre-processing unit 130 may pre-process temperature data, solar radiation data, DC power generation data, and AC power generation data from one or more environment sensors 10 and the photovoltaic system 20 interlocked with each other.

For example, the data pre-processing unit 130 receives the air temperature from the environment sensor 10 and the photovoltaic system 20 installed in region A and the environment sensor 10 and the photovoltaic system 20 installed in region B, respectively. Data, solar radiation data, DC power generation data, and AC power generation data may be stored in separate tables, the two tables may be merged, and temperature data and/or data for which solar radiation data is not input may be removed.

The data pre-processing unit 130 may pre-process DC power generation data and/or AC power generation data using MinMaxScaler.

The data pre-processing unit 130 may select data to be used for learning the multilayer perceptron model by checking the efficiency reduction of the inverter 23 through linear regression analysis.

The data pre-processing unit 130 may select data to be used for learning the multilayer perceptron model through correlation analysis.

The data pre-processing unit 130 according to an embodiment may extract input data from sensing data and photovoltaic power generation data through correlation analysis with AC power generation data.

For example, the data pre-processing unit 130 may select data having a high correlation coefficient with AC power generation data to be predicted later from among the collected sensing data and the collected solar power generation data as input data.

As a result of the correlation analysis by the data pre-processing unit 130, among the sensing data, data having a high correlation coefficient with AC power generation data may be insolation data, and among photovoltaic power generation data, data having a high correlation coefficient with AC power generation data may be DC power generation data. there is.

2 is a graph showing a correlation between input data and output data of a multilayer perceptron model according to an embodiment of the present invention.

(a) of FIG. 2 shows the correlation between DC power generation data (DCP) and AC power generation data (ACP), and (b) shows the correlation between solar radiation data (Levelsolar) and AC power generation data (ACP).

Referring to FIG. 2 , the correlation between DC power generation data (DCP) and AC power generation data (ACP) is stronger than that between solar radiation data (Levelsolar) and AC power generation data (ACP).

The distribution aspect of solar radiation data is greatly influenced by the data collection period, and when sensing data and solar power generation data are collected from a plurality of environmental sensors 10 and the photovoltaic power generation system 20, the data collected in a common period is available.

For example, among all sensing data and solar power generation data collected from the plurality of environmental sensors 10 and the photovoltaic power generation system 20, the sensing data and solar power generation data corresponding to a common period are selectively used for machine learning. It can be.

Referring back to FIG. 1 , the multilayer perceptron model learning unit 150 learns the multilayer perceptron model based on the preprocessed sensing data and the preprocessed solar power generation data.

3 is a diagram for explaining a multi-layer perceptron model according to an embodiment of the present invention.

Referring to FIG. 3 , the multilayer perceptron model may include one input layer, two hidden layers, and one output layer. w in each layer represents a weight.

An input layer may consist of two nodes. For example, the input layer may include a node representing solar radiation data (Solar Radiation) and a node representing DC power generation data (DCP).

One of the two hidden layers may be composed of 4 nodes and the other may be composed of 2 nodes.

An output layer may consist of one node. For example, the output layer may include a node ACP representing AC power generation data.

Referring back to FIG. 1, in the multilayer perceptron model learning unit 150, the data preprocessing unit 130 learns a predetermined amount of data from the preprocessed data to create a multilayer perceptron model, and uses the remaining data except for the learned data to create a multilayered multilayer model. The perceptron model can be verified.

An activation function of the multilayer perceptron model may be a linear function.

Multi-layer perceptron models can be trained through a backpropagation training algorithm. Each neuron constituting the multilayer perceptron model can perform optimization to find optimal parameters that minimize errors while going through a backpropagation process.

는 기계 학습의 학습률,

는 그래디언트(gradient)의 두 번째 모멘트로서 비중심 분산,

은 0으로 나누어 떨어지는 것을 방지하기 위한 매우 작은 수,

는 그래디언트의 첫 번째 모멘트로서 평균을 의미할 수 있다.Optimization may be performed by adaptive moment estimation (Adam), and weights may be derived through Equation 1. Optimization can be used for parameter update of machine learning, through which optimal parameters can be found to minimize the cost function. w is the weight,

is the learning rate of machine learning,

is the non-central variance as the second moment of the gradient,

is a very small number to prevent divisibility by zero,

is the first moment of the gradient and may mean an average.

는 실제 데이터,

는 다층 퍼셉트론 모델을 통해 출력한 예측 데이터일 수 있다. The multi-layer perceptron model learning unit 150 may verify the multi-layer perceptron model through Equation 2. Equation 2 may be an equation for calculating a mean squared error (MES) used as a loss function. n is the number of data,

is the actual data,

may be prediction data output through the multi-layer perceptron model.

는 실제 데이터,

는 예측 데이터일 수 있다.The multilayer perceptron model learning unit 150 may measure the accuracy of prediction data output by the multilayer perceptron model through Equation 3. Equation 4 may be an equation for calculating a Mean Absolute Percentage Error (MAPE). n is the number of data,

is the actual data,

may be prediction data.

The solar power generation estimation unit 170 may predict the solar power generation using a multi-layer perceptron model.

The solar power generation prediction unit 170 according to an embodiment inputs new solar radiation data from the environment sensor 10 and new DC power generation data from the inverter 23 to the multilayer perceptron model, and based on the result obtained, the new AC power generation amount. data can be predicted.

4 is a flowchart illustrating a method for predicting solar power generation amount according to an embodiment of the present invention.

Referring to FIG. 4 , the data collection unit 110 collects sensing data and solar power generation data from the environment sensor 10 and the photovoltaic system 20 (S410).

For example, the sensing data may include insolation data and temperature data, and the solar power generation data may include DC power generation data and AC power generation data of an inverter included in the photovoltaic power generation system.

Subsequently, the data pre-processing unit 130 pre-processes the sensing data and the photovoltaic power generation data (S430).

For example, the data pre-processing unit 130 may extract insolation data and DC power generation data as input data from sensing data and photovoltaic power generation data through correlation analysis with AC power generation data.

Subsequently, the multilayer perceptron model learning unit 150 learns the multilayer perceptron model based on the preprocessed sensing data and the preprocessed solar power generation data (S450).

A multi-layer perceptron model may include one input layer, two hidden layers, and one output layer.

The input layer consists of nodes representing insolation data and nodes representing direct current generation data, one of the two hidden layers is composed of 4 nodes, the other of the two hidden layers is composed of 2 nodes, and the output layer is AC It may be composed of nodes representing power generation data.

Next, the solar power generation prediction unit 170 predicts the solar power generation using the multi-layer perceptron model (S470).

5 is a table showing linear regression progress results for checking efficiency reduction of an inverter according to an embodiment of the present invention.

Referring to FIG. 5 , whether or not the efficiency of the inverter 23 according to an embodiment of the present invention actually decreases can be confirmed through a change in the linear regression coefficient Coef.

The linear regression coefficient (Coef) means a coefficient obtained by performing a linear regression analysis on the DC power generation data and AC power generation data by year. As the linear regression coefficient (Coef) decreases over time, the It can be seen that the ratio decreases.

6 is a graph showing the operation result of the solar power prediction system according to an embodiment of the present invention.

6(a) shows the result of predicting the AC power generation data of an inverter operated for one year using the learned and verified multilayer perceptron model.

6(b) shows the result of predicting AC power generation data of an inverter operated for 2 years using the learned and verified multilayer perceptron model.

Referring to FIG. 6 , it can be seen that the predicted data (pred) generally has a larger value than the actual data (actual), but the predicted data (pred) follows the trend of the actual data (actual) well.

The result of using the multilayer perceptron model in FIG. 6 has a longer prediction period than the linear regression progress result of FIG. 5, and can predict the reduction ratio of AC power generation to DC power generation more accurately.

So far, the present invention has been mainly looked at with respect to preferred embodiments. Those skilled in the art to which the present invention belongs will understand that the present invention can be implemented in a modified form without departing from the essential characteristics of the present invention.

Therefore, the disclosed embodiments should be considered from a descriptive point of view rather than a limiting point of view. The scope of the present invention is shown in the claims rather than the foregoing description, and the inventions claimed by the claims and inventions equivalent to the claimed inventions should be construed as being included in the present invention.

10: environmental sensor
20: solar power system
100: Solar power generation prediction system

Claims (10)

a data collection unit that collects sensing data and photovoltaic power generation data from an environment sensor and a photovoltaic power generation system;
a data preprocessing unit preprocessing the sensing data and the photovoltaic power generation data;
a multi-layer perceptron model learning unit that learns a multi-layer perceptron model based on the pre-processed sensing data and the pre-processed solar power generation data; and
A solar power generation estimation unit for predicting solar power generation using the multi-layer perceptron model;
The sensing data includes insolation data and temperature data,
The photovoltaic power generation data includes DC power generation data and AC power generation data of an inverter included in the photovoltaic power generation system,
The data pre-processing unit
A solar power prediction system for extracting input data from the sensing data and the photovoltaic power generation data through correlation analysis with the AC power generation data. delete delete The method of claim 1,
The data pre-processing unit
A solar power generation prediction system for extracting the insolation data and the DC power generation data as input data. The method of claim 4,
The multilayer perceptron model includes one input layer, two hidden layers, and one output layer,
The input layer includes a node representing the solar radiation data and a node representing the DC power generation data,
The output layer includes a node representing the AC power generation data. The method of claim 4,
The multilayer perceptron model includes one input layer, two hidden layers, and one output layer,
The input layer is composed of a node representing the solar radiation data and a node representing the DC power generation data,
One of the two hidden layers is composed of 4 nodes, and the other of the two hidden layers is composed of 2 nodes,
The output layer is composed of nodes representing the AC power generation data. The method of claim 1,
The multilayer perceptron model includes one input layer, two hidden layers, and one output layer,
The input layer consists of two nodes, one of the two hidden layers consists of four nodes, the other of the two hidden layers consists of two nodes, and the output layer consists of one node. prediction system. The method of claim 1,
The multi-layer perceptron model learning unit
Among the data preprocessed by the data preprocessor, a predetermined amount of data is learned, and the multilayer perceptron model is verified with the remaining data except for the learned data. The method of claim 1,
The solar power generation prediction unit predicts new AC power generation data based on a result obtained by inputting new solar radiation data from the environmental sensor and new DC power generation data from the inverter into the multilayer perceptron model. Solar power generation prediction system. Collecting sensing data and photovoltaic power generation data from an environment sensor and a photovoltaic power generation system by a data collection unit;
pre-processing the sensing data and the photovoltaic generation amount data by a data pre-processing unit;
learning a multi-layer perceptron model based on the pre-processed sensing data and the pre-processed solar power generation data by a multi-layer perceptron model learning unit; and
Predicting the amount of solar power generation using the multi-layer perceptron model by a solar power generation predictor; including,
The sensing data includes insolation data and temperature data,
The photovoltaic power generation data includes DC power generation data and AC power generation data of an inverter included in the photovoltaic power generation system,
The step of preprocessing the data is,
A method for predicting solar power generation, wherein input data is extracted from the sensing data and the photovoltaic power generation data through correlation analysis with the AC power generation data.

Images