KR102544233B1 - Method and apparatus for compensating photovoltaic output prediction uncertainty and intermittency - Google Patents

Abstract

A method performed by an apparatus for compensating for output prediction uncertainty and intermittence of photovoltaic power generation according to an embodiment is based on past power generation data of a photovoltaic power generation facility as a dependent variable and disaster prevention weather observation data as an initial independent variable, Additional creation of independent variables to reflect time-unit delay values and characteristics of time series data, and selection of variables to be used as inputs based on the correlation between the dependent variable, the initial independent variable, and the additionally generated independent variables. a step of inputting the selected variables into a pre-learned deep learning artificial neural network and predicting the amount of photovoltaic power generation of the photovoltaic power generation facility as an output of the deep learning artificial neural network; and Applying intermittence by multiplying the photovoltaic power generation data value by a random value following a Gaussian distribution, comparing the predicted photovoltaic power generation amount and the photovoltaic power generation amount to which the intermittency is applied, and according to the result of the comparison, the BESS and determining an operation mode as one of stop, charging operation, and discharging operation.

Description

Method and device for compensating output prediction uncertainty and intermittency of solar power generation

The present invention relates to a method for compensating for output prediction uncertainty and intermittency of a solar power plant and an apparatus for performing the method.

Carbon-neutral (Net-Zero) policies to curb climate change and temperature rise have become a global paradigm between countries, and after the Paris Agreement and the UN Climate Summit, 121 countries have joined the Climate Alliance for the 2050 carbon-neutral goal. Various carbon reduction policies, such as carbon border tax and shipping greenhouse gas emission trade, are being discussed to achieve carbon neutrality. Diffusion is accelerating the spread of renewable energy.

Korea is also actively expanding the supply of renewable energy according to the Basic Energy Plan and the Renewable Energy 3020 Implementation Plan policy. Compared to 2015, the increase in domestic solar power generation in 2019 showed a steep increase from a minimum of 74 [%] to a maximum of 421 [%] by region. In addition, with the addition of the 2050 carbon neutral promotion strategy along with the existing policy, the expansion of renewable energy supply is expected to accelerate further.

Among currently used renewable energies, variable renewable energy (VRE), which shows output fluctuations according to external factors, inherently has intermittency and predictability of output depending on natural conditions. In the case of prediction uncertainty, it refers to a characteristic that inevitably involves prediction errors in predicting the generation output of variable renewable energy, and intermittency means that the generation output of variable renewable energy changes rapidly depending on external conditions. Based on the background mentioned above, if the share of variable renewable energy in power generation increases within the distribution system, rapid output changes due to intermittence may cause negative effects on power quality, and economic dispatch and shutdown of base power sources due to forecast uncertainty It can affect the plan and cause unexpected output limitations. In expanding the supply of renewable energy, the intermittency and predictive uncertainty of variable renewable energy are essential to be overcome.

Republic of Korea Patent Publication No. 10-2021-0047627, published on April 30, 2021.

According to an embodiment, a method and apparatus for simultaneously compensating for intermittency and predictive uncertainty of photovoltaic power generation, which is a type of variable renewable energy, through a battery energy storage system (BESS) are provided.

The problem to be solved by the present invention is not limited to those mentioned above, and another problem to be solved that is not mentioned will be clearly understood by those skilled in the art from the description below.

The method performed by the solar power generation prediction uncertainty and intermittency compensation device according to the first aspect is based on the past power generation data of the solar power generation facility as a dependent variable and disaster prevention weather observation data as an initial independent variable, the amount of solar power generation Additional creation of independent variables to reflect time-unit delay values and characteristics of time series data, and selection of variables to be used as inputs based on the correlation between the dependent variable, the initial independent variable, and the additionally generated independent variables. a step of inputting the selected variables into a pre-learned deep learning artificial neural network and predicting the solar power generation amount of the photovoltaic power generation facility as an output of the deep learning artificial neural network; The step of applying intermittence by multiplying the actual photovoltaic power generation data value of the power generation facility by a random value following the Gaussian distribution, the step of comparing the predicted photovoltaic power generation amount and the photovoltaic power generation amount to which the intermittency was applied, and the result of the comparison Accordingly, determining an operation mode of the BESS as one of stop, charging operation, and discharging operation.

An apparatus for compensating for output prediction uncertainty and intermittence of photovoltaic power generation according to a second aspect includes a data acquisition unit that acquires power generation data and disaster prevention weather observation data of a photovoltaic power generation facility, and the photovoltaic power generation as an output corresponding to a predetermined input. It includes a deep learning artificial neural network pretrained to predict the amount of solar power generation of the facility, and a processor unit, wherein the processor unit takes past power generation data of the photovoltaic power generation facility acquired by the data acquisition unit as a dependent variable, and The disaster prevention meteorological observation data is taken as an initial independent variable, and independent variables are additionally created to reflect the time-unit delay value of solar power generation and the characteristics of time-series data, and the dependent variable, the initial independent variable, and the additionally generated independent variables are created. Based on the correlation between variables, variables to be used as inputs are selected, and the selected variables are input to the pre-learned deep learning artificial neural network, and the solar power generation amount of the solar power generation facility is output as the output of the deep learning artificial neural network. Obtaining a prediction value, applying intermittence by multiplying the actual photovoltaic power generation data value of the photovoltaic facility by a random value following a Gaussian distribution, and comparing the photovoltaic power generation prediction value with the photovoltaic power generation amount to which the intermittency was applied, According to the result of the comparison, the operation mode of the BESS is determined to be one of stop, charging operation, and discharging operation.

The computer readable recording medium storing the computer program according to the third aspect may, when the computer program is executed by the processor, perform the method performed by the apparatus for compensating for the output prediction uncertainty and intermittence of the photovoltaic power generation. Contains commands to do so.

The computer program stored in the computer readable recording medium according to the fourth aspect is, when the computer program is executed by the processor, the processor performs the method performed by the output prediction uncertainty and intermittency compensation device of the photovoltaic power generation. Contains commands to do so.

According to an embodiment of the present invention, intermittency and predictive uncertainty of photovoltaic power generation, which is a type of variable renewable energy, are simultaneously compensated through the BESS. In this way, since forecast uncertainty and intermittency are compensated for, it is possible to minimize the change in the start-up plan or economic dispatch plan of base-load generation by supplying the same amount of power as the predicted generation amount, and to solve the problem of limiting renewable energy output due to supply-demand imbalance. can be alleviated

1 is a block diagram of an apparatus for compensating for output uncertainty and intermittence of photovoltaic power generation according to an embodiment of the present invention.
2 is a flowchart illustrating a network topology visualization method performed by an apparatus for compensating for output uncertainty and intermittence of photovoltaic power generation according to an embodiment of the present invention.
3 is a result of correlation coefficient analysis between variables.
4 is a result of measuring the performance of the least squares method according to an increase in an independent variable.
5 is a configuration diagram of a deep learning artificial neural network.
6 shows the results of optimizing hyperparameters.
7 is a result of predicting solar power generation output through RNN.
8 is data for the first 7 days of prediction results of photovoltaic power generation output through RNN.
9 is a graph of actual photovoltaic power generation to which intermittency is applied.
10 is a ramp rate (RR) of actual photovoltaic power generation to which intermittency is applied.
11 shows a BESS charge/discharge section.
12 is a SoC result of BESS.
13 shows the results of intermittency and uncertainty compensation through BESS operation.

Advantages and features of the present invention, and methods of achieving them, will become clear with reference to the embodiments described below in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, and only these embodiments make the disclosure of the present invention complete, and common knowledge in the art to which the present invention belongs. It is provided to completely inform the person who has the scope of the invention, and the present invention is only defined by the scope of the claims.

The terms used in this specification will be briefly described, and the present invention will be described in detail.

The terms used in the present invention have been selected from general terms that are currently widely used as much as possible while considering the functions in the present invention, but these may vary depending on the intention of a person skilled in the art or precedent, the emergence of new technologies, and the like. In addition, in a specific case, there is also a term arbitrarily selected by the applicant, and in this case, the meaning will be described in detail in the description of the invention. Therefore, the term used in the present invention should be defined based on the meaning of the term and the overall content of the present invention, not simply the name of the term.

In the entire specification, when a part is said to 'include' a certain component, it means that it may further include other components, rather than excluding other components, unless otherwise stated.

In addition, the term 'unit' used in the specification means software or a hardware component such as FPGA or ASIC, and 'unit' performs certain roles. However, 'wealth' is not limited to software or hardware. 'Unit' may be configured to reside in an addressable storage medium and may be configured to reproduce one or more processors. Thus, as an example, 'unit' refers to components such as software components, object-oriented software components, class components and task components, processes, functions, properties, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays and variables. Functions provided within components and 'parts' may be combined into a smaller number of elements and 'parts' or further separated into additional elements and 'parts'.

Hereinafter, with reference to the accompanying drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily carry out the present invention. And in order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted.

1 is a block diagram of an apparatus for compensating for output uncertainty and intermittence of photovoltaic power generation according to an embodiment of the present invention.

Referring to FIG. 1, an apparatus 100 for compensating for output prediction uncertainty and intermittence of photovoltaic power generation according to an embodiment includes a data acquisition unit 110, a deep learning artificial neural network 120, and a processor unit 130, Study (140) and / or BESS (150) may be further included.

The data acquisition unit 110 acquires power generation data and disaster prevention weather observation data of the photovoltaic power generation facility and provides them to the deep learning artificial neural network 120 and/or the processor unit 130. For example, the data acquisition unit 110 may receive power generation data and disaster prevention weather observation data from an external server through a communication channel. Alternatively, power generation data and disaster prevention weather observation data of the photovoltaic power generation facility may be input from a peripheral device through a serial interface.

The deep learning artificial neural network 120 is pre-learned to predict the solar power generation amount of the solar power generation facility as an output corresponding to a predetermined input. For example, the deep learning artificial neural network 120 may be in a pre-learned state of a learning dataset including variables as inputs and a predicted value of solar power generation as a label. For example, the pre-learned deep learning artificial neural network 120 may output predicted values of the solar power generation amount of the solar power generation facility in response to input variables. For example, the pre-learned deep learning artificial neural network 120 may be optimized for at least one item of sequence size, batch size, dropout, and epoch among hyperparameters. For example, the deep learning artificial neural network 120 may use a multi layered perceptron (MLP), a recurrent neural network (RNN), a long short term memory (LSTM), and a gate recurrent unit (GRU).

The processor unit 130 takes the past power generation data of the photovoltaic power generation facility acquired by the data acquisition unit 110 as a dependent variable, takes the disaster prevention weather observation data as an initial independent variable, and sets the time unit delay value of the solar power generation amount and additionally create independent variables to reflect the characteristics of time series data. Here, the processor unit 130 may integrate the dependent variable and the initial independent variable into one data set, replace missing values among disaster prevention weather observation data with other disaster prevention weather observation data, and then additionally generate independent variables. For example, the processor unit 130 divides the dependent variable and the initial independent variable again, and then additionally generates at least one independent variable in addition to the initial independent variable among the solar power generation trend, seasonality, hour, day, week, month, and year. can do.

Then, the processor unit 130 selects variables to be used as inputs based on the correlation between the dependent variable, the initial independent variable, and the additionally generated independent variable. For example, the processor unit 130 may sort the initial independent variables and additionally generated independent variables according to the height of the correlation, sequentially increase the sorted independent variables, and measure the performance of the least squares method according to the increase in the variables. and at least one of the sorted independent variables may be selected as a variable to be used as an input, based on a result of measuring the performance of the least squares method.

Then, the processor unit 130 inputs the selected variables to the pre-learned deep learning artificial neural network, and obtains a predicted value of the solar power generation amount of the photovoltaic power generation facility as an output of the deep learning artificial neural network. Then, the processor unit 130 applies intermittence by multiplying the actual photovoltaic power generation data value of the photovoltaic facility by a random value following a Gaussian distribution. Then, the processor unit 130 compares the estimated solar power generation amount with the solar power generation amount applied with intermittency. And, the processor unit 130 determines the operation mode of the BESS to one of stop, charge operation, and discharge operation according to the result of the comparison.

The provision unit 140 provides a result of processing by the processor unit 130 . For example, the provision unit 140 may output information about the operation mode of the BESS determined by the processor unit 130 to a peripheral device connected through a serial interface. Alternatively, the provision unit 140 may transmit information about the operation mode of the BESS determined by the processor unit 130 to the outside through a communication channel. Alternatively, information on the operation mode of the BESS determined by the processor unit 130 may be visualized through a display device, a printer, or the like.

The BESS 150 may stop or operate according to the result of processing by the processor unit 130 . For example, the BESS 150 may be driven in one operation mode among stop, charge, and discharge according to a result of processing by the processor unit 130 .

2 is a flowchart illustrating a network topology visualization method performed by an apparatus for compensating for output uncertainty and intermittence of photovoltaic power generation according to an embodiment of the present invention.

Hereinafter, a network topology visualization method performed by the apparatus 100 for compensating for output uncertainty and intermittence of photovoltaic power generation according to an embodiment of the present invention will be described in detail with reference to FIGS. 1 to 13 .

In order to check the output forecast uncertainty of photovoltaic power generation, the generation amount prediction should be performed first. Accordingly, the apparatus 100 for compensating for output prediction uncertainty and intermittence of photovoltaic power generation according to an embodiment of the present invention proceeds with prediction of power generation using a deep learning artificial neural network that can effectively reflect nonlinear characteristics. In addition, a deep learning artificial neural network that prevents overfitting and has excellent predictive performance by creating additional independent variables or removing highly correlated independent variables through a data preprocessing process before performing prediction through the deep learning artificial neural network. Select the optimal input variable for

The data acquisition unit 110 acquires power generation data and disaster prevention weather observation data of the photovoltaic power generation facility and provides them to the processor unit 130 (S201). For the description of an embodiment of the present invention, the output data of Korea South-East Power's third photovoltaic (350 [kW]) power generation facility provided through the public data portal was used as a dependent variable, and as an initial independent variable, the Korea Meteorological Administration's The disaster prevention meteorological observation data of Goseong-gun, Gyeongsangnam-do, provided through the meteorological data open portal, were used. Table 1 shows the items of disaster prevention meteorological observation data.

No. Independent Variables One Temperatures [℃] 2 Wind speed [m/s] 3 wind direction [deg] 4 Rainfall [mm] 5 Humidity [%]

The processor unit 130 integrates the dependent variable and the independent variable into one dataset. Then, the processor unit 130 performs a process of filling in missing values that are not recorded or measured with respect to the power generation data and disaster prevention meteorological observation data of the photovoltaic power generation facility provided from the data acquisition unit 110 . For example, if only 1 hour is missed, it can be replaced by taking the average of the values before and after the unmeasured time, and if it is not measured for more than 1 hour, it can be replaced with disaster prevention meteorological observation data of Saryangdo Island close to Goseong-gun. If both Goseong-gun and Saryangdo's disaster prevention meteorological observation data are missing, the values of other years of the missing time can be averaged and replaced.

Thereafter, the processor unit 130 classifies the independent variable and the dependent variable again, and additionally creates an independent variable to reflect the characteristics of the time-unit delay value and time-series data of the amount of solar power generation in addition to the initially configured independent variable (S202). . For example, the processor unit 130 determines the trend, seasonality, hour, and day of the solar power generation amount to reflect the delay values of 1 hour and 2 hours and the characteristics of time-series data. , week, month, and year can be additionally created as independent variables. Here, the trend can be called a trend factor, and means a long-term variable factor. These trends are available for the entire period of training data. Seasonality can be called a seasonal factor, and means changes that repeat with a cycle. The period of seasonality uses a cycle of 12 months (1 year).

Also, the processor unit 130 determines correlation coefficients between variables including independent variables additionally generated through the previous process. Here, correlation coefficients between variables that can be grasped by the processor unit 130 are illustrated in FIG. 3 . Subsequently, the processor unit 130 selects variables to be used as inputs of the deep learning artificial neural network 120 based on the correlation between the dependent variable, the initial independent variable, and the additionally generated independent variable. For example, the processor unit 130 sorts the initial independent variables and additionally generated independent variables in descending order of inter-variable correlation according to the height of the correlation, and then sequentially increases the sorted independent variables, Ordinary least squares (OLS) performance may be measured, and at least one of the aligned independent variables may be selected as a variable to be used as an input based on a result of measuring the least squares performance. Here, a variance inflation factor (VIF) can be used to remove highly correlated independent variables. VIF, also called the variance expansion coefficient, is an index for determining whether there is a correlation between independent variables. The elimination of such highly correlated independent variables is called multicollinearity elimination. A high correlation between independent variables means that the correlation between two variables is high, so it is not necessary to use both variables, and if two variables are used, errors may occur during analysis. When there are several independent variables, the method of determining VIF determines multicollinearity by measuring the relationship between variables by performing a regression analysis with a specific independent variable as the dependent variable and the remaining independent variables as independent variables when there are several independent variables.

For example, the processor unit 130 may measure the performance of the least squares method according to the increase of the independent variable using mean absolute error (MAE) and mean square error (MSE), which are expressed in Equations 1 and 2 below. is defined as The results of measuring the performance of the least squares method according to the increase of the independent variable are illustrated in FIG. 4 .

는 측정값이고,

는 예측값이며,

은 데이터의 수이다.here,

is the measured value,

is the predicted value,

is the number of data.

Then, the processor unit 130 selects an optimal variable to be used as an input to the deep learning artificial neural network 120 based on the previous processing result (S203). According to the example of FIG. 4, when 13 independent variables are applied to the least squares method in the test set, the lowest MAE and MSE are shown, compared to 14 independent variables, and the least squares method performance is the best. In this case, a total of 13 independent variables to be used as inputs of the deep learning artificial neural network 120 are determined.

According to one embodiment, the deep learning artificial neural network 120 may use MLP, RNN, LSTM, or GRU, and the configuration of the deep learning artificial neural network 120 is illustrated in FIG. 5 . Here, the deep learning artificial neural network 120 may be in a state in which hyperparameters directly set by the user are optimized. For example, the deep learning artificial neural network 120 may be optimized for at least one of hyperparameters of sequence size, batch size, dropout, and epoch. . For example, in order to optimize the hyperparameters of the deep learning artificial neural network 120, a grid search, which is a method of finding optimal hyperparameter values while constantly increasing parameter values, may be used. Hyperparameter optimization results of the deep learning artificial neural network 120 are illustrated in FIG. 6 and Table 2. In FIG. 6, (a) is an MLP, (b) is an RNN, (c) is an LSTM, and (d) is a GRU.

Hyperparameter MLP RNN LSTM GRU Sequence Size - 28 41 43 Batchsize 28 10 13 11 Dropout 0 [%] 0 [%] 0 [%] 0 [%] Epoch 30 55 27 16

Table 3 shows the performance of the deep learning artificial neural network 120 with hyperparameter optimization applied. In particular, the recursive circuit-based deep learning artificial neural networks RNN, LSTM, and GRU showed superior performance compared to MLP. This means that the recursive circuit-based deep learning artificial neural network is a suitable prediction model for time-series data such as solar power generation. It can be seen that the RNN showed the lowest error after optimizing the hyperparameters. The result of predicting the amount of solar power generation through the deep learning artificial neural network 120 employing RNN is the same as the example of FIG. 7 (S204). 8 shows the enlarged results of the first 7 days of prediction. These results indicate that the prediction through the deep learning artificial neural network 120 also has uncertainty.

Hyperparameter Optimization Performance Evaluation MAE MSE Train Test Train Test MLP 7.23 210.81 8.37 277.60 RNN 2.61 13.67 2.79 16.94 LSTM 6.07 143.18 6.03 153.63 GRU 6.07 149.73 6.57 169.02

Output intermittence of photovoltaic power generation is caused by external factors such as weather and season, and its characteristics can be clearly identified in the minute unit solar power output. When the data acquisition unit 110 acquires actual solar power generation data per hour and provides it to the processor unit 130, there is a limit to indicating intermittence. In this case, the processor unit 130 may generate intermittency probabilistically and apply it to the simulation. For example, the processor unit 130 is the deep learning artificial neural network 120, and the amount of power generation having a negative value among the results of the test set output through the RNN and the amount of power generated during the time when the solar power generation facility cannot produce power can be treated as 0 [kW]. Afterwards, in order to consider the most extreme situation when applying the BESS 150, intermittency may be applied based on the day when the largest error occurred among the first 7 days of prediction (S205). Table 4 shows the performance of the first 7 days of prediction, and as a result, the first day of prediction (7/2) showed the largest error.

Date 1st
(7/2) 2nd
(7/3) 3rd
(7/4) 4th
(7/5) 5th
(7/6) 6th
(7/7) 7th
(7/8) MAE 2.46 1.36 0.95 1.15 1.87 1.23 2.09 MSE 22.73 3.73 3.73 4.30 14.69 8.63 11.15

According to the embodiment, based on the results of Table 4, the initial first day of prediction may be selected as an intermittent application target. The application method can change actual solar power generation data per hour to solar power generation data per minute through linear interpolation. Thereafter, intermittency may be applied by multiplying the actual photovoltaic power generation data value changed in units of 1 minute by a random value following a Gaussian distribution having an average of 1 and a standard deviation of 0.2. As a result, as shown in FIG. 9 , it can be seen that intermittency is applied to the actual amount of photovoltaic power generation.

The high intermittence of solar power generation means that the output of solar power generation fluctuates rapidly. In order to numerically represent such intermittency, a ramp rate (RR) can be used, and through this, a change in current output compared to previous output can be represented. The mathematical expression of RR is defined by Equation 3 below.

는 시간

에서의 태양광 발전량이고,

은

의 시간 변화이다.here,

time

is the amount of solar power in

silver

is the time change of

For example, when RR is limited to ±10 and ±1 to 5 [%/min], respectively, in order to maintain power quality from the intermittency of variable renewable energy, the RR of the actual solar power generation amount applied with intermittence as shown in FIG. 9 is As shown in FIG. 10, it can be seen that a number of intermittents exceeding RR ±10 [%/min] are included.

BESS (150) is an energy storage device that supplies power at a time when it is necessary after storing the produced power, and is used for frequency adjustment, renewable energy connection, peak reduction, and emergency generator replacement. Uncertainty and intermittency in solar power generation can cause supply and demand imbalances, shutdowns of base generation and changes in economic dispatch plans, and can cause output limitations. According to the present invention, prediction uncertainty and intermittence can be simultaneously compensated for through the BESS 150 in order to minimize a change in the base power plant operation plan.

The processor unit 130 compares the actual amount of photovoltaic power generation with intermittence applied and the predicted amount of photovoltaic power generation (S206). Then, the processor unit 130 determines the operation mode of the BESS 150 as an idle state when the actual amount of photovoltaic power generation applied with intermittency and the predicted amount of photovoltaic power generation are the same as the comparison result (S209). Here, the processor unit 130 determines the operation mode of the BESS 150 as a charging operation when the actual amount of photovoltaic power generation to which intermittency is applied is greater than the predicted amount of photovoltaic power generation as a result of the comparison (S210). Conversely, the processor unit 130 determines the operation mode of the BESS 150 as a discharge operation when the actual amount of photovoltaic power generation to which intermittency is applied is smaller than the predicted amount of photovoltaic power generation (S211).

In this way, information on the operation mode of the BESS 150 determined by the processor unit 130 may be provided to the outside through the providing unit 140 . For example, the provision unit 140 may output information about the operation mode of the BESS determined by the processor unit 130 to a peripheral device connected through a serial interface. Alternatively, the provision unit 140 may transmit information about the operation mode of the BESS determined by the processor unit 130 to the outside through a communication channel. Alternatively, information on the operation mode of the BESS determined by the processor unit 130 may be visualized through a display device, a printer, or the like.

In addition, according to the control signal of the processor unit 130, the BESS 150 may be driven in the operation mode determined in steps S209 to S211 among the idle state, charging operation, and discharging operation.

Meanwhile, the processor unit 130 sequentially and repeatedly performs operations from step S206 until the preset time is reached, but ends the operation mode determination algorithm for the BESS 150 after the preset time is reached. For example, the processor unit 130 ends the operation mode determination algorithm for the BESS 150 after late evening hours (eg, 23:00) when solar power is not generated (S212).

Also, the processor unit 130 may calculate the required capacity of the BESS 150. Through Table 4, the initial operating capacity of the BESS (150) can be calculated based on the first day of the prediction (FIG. 9) with the largest error, and when the BESS operation mode decision algorithm of FIG. 2 is applied, the BESS (150) The charge and discharge section of is shown in FIG. 12. The initial operating capacity of the BESS 150 can be calculated using Equation 4, and the initial operating capacity of the BESS 150 can be calculated by integrating the output according to the charging and discharging of the BESS 150 for one day. As a result, the initial operating capacity of BESS was calculated as 21.20 [kWh].

는 BESS 초기 운영 용량이며,

는 시간

에서의 BESS 출력이다.here,

is the BESS initial operating capacity,

time

This is the BESS output from

When the BESS (150) starts operating, assuming that the initial state of charge (SoC) is 50 [%], it is operated for one day to reduce the lifespan of the BESS (150) due to complete discharge and to prevent fire due to overcharging When the SoC range is limited to 20∼80 [%], the minimum capacity of the BESS (150) is calculated to be 46.30 [kWh]. The SoC change according to the daily operation of the BESS is shown in FIG. 13, and it can be seen that the SoC standard is satisfied.

When the BESS 150 performs the charging/discharging operation as shown in FIG. 12 under the control of the processor unit 130, solar power generation having the same output as the predicted solar power generation amount is possible, and intermittent and predictive solar power generation is possible as shown in FIG. 13 It can be seen that the uncertainty is compensated.

The renewable energy market continues to grow towards achieving carbon neutrality. Accordingly, the problem of forecast uncertainty and output intermittence of renewable energy is a problem that must be solved in expanding the base of variable renewable energy. According to an embodiment of the present invention, prediction uncertainty and intermittence can be simultaneously compensated for through BESS.

Combinations of each step in each flowchart attached to the present invention may be performed by computer program instructions. Since these computer program instructions may be loaded into a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing equipment, the instructions executed by the processor of the computer or other programmable data processing equipment function as described in each step of the flowchart. create a means to do them. These computer program instructions can also be stored on a computer usable or computer readable medium that can be directed to a computer or other programmable data processing equipment to implement functions in a particular way, so that the computer usable or computer readable It is also possible that the instructions stored on the recording medium produce an article of manufacture containing instruction means for performing the functions described in each step of the flowchart. The computer program instructions can also be loaded on a computer or other programmable data processing equipment, so that a series of operational steps are performed on the computer or other programmable data processing equipment to create a computer-executed process to generate computer or other programmable data processing equipment. Instructions for performing processing equipment may also provide steps for executing the functions described at each step in the flowchart.

Further, each step may represent a module, segment or portion of code that includes one or more executable instructions for executing the specified logical function(s). It should also be noted that in some alternative embodiments it is possible for the functions mentioned in the steps to occur out of order. For example, two steps shown in succession may in fact be performed substantially concurrently, or the steps may sometimes be performed in reverse order depending on the function in question.

The above description is merely an example of the technical idea of the present invention, and various modifications and variations can be made to those skilled in the art without departing from the essential qualities of the present invention. Therefore, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but to explain, and the scope of the technical idea of the present invention is not limited by these embodiments. The protection scope of the present invention should be construed according to the claims below, and all technical ideas within the scope equivalent thereto should be construed as being included in the scope of the present invention.

100: Output prediction uncertainty and intermittency compensation device for photovoltaic power generation
110: data acquisition unit
120: Deep learning artificial neural network
130: processor unit
140: provision unit
150: BESS

Claims (13)

delete As a method of performing the output prediction uncertainty and intermittency compensation device of photovoltaic power generation,
Based on the past power generation data of the solar power generation facility as a dependent variable and disaster prevention weather observation data as an initial independent variable, additionally generating an independent variable to reflect the characteristics of the time unit delay value and time series data of the solar power generation amount;
Selecting variables to be used as inputs based on correlations between the dependent variable, the initial independent variable, and the additionally generated independent variable;
Inputting the selected variables into a pre-learned deep learning artificial neural network and predicting the solar power generation amount of the solar power generation facility as an output of the deep learning artificial neural network;
Applying intermittence by multiplying the actual photovoltaic power generation data value of the photovoltaic power generation facility by a random value following a Gaussian distribution;
Comparing the estimated amount of photovoltaic power generation with the amount of photovoltaic power generation to which the intermittency is applied;
According to the result of the comparison, determining an operation mode of a battery energy storage system (BESS) as one of stop, charge operation, and discharge operation,
The step of additionally generating the independent variable,
Integrating the dependent variable and the initial independent variable into one data set, and replacing the missing values among the disaster prevention weather observation data with other disaster prevention weather observation data
Output prediction uncertainty and intermittency compensation method of photovoltaic power generation. According to claim 2,
The step of additionally generating the independent variable,
After dividing the dependent variable and the initial independent variable again, the trend, seasonality, hour, day, week, month, and year of the solar power generation (year) to additionally create at least one independent variable in addition to the initial independent variable
Output prediction uncertainty and intermittency compensation method of photovoltaic power generation. As a method of performing the output prediction uncertainty and intermittency compensation device of photovoltaic power generation,
Based on the past power generation data of the solar power generation facility as a dependent variable and disaster prevention weather observation data as an initial independent variable, additionally generating an independent variable to reflect the characteristics of the time unit delay value and time series data of the solar power generation amount;
Selecting variables to be used as inputs based on correlations between the dependent variable, the initial independent variable, and the additionally generated independent variable;
Inputting the selected variables into a pre-learned deep learning artificial neural network and predicting the solar power generation amount of the solar power generation facility as an output of the deep learning artificial neural network;
Applying intermittence by multiplying the actual photovoltaic power generation data value of the photovoltaic power generation facility by a random value following a Gaussian distribution;
Comparing the estimated amount of photovoltaic power generation with the amount of photovoltaic power generation to which the intermittency is applied;
According to the result of the comparison, determining an operation mode of a battery energy storage system (BESS) as one of stop, charge operation, and discharge operation,
In the step of selecting variables to be used as the input,
After arranging the initial independent variable and the additionally generated independent variable according to the height of the correlation, the sorted independent variable is sequentially increased, the least squares performance according to the increase in the variable is measured, and the least squares performance is measured Based on the result, selecting at least one of the sorted independent variables as a variable to be used as the input
Output prediction uncertainty and intermittence compensation method of photovoltaic power generation. According to claim 4,
The pre-learned deep learning artificial neural network is optimized for at least one item of sequence size, batch size, dropout, and epoch among hyperparameters
Output prediction uncertainty and intermittency compensation method of photovoltaic power generation. According to claim 4,
After the determining step, repeating the following steps from the predicting step until a predetermined time is reached
Output prediction uncertainty and intermittency compensation method of photovoltaic power generation. delete A data acquisition unit for acquiring power generation data and disaster prevention weather observation data of the photovoltaic power generation facility;
A deep learning artificial neural network pretrained to predict the solar power generation amount of the solar power generation facility as an output corresponding to a predetermined input;
Including a processor unit,
The processor unit,
The past power generation data of the photovoltaic power generation facility acquired by the data acquisition unit is taken as a dependent variable, the disaster prevention weather observation data is taken as an initial independent variable, and the time unit delay value of the photovoltaic power generation amount and the characteristics of time series data Create additional independent variables to reflect,
Based on the correlation between the dependent variable, the initial independent variable, and the additionally generated independent variable, variables to be used as input are selected;
Inputting the selected variables to the pre-learned deep learning artificial neural network to obtain a predicted value of the solar power generation amount of the photovoltaic power generation facility as an output of the deep learning artificial neural network,
Applying intermittence by multiplying the actual photovoltaic power generation data value of the photovoltaic power plant by a random value following a Gaussian distribution,
Comparing the solar power generation predicted value and the solar power generation to which the intermittence was applied,
According to the result of the comparison, the operation mode of the BESS is determined as one of stop, charging operation, and discharging operation,
Integrating the dependent variable and the initial independent variable into one data set, replacing the missing value among the disaster prevention weather observation data with other disaster prevention weather observation data, and then generating additional independent variables
Output prediction uncertainty and intermittence compensation device for photovoltaic power generation. According to claim 8,
A data acquisition unit for acquiring power generation data and disaster prevention weather observation data of the photovoltaic power generation facility;
A deep learning artificial neural network pretrained to predict the solar power generation amount of the solar power generation facility as an output corresponding to a predetermined input;
Including a processor unit,
The processor unit,
The past power generation data of the photovoltaic power generation facility acquired by the data acquisition unit is taken as a dependent variable, the disaster prevention weather observation data is taken as an initial independent variable, and the time unit delay value of the photovoltaic power generation amount and the characteristics of time series data Create additional independent variables to reflect,
Based on the correlation between the dependent variable, the initial independent variable, and the additionally generated independent variable, variables to be used as input are selected;
Inputting the selected variables to the pre-learned deep learning artificial neural network to obtain a predicted value of the solar power generation amount of the photovoltaic power generation facility as an output of the deep learning artificial neural network,
Applying intermittence by multiplying the actual photovoltaic power generation data value of the photovoltaic power plant by a random value following a Gaussian distribution,
Comparing the solar power generation predicted value and the solar power generation to which the intermittence was applied,
According to the result of the comparison, determining the operation mode of the BESS to one of stop, charge operation, and discharge operation
The processor unit,
After dividing the dependent variable and the initial independent variable again, generating at least one additional independent variable in addition to the initial independent variable among the trend, seasonality, time, day, week, month, and year of the solar power generation
Output prediction uncertainty and intermittence compensation device for photovoltaic power generation. A data acquisition unit for acquiring power generation data and disaster prevention weather observation data of the photovoltaic power generation facility;
A deep learning artificial neural network pretrained to predict the solar power generation amount of the solar power generation facility as an output corresponding to a predetermined input;
Including a processor unit,
The processor unit,
The past power generation data of the photovoltaic power generation facility acquired by the data acquisition unit is taken as a dependent variable, the disaster prevention weather observation data is taken as an initial independent variable, and the time unit delay value of the photovoltaic power generation amount and the characteristics of time series data Create additional independent variables to reflect,
Based on the correlation between the dependent variable, the initial independent variable, and the additionally generated independent variable, variables to be used as input are selected;
Inputting the selected variables to the pre-learned deep learning artificial neural network to obtain a predicted value of the solar power generation amount of the photovoltaic power generation facility as an output of the deep learning artificial neural network,
Applying intermittence by multiplying the actual photovoltaic power generation data value of the photovoltaic power plant by a random value following a Gaussian distribution,
Comparing the solar power generation predicted value and the solar power generation to which the intermittence was applied,
According to the result of the comparison, the operation mode of the BESS is determined as one of stop, charging operation, and discharging operation,
After sorting the initial independent variable and the additionally generated independent variable according to the height of the correlation, sequentially increasing the sorted independent variable and measuring the performance of the least squares method according to the increase in the variable,
Selecting at least one of the sorted independent variables as a variable to be used as the input based on the result of measuring the least squares performance
Output prediction uncertainty and intermittence compensation device for photovoltaic power generation. According to claim 10,
The pre-learned deep learning artificial neural network is optimized for at least one item of sequence size, batch size, dropout, and epoch among hyperparameters
Output prediction uncertainty and intermittence compensation device for photovoltaic power generation. A computer-readable recording medium storing a computer program,
When the computer program is executed by a processor,
A computer-readable recording medium comprising instructions for causing the processor to perform the method of any one of claims 2 to 6. As a computer program stored on a computer-readable recording medium,
When the computer program is executed by a processor,
A computer program comprising instructions for causing the processor to perform the method of any one of claims 2 to 6.

Images