KR-20260063200-A - SYSTEM OF GENERATING MODEL FOR AUTOMATIC IDENTIFICATION OF BALLAST WATER TREATMENT STATUS AND METHOD OF GENERATING MODEL USING THE SAME

KR20260063200AKR 20260063200 AKR20260063200 AKR 20260063200AKR-20260063200-A

Abstract

The present invention discloses a model generation system for automatically identifying a ballast water treatment state and a method for generating a model using the same, wherein the ballast water treatment state can be automatically identified using data on ballast water treatment.

Inventors

오정희
신경순
장풍국
장민철

Assignees

한국해양과학기술원

Dates

Publication Date: 20260507
Application Date: 20241030

Claims (15)

At least part of each step is performed by a processor, and as a method for generating an automatic identification model of ship ballast water treatment status, A step of collecting processing data for processing ballast water based on preset conditions in each of a plurality of vessels; A step of generating a scatter plot for treatment concentration values of ballast water treated based on the above processing data, and defining the ballast water treatment state based on the generated scatter plot; A step of deriving an identification rule for automatically identifying the above-mentioned ship ballast water treatment status; and A step comprising generating a machine learning-based ballast water treatment status identification model using training data created based on the derived identification rules, Method for generating an automatic identification model for ship ballast water treatment status.
In paragraph 1, The above-mentioned collecting step is, A step of generating related data by adding statistical values of the processing data and identification values of the processing status of derived variables to the above processing data; and A step comprising analyzing the characteristics of the above-mentioned processing data and the above-mentioned related data, Method for generating an automatic identification model for ship ballast water treatment status.
In paragraph 2, The method further includes the step of generating the training data labeled with the relevant data, The step of generating the above training data is, A step of dividing the above learning data into training data and validation data; A step comprising training by crossing the separated training data and the verification data n times, Method for generating an automatic identification model for ship ballast water treatment status.
In paragraph 3, The method further includes the step of classifying the above training data by allocating it to 65% to 75% and the above validation data by allocating it to 35% to 25%. Method for generating an automatic identification model for ship ballast water treatment status.
In paragraph 1, The above-mentioned ballast water treatment status identification model is based on the XGBoost algorithm and is a model trained to apply the values of optimal hyperparameters calculated using the GridSearchCV function to the training data to optimize the ballast water treatment status identification model, Method for generating an automatic identification model for ship ballast water treatment status.
In paragraph 1, The step of deriving the above identification rule is, A step comprising reviewing the validity of the derived rule through comparison between the derived identification rule and visual identification, segmented clustering analysis, and hierarchical clustering analysis. Method for generating an automatic identification model for ship ballast water treatment status.
In paragraph 1, The above treatment concentration value is within a preset allowable range, and the method includes a step of calculating the coefficient of variation (CV) of the treatment concentration value to measure the rate of change of the treatment concentration value to determine whether the treatment concentration value changes within the allowable range. The coefficient of variation of the above treatment concentration value is the value obtained by dividing the standard deviation of the above treatment concentration value by the average value of the above treatment concentration value, Method for generating an automatic identification model for ship ballast water treatment status.
In paragraph 1, The above treatment concentration value is within a preset allowable range, and a treatment concentration value fluctuation rate is measured to determine whether the above treatment concentration value deviates from the allowable range, The step of deriving the above identification rule is, A step of deriving a plurality of boundary values to identify the ship ballast water treatment state based on the above treatment concentration value fluctuation rate; and A step that further includes deriving multiple classification rules, Method for generating an automatic identification model for ship ballast water treatment status.
In paragraph 8, The step of deriving the above boundary value is, A step of setting an appropriate treatment concentration value for treating the above-mentioned ship ballast water; The method further includes the step of determining the lower and upper limits of the appropriate operating range of the treatment device for treating the ship ballast water based on the above appropriate treatment concentration value.
In Paragraph 9, It further includes a step of calculating a state distinction boundary value for distinguishing the occurrence rate of an outlier in the treatment concentration value based on the appropriate treatment concentration value, and The above state classification boundary value is calculated such that, taking into account the initial warm-up time of the treatment device for treating the ballast water, the treatment state of the ballast water is within the normal operating range if the rate of occurrence of an outlier in the treatment concentration value is less than 10%, the treatment state of the ballast water is within the unstable operating range if the rate of occurrence of an outlier is between 10% and less than 20%, and the treatment state of the ballast water is within the abnormal operating range if the rate of occurrence of an outlier is 20% or more. Method for generating an automatic identification model for ship ballast water treatment status.
In paragraph 8, The step of deriving the above multiple classification rules is, The method further includes a step of deriving a variation boundary value for distinguishing the variation rate of the treatment concentration value, wherein the variation boundary value determines, considering the influence of abnormal values occurring during initial warm-up operation, that the treatment concentration value is determined to be a normal treatment concentration value if the variation rate of the treatment concentration value is less than 20%, is determined to be an unstable treatment concentration value if the variation rate of the treatment concentration value is between 20% and less than 30%, and is determined to be an abnormal treatment concentration value if the variation rate of the treatment concentration value is 30% or more. Method for generating an automatic identification model for ship ballast water treatment status.
In paragraph 8, A step further comprising generating a classification rule for classifying the above-mentioned ballast water treatment status, Method for generating an automatic identification model for ship ballast water treatment status.
In paragraph 1, After the step of generating the above training data, A step comprising verifying the performance of the ballast water treatment identification model based on the above training data, Method for generating an automatic identification model for ship ballast water treatment status.
As a system for generating a model that automatically identifies the state of ship ballast water treatment, processor; and It includes a memory electrically connected to the processor and storing at least one code executed in the processor, When the above memory is executed through the above processor, the processor Collecting processing data for ballast water treatment from each of multiple vessels, generating a scatter plot for the treatment concentration values of the ballast water based on the processing data, defining the ballast water treatment state based on the generated scatter plot, deriving an identification rule for automatically identifying the ballast water treatment state, and storing codes that cause to generate a machine learning-based ballast water treatment state identification model using training data created based on the derived identification rule. Model generation system for automatically identifying ship ballast water treatment status.
In Paragraph 14, The above-mentioned ballast water treatment status identification model is based on the XGBoost algorithm and is a model trained to apply the values of optimal hyperparameters calculated using the GridSearchCV function to the training data to optimize the ballast water treatment status identification model, Model generation system for automatically identifying ship ballast water treatment status.

Description

System of generating model for automatic identification of ballast water treatment status and method of generating model using the same The present invention relates to a technology that enables the automatic identification of the treatment status of ship ballast water. Ships are widely used as a means of transportation, accounting for more than half of global cargo transport. These ships include ballast tanks for storing ballast water. Ballast water performs the function of regulating the ship's balance and draft by adjusting the amount stored in the ballast tanks according to the volume of cargo loaded on the vessel. Most ballast water is seawater. The process of regulating the amount of ballast water stored in the ballast tanks is achieved by discharging the stored ballast water to the outside of the ship or by supplying seawater located outside the ship to the ballast tanks. For example, when a vessel located in a first area is sailing to load cargo located in a second area, the vessel can control the amount of ballast water by storing seawater located in the first area in a ballast tank. When the vessel reaches the second area, the vessel can control the amount of ballast water by discharging the ballast water stored in the ballast tank to the outside according to the amount of cargo being loaded. As such, when ballast water stored in the ballast tank in the first area is discharged in the second area, inorganic materials such as gravel, sand, and mud located in the first area, as well as organisms such as fish, shellfish, crustaceans, and plankton, and organic materials, are transferred to the second area, causing problems such as ecological disturbance and the spread of harmful pathogens. To prevent this, the International Maritime Organization (IMO) adopted the International Convention for the Control and Management of Ship's Ballast Water and Sediments. By mandating the installation of ballast water treatment systems, the aim is to prevent damage caused by ballast water, such as ecosystem disturbance and the spread of harmful pathogens. To this end, conventional methods have been used to treat ballast water using heat treatment, chemical treatment, and electrolysis. Heat treatment methods have the problem of being difficult to treat heat-resistant microorganisms or inorganic contaminants. Chemical treatment methods have the problem of potential secondary contamination after ballast water treatment. Accordingly, electrolysis methods are widely used recently to perform sterilization treatment on ballast water by electrolyzing a portion of the seawater supplied to the ballast tank and supplying the generated sterilized water back to the seawater. Inspectors at the Port State Control must verify whether ballast water was treated by skilled technicians through various methods in accordance with the implementation procedures of the Ballast Water Management Convention, and must verify the status of the treated ballast water once treatment is complete. However, due to inspection procedures that require multiple steps, it is difficult to quickly verify the status of ballast water treatment, and there is currently no method to consistently check the treatment status of ballast water treated by various methods. The aforementioned background technology is technical information that the inventor possessed for the derivation of the present invention or acquired during the process of deriving the present invention, and it cannot be considered as prior art disclosed to the general public prior to the filing of the present invention. FIG. 1 is a diagram illustrating an environment in which a system for generating a model that automatically identifies the state of ballast water treatment according to an embodiment of the present invention is implemented. FIG. 2 is a diagram illustrating the configuration of a system that generates a model for automatically identifying the state of ballast water treatment according to an embodiment of the present invention. FIGS. 3 and 4 are drawings illustrating an example of analyzing data characteristics of ballast water processed to create a model that automatically identifies the ballast water processing status according to an embodiment of the present invention. FIG. 5 is a diagram illustrating the correlation between the treatment concentration value of the ballast water and the occurrence rate of an outlier in the treatment concentration value to create a model that automatically identifies the ballast water treatment status according to an embodiment of the present invention. FIGS. 6 and 7 are drawings illustrating conditions for deriving identification rules for ballast water treatment status in order to create a model that automatically identifies ballast water treatment status according to an embodiment of the present invention. FIG. 8 is a diagram illustrating the process of generating a model that automatically identifies the state of ballast water treatment according t