System for analyzing solar power plant siting based on maps and the controlling method thereof

Abstract

A system for analyzing map-based solar power plant locations according to various embodiments of the present invention may include: a data collection server that collects solar power plant information, including weather data, regional solar radiation data, power plant information, electricity sales price data, and power generation data for each solar power plant, from a plurality of solar power plants stored in a database; a solar power generation prediction analysis server that predicts solar power generation based on solar power plant information received from the data collection server; and a solar power plant location analysis server that performs map-based solar power plant location analysis based on information on solar power generation by location received from the solar power generation prediction analysis server.

Description

Technology Field

[0001] The present invention relates to the field of solar power plant site selection technology, and more specifically, to a system and method for predicting the power generation of a solar power plant using GIS information and optimizing the site selection of a new power plant based thereon. Background Technology

[0002] Recently, as the stability of energy supply and demand and the response to climate change have emerged as critical global issues, new and renewable energy businesses, including solar power generation, are rapidly expanding. Solar power plants have the advantage of being installable in various regions due to their relatively short construction periods and modular structures; however, power generation efficiency varies significantly depending on the characteristics of the installation site.

[0003] For example, even for solar power plants of the same capacity, actual power generation and profitability vary significantly depending on regional solar irradiance, weather conditions (temperature, humidity, cloud cover, precipitation, etc.), surrounding topography (altitude, slope, shading, etc.), installation angle, panel type, grid connection costs, and electricity sales prices. Accordingly, the process of selecting the optimal site for a solar power plant is emerging as a critical task for improving power production efficiency and economic viability.

[0004] In the past, it was common practice to calculate approximate power generation by simply considering average solar radiation data or specific meteorological data. In this case, however, it was difficult to adequately reflect various variables—such as topographical characteristics surrounding the actual installation site (shading, elevation differences, etc.), panel characteristics, and operational information—resulting in low prediction accuracy.

[0005] If sites are selected using power generation indicators estimated uniformly across fixed units (administrative districts, large-scale areas, etc.), it is difficult to properly reflect meteorological and topographical factors that vary subtly even within that area. For instance, mountainous or coastal regions differ from flatlands in terms of solar radiation, wind speed, and cloud patterns, so predictions based on simple average values ​​can deviate significantly from reality.

[0006] Some existing technologies often failed to comprehensively analyze factors such as the installation history, maintenance information, fluctuations in electricity sales prices, and grid connection status of power plants, frequently limiting themselves to verifying only restricted weather data or licensing information. Consequently, it was difficult to account for operational issues that occur during the actual operation of solar power plants (such as reduced panel efficiency, impact on surrounding facilities, and fluctuations in subsidies and sales prices), leading to limitations in the reliability of long-term profitability forecasts.

[0007] Solar power generation is determined not only by meteorological factors such as solar irradiance, temperature, humidity, and cloud cover, but also by interactions with the panel manufacturer, efficiency, installation angle, surrounding obstacles (shadow effects), and terrain slope. Existing simple regression models or partial machine learning approaches have often failed to adequately reflect the complex interactions among these multidimensional variables.

[0008] As such, it is difficult to select the optimal site based solely on temporary estimates derived from simple historical data. Furthermore, if actual power generation falls significantly short of expectations after selection, problems may arise such as delayed profit generation relative to investment costs or a severe decline in facility operational efficiency. The problem to be solved

[0009] In order to overcome the technical limitations mentioned above, the present invention provides a system and method that supports intuitive decision-making when selecting a site by comprehensively learning and analyzing various actual measurement data collected from multiple solar power plants, correcting residuals for first-order model predictions second-order through additional parameters (higher-order terms, interaction terms, topographic characteristics, etc.), and visualizing the results based on GIS, thereby enabling a significant increase in the efficiency of constructing and operating solar power plants. means of solving the problem

[0010] A system for analyzing map-based solar power plant locations according to various embodiments of the present invention may include: a data collection server that collects solar power plant information, including weather data, regional solar radiation data, power plant information, electricity sales price data, and power generation data for each solar power plant, from a plurality of solar power plants stored in a database; a solar power generation prediction analysis server that predicts solar power generation based on solar power plant information received from the data collection server; and a solar power plant location analysis server that performs map-based solar power plant location analysis based on information on solar power generation by location received from the solar power generation prediction analysis server.

[0011] According to one embodiment, the solar power generation prediction analysis server can process solar power plant information into GIS information by performing data cleaning on information received from a data collection server.

[0012] According to one embodiment, the solar power generation prediction analysis server sets weather data of a solar power plant, regional solar radiation data, power plant information, electricity sales price data, and power generation data as a plurality of parameters; calculates regression coefficients of a linear model representing the relationship between the plurality of parameters and the result value, which is solar power generation; calculates correction values ​​when adding the plurality of parameters when calculating the predicted solar power generation amount by the linear model; and determines a linear model by determining at least one parameter and regression coefficient that minimizes the correction value.

[0013] According to one embodiment, the correction value is calculated using the following mathematical formula, and

[0014]

[0015] SSE is the sum of squared residuals, SST is the sum of squared total deviations, n is the number of observations, and p may be the number of parameters + 1.

[0016] According to one embodiment, the solar power generation prediction analysis server has a solar power generation prediction value calculated by the correction value ( 1) and the actual amount of power generated (y) collected from nearby solar power plants real Calculate the difference between ); train a second-order correction model by setting the above residual as the dependent variable and a plurality of parameters including a polynomial or interaction term as independent variables; and the correction value derived from the second-order correction model ( Recall ) In addition to 1, the final predicted power generation ( ) can be produced.

[0017] According to one embodiment, the second correction model comprises, together with actual power generation data obtained from a nearby solar power plant, a seasonal periodic function including a monthly or seasonal index, surrounding terrain information (elevation, terrain slope, shading effect), transformer load factor, panel efficiency coefficient, sin(ωt), cos(ωt), etc., and a squared term (x) for the input variables of the first model. j 2 ) or cross term(x j ×x k It can be characterized by being learned by setting at least one of the parameters as an independent variable.

[0018] According to one embodiment, the secondary correction model is,

[0019]

[0020] Residual (r) in the form of i Predicting ), where z k,i is the k-th input value, γ0,γ among multiple parameters including raw data, additional variables, and non-linear terms used in the quadratic correction model. k ,γ k,l can be a constant term, a first-order term coefficient, a second-order term coefficient, or a cross-term coefficient, respectively.

[0021] According to one embodiment, the training of the second correction model is performed on the residual (r i A step of setting ) as the dependent variable and generating a polynomial expansion vector (zi) composed of the above parameters as independent variables; a step of estimating γ satisfying the following equation through at least one method among the normal equation, gradient descent, and least squares quenching (LSQ);

[0022]

[0023] Residual prediction value i and first-order model prediction 1,i Summing up the final predicted power generation i = 1,i + i It may include a step of deriving.

[0024] According to one embodiment, the solar power generation prediction analysis server updates γ by re-executing the secondary correction model training whenever additional actual power generation data continuously measured from a nearby solar power plant is collected, and

[0025] Using the updated γ, the final predicted power generation for map-based candidate points can be recalculated in real-time or periodically.

[0026] According to one embodiment, the solar power generation prediction analysis server can re-run the secondary correction model learning to update γ whenever actual power generation data continuously measured at nearby solar power plants is additionally collected, and recalculate the final predicted power generation for a map-based candidate point in real time or periodically using the updated γ. Effects of the invention

[0027] According to the present invention, a system and method capable of significantly improving prediction accuracy are provided by obtaining a basic predicted value through multiple linear regression (first-order model) and then modeling the error (residual) once again (second-order correction) by comparing it with the actual power generation of a nearby solar power plant. The present invention aims to enable automated improvement of prediction accuracy over a long period through dynamic (real-time) updates by selecting and analyzing candidate sites in conjunction with a map (GIS) system. Through this, the present invention aims to contribute to increasing investment efficiency by comprehensively considering economic and environmental factors when selecting the optimal site for a solar power plant. Brief explanation of the drawing

[0028] FIG. 1 illustrates a system for analyzing map-based solar power plant locations according to an embodiment of the present invention. FIG. 2 is a block diagram of the internal configuration of the servers of the system for analyzing the location of a map-based solar power plant according to the present invention. FIG. 3 is an exemplary flowchart of the operation of a system for analyzing map-based solar power plant locations according to an embodiment of the present invention. FIG. 4 is an exemplary diagram of a system for analyzing map-based solar power plant locations according to an embodiment of the present invention. FIG. 5 is an exemplary drawing of a data collection server according to an embodiment of the present invention. FIG. 6 is an exemplary drawing of a solar power generation prediction analysis server according to an embodiment of the present invention. FIG. 7 is an exemplary drawing of a solar power plant location analysis server according to an embodiment of the present invention. Specific details for implementing the invention

[0029] The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined only by the scope of the claims. Accordingly, in some embodiments, well-known process steps, well-known device structures, and well-known techniques are not specifically described to avoid the present invention being interpreted ambiguously. Throughout the specification, like reference numerals refer to like components.

[0030] In this specification, terms such as first, second, third, etc., may be used to describe various components, but these components are not limited by these terms. The terms are used for the purpose of distinguishing one component from other components. For example, without departing from the scope of the present invention, a first component may be named a second or third component, etc. Since the present invention is capable of various modifications and may have various embodiments, specific embodiments are illustrated in the drawings and described in detail.

[0031] However, this is not intended to limit the invention to specific embodiments and should be understood to include all modifications, equivalents, and substitutions that fall within the spirit and scope of the invention. Similar reference numerals have been used for similar components in the description of each figure.

[0032] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0033] The terms used in this application are used merely to describe specific embodiments and are not intended to limit the invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, terms such as "comprising" or "having" are intended to specify the presence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0034] Hereinafter, embodiments of the present invention will be described with reference to the attached drawings. Identical reference numerals in each drawing indicate identical components. In describing the present invention, specific descriptions regarding related known functions or configurations are omitted to avoid obscuring the essence of the invention.

[0036] FIG. 1 is a schematic diagram illustrating the configuration of a system (100) for analyzing map-based solar power plant locations.

[0037] The system (100) may be configured to include a data collection server (110), a solar power generation prediction analysis server (120), and a solar power plant location analysis server (130). The data collection server (110) may be configured to collect and store solar power plant information, including weather data, regional solar radiation data, power plant information, electricity sales price data, and power generation data for each solar power plant, from a plurality of solar power plants stored in a database.

[0038] The solar power generation prediction analysis server (120) may be configured to predict solar power generation for a specific period in the future by comprehensively analyzing solar radiation and weather information by location based on multiple solar power plant information received from the data collection server (110).

[0039] The solar power plant location analysis server (130) may be configured to analyze and recommend the optimal location of a solar power plant based on a map, based on the solar power generation prediction results by location provided by the solar power generation prediction analysis server (120).

[0040] In step 310, the data collection server (110) can collect solar power plant information including weather data, local solar radiation data, power plant information, electricity sales price data, and power generation amount data from a plurality of solar power plants stored in a database.

[0041] To this end, the data collection server (110) is connected to a network with sensors and monitoring devices corresponding to each solar power plant, so that it can periodically collect information or transmit and receive information in real time.

[0042] The collected solar power plant information is stored in a database and can be transmitted to a solar power generation prediction analysis server (120) for subsequent processing and analysis.

[0043] In step 320, the solar power generation prediction analysis server (120) can predict solar power generation based on multiple solar power plant information received from the data collection server (110).

[0044] At this time, the solar power generation prediction analysis server (120) can construct a prediction model by integrating meteorological information, satellite data, and past power generation history, and can increase accuracy by considering solar radiation and environmental variables that vary by location and season.

[0045] The predicted solar power generation information can be processed into a power generation pattern by location, time, or specific period and transmitted to a solar power plant location analysis server (130).

[0046] In step 330, the solar power plant location analysis server (130) can perform a map-based solar power plant location analysis based on location-specific solar power generation prediction information received from the solar power generation prediction analysis server (120).

[0047] For example, the solar power plant location analysis server (130) can recommend a location suitable for installing a power plant on a map, or select candidate sites by considering environmental regulations, natural disaster risk, and existing infrastructure conditions.

[0048] In addition, the solar power plant location analysis server (130) can visualize the simulation results and provide them to the user so that they can compare and evaluate the power generation efficiency or economic feasibility at a glance.

[0049] As described above, the system (100), composed of a data collection server (110), a solar power generation prediction analysis server (120), and a solar power plant location analysis server (130), can systematically collect and analyze information on multiple solar power plants to efficiently derive the optimal solar power plant location based on a map. The entire system may be diagrammed as shown in FIG. 4, and each server may be diagrammed as shown in FIG. 5 to 7.

[0050] FIG. 2 is a block diagram showing the internal configuration of a server according to one embodiment of the present invention.

[0051] Referring to FIG. 2, a server (200) according to various embodiments of the present invention includes a processor (210), memory (220), and a communication module (230). The server (200) of FIG. 2 may be a data collection server (110), a solar power generation prediction analysis server (120), or a solar power plant location analysis server (130).

[0052] The processor (210) may be composed of one or more cores and may include a processor (210) for data analysis and deep learning, such as a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), or a tensor processing unit (TPU) of an electronic device (100). The processor (210) can read a computer program stored in memory (220) and perform data processing for machine learning according to one embodiment of the present invention. In addition, the processor (210) can control the configuration of the server (200) to operate and implement the operation of the overall system.

[0053] For example, the processor (210) can typically control the overall operation of the server (200). The processor (210) can provide or process appropriate information or functions to the user by processing signals, data, information, etc. that are input or output through the components described above, or by running applications stored in memory (220).

[0054] Additionally, the processor (210) can control at least some of the components of the server (200) to run an application stored in memory (220). Furthermore, the processor (210) can operate at least two or more of the components included in the server (200) in combination to run the application.

[0055] According to one embodiment of the present invention, the processor (210) can perform operations for training a neural network. The processor (210) can perform operations for training a neural network, such as processing input data for training in deep learning (DL), extracting features from input data, calculating errors, and updating the weights of the neural network using backpropagation. At least one of the CPU, GPGPU, and TPU of the processor (210) can process the training of the network function. For example, the CPU and GPGPU can together process the training of the network function and data classification using the network function. In addition, the communication module (130) has a general configuration, and a detailed description is omitted.

[0056] According to one embodiment, the solar power generation prediction analysis server can process solar power plant information into GIS information by performing data cleaning on information received from a data collection server.

[0057] For example, the location (latitude and longitude), elevation, and information on nearby terrain and features of each solar power plant can be converted into spatial attributes in a database, allowing them to be intuitively displayed on a map. Furthermore, it is possible to apply coordinate transformations (e.g., WGS84, UTM-K, etc.) required for map integration, or perform GIS processing logic such as calculating distances between multiple points and calculating terrain slope.

[0058] For example, regarding a specific point (candidate site), it is possible to comprehensively query and analyze factors such as the existence of data on surrounding power plants, the distribution of topography and features, and the distribution of solar radiation by region. Through this, various factors considered when selecting a solar power plant site (such as power generation efficiency, ease of installation, and profitability of electricity sales) can be systematically compared.

[0059] As in this embodiment, by converting information received from a data collection server into GIS information, the user (or administrator) can check the predicted power generation or profitability of multiple candidate sites at a glance on a map screen, and furthermore, select the optimal solar power plant location through the analysis of various scenarios.

[0060] In this way, the solar power generation prediction analysis server performs data cleaning and processes the cleaned information into GIS information, thereby enabling accurate and consistent data-based analysis, which can increase the efficiency and reliability of the solar power generation business.

[0061] According to one embodiment, the solar power generation prediction analysis server can set weather data of the solar power plant, regional solar radiation data, power plant information, electricity sales price data, and power generation data as a plurality of parameters.

[0062] The solar power generation prediction analysis server can set weather data collected from the solar power plant (e.g., temperature, humidity, wind speed, etc.), local solar radiation data, power plant information (e.g., installation location, panel capacity, tilt angle, direction, etc.), electricity sales price data (e.g., unit price / amount of electricity based on generation amount, etc.), and generation data (e.g., historical generation measured by time period) as multiple parameters, respectively.

[0063] For example, x1 can be defined as regional solar radiation, x2 as average temperature, x3 as panel capacity, and x4 as an indicator of electricity sales revenue, respectively.

[0064] The solar power generation prediction analysis server can calculate regression coefficients of a linear model representing the relationship between multiple parameters and the resulting value, solar power generation.

[0065] The above-mentioned solar power generation prediction analysis server assumes that the relationship between the above-mentioned plurality of parameters and the result value, solar power generation (y), is linear, and can construct a linear model of the following form.

[0066]

[0067] Here represents the predicted solar power generation, and β0, β1, … , β p is a regression coefficient corresponding to the above multiple parameters.

[0068] The above-mentioned solar power generation prediction analysis server uses, for example, the least squares method or conventional regression analysis techniques to β0 to β p It can be estimated.

[0069] The solar power generation prediction analysis server can calculate correction values ​​when multiple parameters are added when calculating the predicted solar power generation amount based on the above linear model.

[0070] According to one embodiment, the solar power generation prediction analysis server evaluates the performance of the model by adding a plurality of parameters one by one or in batches. Correction value It can calculate.

[0071] In this case, the above correction value is the adjusted coefficient of determination (Adjusted R 2 It can be defined in the form of ), and, for example, can be expressed as in the mathematical formula below:

[0072]

[0073] Here, SSE stands for Sum of Squared Errors, which can be the sum of squared errors between the predicted values ​​and the actual values. SST stands for Total Sum of Squares, which can be the sum of squared deviations between the actual values ​​and the mean values. n is the number of observations (samples), and p can represent the number of parameters + 1 (including the constant term).

[0074] The solar power generation prediction analysis server can determine a linear model by determining at least one parameter and regression coefficient that minimizes the correction value. The solar power generation prediction analysis server comprises the plurality of parameters and the result value Solar power generation Assuming the relationship between (y) is linear, of the following form linear model It can be composed of:

[0075] The correction value (Adjusted R) defined in the mathematical formula of the above correction value 2 ) is generally a value The larger the The model's explanatory power can be evaluated as high. However, if the "correction value" is based on the "sum of squared residuals" (e.g., when using SSE directly), this minimum A method of derivation is also possible, and the adjusted coefficient of determination (Adjusted R 2 Making ) “maximum” can also be seen as having an equivalent effect.

[0076] Therefore, in one embodiment, the adjusted coefficient of determination (Adjusted R²) when including newly added parameters 2 )go maximum By finding the condition where (or SSE is minimized), the optimal model configuration (parameter combination and β coefficient) can be determined.

[0077] For example, parameter x p+1 Adjusted R of the model with added 2 Calculate and parameter x p+1 , x p+2 Adjusted R of the model with added 2 After calculating, the highest Adjusted R 2 The model showing the value can be adopted as the final linear model.

[0078] Finally, the above-mentioned solar power generation prediction analysis server uses the determined model (parameter set and regression coefficients) to predict the solar power generation value ( ) can be produced.

[0079] Calculated predicted value ( ) can be transmitted, for example, to a GIS-based solar power plant location analysis server and can be used to visualize the expected power generation of candidate sites on a map or to perform an economic evaluation (such as electricity sales revenue).

[0080] That is, according to the system of the present invention, the input may be weather data of a solar power plant (temperature, humidity, solar irradiance, etc.), regional solar irradiance data, power plant information, electricity sales revenue data, power generation data (actual measurement), and additional multiple parameters (additional topographic characteristics, seasonal indicators, etc.). In this regard, the system of the present invention β0 to β p Construct a basic linear regression model including, and whenever a new parameter is added, Adjusted R 2 Calculate the base correction value and the correction value The best (e.g., Adjusted R 2 (maximum) The set of parameters and coefficients that drive the result can be determined. Through this, the output value is finally determined Linear model (parameters and regression coefficients) and Predicted power generation calculated from the above model ( Can have ).

[0081] The above solar power generation prediction analysis server is a solar power generation prediction value calculated by the above correction value ( 1) and the actual amount of power generated (y) collected from nearby solar power plants real The difference between ) can be calculated.

[0082] According to one embodiment, the solar power generation prediction analysis server can calculate a solar power generation prediction value (y1) from a first-order model determined using correction values ​​(e.g., adjusted coefficient of determination or SSE-based indicators). In addition, actual power generation (y) collected from nearby solar power plants real ) can be prepared. Based on the same point in time or similar environmental conditions, the above server, y1 and y real By calculating the difference between them, the residual (r) can be obtained.

[0083] The above solar power generation prediction analysis server can learn a second-order correction model by setting the above residual as the dependent variable and multiple parameters including a polynomial or an interaction term as independent variables.

[0084] The above solar power generation prediction analysis server, the above residual r dependent variable It can be set as. Subsequently, multiple parameters including polynomials or interaction terms independent variable A second-order correction model composed of can be trained.

[0085] According to one embodiment, the parameter used at this time is Actual power generation data obtained from a nearby solar power plant Along with, the following items may be included:

[0086] Monthly or seasonal index (e.g., January, February, spring, summer, autumn, winter, etc.)

[0087] Surrounding terrain information (elevation, terrain slope, shading influence, etc.)

[0088] Transformer load rate , Panel efficiency coefficient

[0089] seasonal periodic function (sin(ωt), cos(ωt), etc.)

[0090] Input variables of the first-order model for squared term (x j 2 ) or Intersection (x j ×x k )

[0091] These parameters {z k,i When} is used, the second correction model can be trained in the form of the following mathematical equation:

[0092]

[0093] Here, z k,ican be the k-th input value among multiple parameters (including raw data, additional variables, and non-linear terms) used in the quadratic correction model. γ0,γ k ,γ k,l are each constant term , 1st term coefficient , Secondary or cross term coefficients It may apply to.

[0094] The solar power generation prediction analysis server uses the correction value derived from the above secondary correction model ( Recall ) In addition to 1, the final predicted power generation ( ) can be produced.

[0095] According to one embodiment, the solar power generation prediction analysis server uses conventional methods such as least squares method (LSQ), gradient descent (GD), and normal equations. Regression analysis techniques By applying, γ0,γ k The back can be estimated.

[0096] As a result, the predicted residual value for each observation i ( i )this Actual residual (r i The model can be trained to minimize the error with ).

[0097] The trained second-order correction model is, Correction value ( i) can be calculated.

[0098] The above solar power generation prediction analysis server, as follows, predicts the value from the first model ( i 1 The correction value calculated by the second correction model in ) i )second Addition As, Final predicted power generation ( i ) can be obtained:

[0099]

[0100] Through this multi-stage correction process, multiple parameters including seasonal, topographical, and periodic characteristics are taken into account while reflecting the actual power generation of nearby solar power plants, thereby obtaining power generation prediction results with improved accuracy compared to simple linear models.

[0101] The training of the above second correction model can be performed by the following steps.

[0102] Residual (r i A polynomial expansion vector (zi) can be generated by setting ) as the dependent variable and configuring the above parameters as independent variables.

[0103] According to one embodiment, the solar power generation prediction analysis server, the predicted power generation amount calculated from the first model ( 1,i ) and actual power generation collected from nearby solar power plants (y real,i The difference between ) is the residual (r i It can be defined as ).

[0104] At this time, the residual r i By setting as the dependent variable and configuring multiple parameters to be used for secondary correction (e.g., monthly / seasonal index, topographic information, transformer load factor, intersection term, squared term, etc.) as independent variables, a polynomial expansion vector (z i Can generate ).

[0105] For example, (z i ) is (1, z 1,i , z 2,i , ..., z m,i , z 1,I, z 2,i , {z 2,i It can be constructed in a form that includes all original variables, cross terms, square terms, etc., such as}^2, ...).

[0106] After that, γ satisfying the following equation can be estimated through at least one of the normal equation, gradient descent, and least squares method (LSQ).

[0107]

[0108] According to one embodiment, the solar power generation prediction analysis server can estimate γ satisfying the objective equation as shown in (Equation 1) below:

[0109] Here, r i is the first-order model prediction value ( 1,i ) and actual value(y real,i It is the error (residual) between ), and (z i ) can represent a vector created by polynomial expansion of the above independent variables (parameters), and γ can represent the coefficient vector of the second-order correction model.

[0110] The above objective function can be solved using at least one method, such as the normal equation, gradient descent, or least squares (LSQ). Through this process, the predicted residual value ( i = (z i ) T γ) is the actual residual (r i γ can be determined so that the error with ) is minimized.

[0111] Finally, the residual prediction value i and first-order model predictions 1,i Summing up the final predicted power generation i= 1,i + i It can derive.

[0112] According to one embodiment, as a result of the above second-order correction model, for each observation i, the predicted residual value ( i ) can be obtained. After that, the first-order model prediction value ( 1,i ) and residual predictions calculated from the second-order correction model ( i Summing ) to get the final predicted power generation ( i) can be derived as follows:

[0113]

[0114] Through this, it is possible to improve prediction accuracy by correcting for non-linear influences (cross-terms, seasonality, terrain effects, etc.) that were missed by existing first-order models.

[0115] According to one embodiment, the solar power generation prediction analysis server can update γ by re-executing the secondary correction model learning whenever additional actual power generation data continuously measured from a nearby solar power plant is collected.

[0116] At this time, using the updated γ, the final predicted power generation for the map-based candidate point (or already installed power plant) ( By recalculating ) in real-time or periodically, model accuracy can be gradually improved.

[0117] Input:

[0118] 1. Residual data {r i} - Difference between the first predicted value and the actual value

[0119] 2. Parameter (Independent Variable) Vector z i - Month / season index, topographic characteristics, intersection term, squared term, etc.

[0120] 3. Polynomial expansion (z generated by (Polynomial Expansion) i )

[0121] · Output:

[0122] 1. Estimated by the second-order correction model coefficient γ - (Equation 1).

[0123] 2. Predicted residuals i =(z i )γ

[0124] 3. Final predicted power generation i=y i1 + i

[0125] Technical Execution Method (Summary)

[0126] Set the residual as the dependent variable : Predicted value of the first-order model( 1 ) and actual value(y real Calculate the difference r i It is possible to prepare.

[0127] Polynomial expansion vector generation : Multiple parameters expansion (Polynomial / Interaction) i It can constitute ).

[0128] learning coefficient γ γ can be obtained by solving the optimization problem using the normal equation, gradient descent, least squares method, etc.

[0129] Calculation of final forecast : Predicted residuals using learned γ ( i Calculate ) and use this as the first predicted value ( i 1 In addition to ), the final power generation amount ( i It can be confirmed as ).

[0130] Dynamic Update : When new actual power generation data is added, (Equation 1)-based learning is performed again to update γ, and for the map-based candidate points It can be recalculated.

[0131] As in this embodiment, Secondary correction model Polynomial expansion during learning ((z i By using ), difficulties that are difficult to reflect with existing linear models alone Non-linear interaction and local characteristics It can handle it easily. In addition, the nearby power plant's Actual power generation As this accumulates Model Update Since prediction accuracy is gradually improved through this, for solar power plant site analysis and maximizing operational efficiency Continuous contribution Can do.

[0132] The solar power generation prediction analysis server according to one embodiment can re-run the secondary correction model learning to update γ whenever actual power generation data continuously measured at nearby solar power plants is additionally collected, and recalculate the final predicted power generation for a map-based candidate point in real time or periodically using the updated γ.

[0133] According to one embodiment, the solar power generation prediction analysis server can re-run the secondary correction model learning to update γ whenever actual power generation data continuously measured at nearby solar power plants is additionally collected, and recalculate the final predicted power generation for a map-based candidate point in real time or periodically using the updated γ.

[0134] According to one embodiment, the solar power generation prediction analysis server continuously measures at a nearby solar power plant Actual power generation data cast Additionally Can be collected.

[0135] For example, measurement sensor values ​​can be received on a daily, weekly, monthly, or real-time basis and accumulated and stored in a database.

[0136] According to one embodiment, the solar power generation prediction analysis server is, Additional collected actual power generation data In order to reflect, Secondary correction model Learning about (i.e., regression analysis including the γ estimation process) Re-execute You can do it.

[0137] At this point, new observations (residuals based on actual power generation and first-order model predictions) are added to the previously used residual information (ri), and based on this Normal equation, gradient descent, least squares method γ with (LSQ), etc. renewal You can do it.

[0138] The updated γ is a second-order correction model (e.g., i =(z iIn )γ) Predicted residuals ( i To calculate ) New coefficient It can be.

[0139] According to one embodiment, the solar power generation prediction analysis server immediately applies the updated γ to the map-based candidate point (or already selected point) Final predicted power generation ( )second Real-time or Periodic by Re-production You can do it.

[0140] If the user wants to check prediction results periodically, reflect new actual data at regular time intervals (e.g., every morning, every week, etc.). Model Update Execute, and based on the update results Final predicted power generation It can be recalculated.

[0141] In addition, in specific situations (e.g., new power plant installation, grid changes, etc.) Real-time When data flows in, through the updated second-order correction model immediately The final predicted power generation amount of the candidate points Real-time It can be updated to.

[0142] With the above configuration, as actual measurement data from nearby solar power plants accumulates over time, the prediction accuracy can be improved by continuously updating the second correction model (γ).

[0143] By displaying the final predicted power generation calculated from the updated model on a map (GIS), the potential power generation for each candidate location can be identified more accurately.

[0144] The shorter the model update cycle (or the more real-time data is secured), the faster and more accurately users can make contract, investment, and operational decisions.

[0145] As described above, the present invention according to one embodiment can update γ by re-executing the second correction model learning whenever the actual power generation amount continuously measured at a nearby solar power plant is accumulated, and by recalculating the final predicted power generation amount of a map-based candidate point in real time or periodically based on the updated γ, it is possible to enable dynamic solar power generation prediction and efficient site selection and operation.

[0147] The above description is merely an illustrative explanation of the technical concept of the present embodiment, and a person skilled in the art to which the present embodiment belongs would be able to make various modifications and variations within the scope of the essential characteristics of the present embodiment. Accordingly, the present embodiments are intended to explain, not limit, the technical concept of the present embodiment, and the scope of the technical concept of the present embodiment is not limited by these embodiments. The scope of protection of the present embodiment shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of the present embodiment.

Claims

Claim 1 A system for analyzing map-based solar power plant locations, comprising: a data collection server that collects solar power plant information, including weather data, regional solar radiation data, power plant information, electricity sales price data, and power generation data for each solar power plant, from a plurality of solar power plants stored in a database; a solar power generation prediction analysis server that predicts solar power generation based on the solar power plant information received from the data collection server; and a solar power plant location analysis server that performs map-based solar power plant location analysis based on information on solar power generation by location received from the solar power generation prediction analysis server. Claim 2 In claim 1, the solar power generation prediction analysis server is a system for analyzing map-based solar power plant locations, which processes solar power plant information into GIS information by performing data cleaning on information received from a data collection server. Claim 3 In claim 1, the solar power generation prediction analysis server sets the solar power plant weather data, regional solar radiation data, power plant information, electricity sales price data, and power generation data as multiple parameters; calculates the regression coefficients of a linear model representing the relationship between the multiple parameters and the resulting solar power generation amount; calculates a correction value when adding multiple parameters in calculating the predicted solar power generation amount by the linear model; and determines the linear model by determining at least one parameter and regression coefficient that minimizes the correction value, a system for analyzing the location of a map-based solar power plant. Claim 4 In Paragraph 3, the above correction value is calculated using the following mathematical formula, and A system for analyzing map-based solar power plant locations, where SSE is the sum of squared residuals, SST is the sum of squared total deviations, n is the number of observations, and p is the number of parameters + 1. Claim 5 In paragraph 4, the solar power generation prediction analysis server comprises the solar power generation prediction value calculated by the correction value ( 1) and the actual amount of power generated (y) collected from nearby solar power plants real Calculate the difference between ); train a second-order correction model by setting the above residual as the dependent variable and multiple parameters including a polynomial or interaction term as independent variables; and the correction value derived from the above second-order correction model ( Recall ) In addition to 1, the final predicted power generation ( A system for analyzing map-based solar power plant sites that calculates ). Claim 6 In paragraph 5, the second correction model comprises, together with actual power generation data obtained from a nearby solar power plant, a monthly or seasonal index, surrounding topographic information (elevation, terrain slope, shading influence), a transformer load factor, a panel efficiency coefficient, a seasonal periodic function including sin(ωt) and cos(ωt), and a squared term (x) for the input variables of the first model. j 2 ) or cross term(x j ×x k A system for analyzing map-based solar power plant locations, characterized by being trained by setting at least one of the following parameters as independent variables. Claim 7 In paragraph 6, the above secondary correction model is, Residual (r) in the form of i Predicting ), where z k,i is the k-th input value, γ0,γ among multiple parameters including raw data, additional variables, and non-linear terms used in the quadratic correction model. k ,γ k,l A system for analyzing map-based solar power plant locations, characterized in that each is a constant term, a first-order term coefficient, and a second-order or cross-term coefficient. Claim 8 In claim 7, the learning of the above-mentioned second correction model is the residual (r i A polynomial expansion vector (z) constructed by setting ) as the dependent variable and the above parameters as independent variables i A step of generating ); a step of estimating γ satisfying the following equation through at least one method among the normal equation, gradient descent, and least squares quenching (LSQ); Residual prediction value i and first-order model prediction 1,i Summing up the final predicted power generation i = 1,i + i A system for analyzing map-based solar power plant locations, characterized by including a step of deriving [the following]. Claim 9 A system for analyzing map-based solar power plant locations, wherein, in claim 8, the solar power generation prediction analysis server re-executes a secondary correction model to update γ whenever actual power generation data continuously measured at a nearby solar power plant is additionally collected, and recalculates the final predicted power generation amount for a map-based candidate point in real-time or periodically using the updated γ. Claim 10 delete

Images