Effects of Changes in Bus Supply Level on Urban Rail Demand Forecasting

박성희 서울대학교 대학원 2018 국내박사

The problem of misleading forecasts on the road is less severe and less one-sided than for rail. As the cause of this severe and nonrandom error of rail demand forecasting, this study focuses on the unexpected increase of bus supply level and the modal share prediction errors caused by the unexpected bus supply level change in the process of mode split. If a new rail line is planned for an area where large land developments are expected, the level of bus supply at the time of development planning and level of bus supply at the time of development completion would be different, and this may cause overestimation of rail demand. Therefore, this study developed a model to forecast the bus supply level and suggested a method to apply the model in the urban rail demand forecasting process. To consider the effect of change of future bus supply level, post-processing analysis which re-estimates urban rail demand by using cross-elasticity and the differential rate between bus supply level of existing model and the proposed model. By using the proposed model, four cases of previous urban rail demand forecasting studies were re-reviewed. In Shinbundang-line, Yongin light rail, Gimpo metro line and Byollae-line case studies, the impacts of bus supply level change on the rail demand forecasts ranged from 16% to 41%. In this proposed model, the error was less than 5% and showed high predictability except for the case of Byollae-line. In Byollae-line case, the results of both of the existing model and proposed model showed large error rate to the observed demand. The source of this relatively large error of Byeollae-line case is supposed that several land development plans are neighboring, and buses running through the districts may be overlapped. It is assumed that the overlapping buses cause errors in several districts, as a result, causing relatively large errors. This study suggests the unexpected change of bus supply level as a major source of rail demand forecasting error, which has been failed to be considered in previous studies and guidelines. This study is distinct from previous studies regarding the impact of change of bus supply level on urban rail demand forecast was quantified, and a model to forecast the bus supply level was developed and applied to improve the reliability of demand forecasting.

에너지 수요관리제도에 관한 공법적 연구

박기선 중앙대학교 대학원 2018 국내박사

에너지는 현대사회에서 인간다운 삶을 영위하기 위한 필수적인 재화이다. 에너지공급은 급부행정에서 보장행정으로 임무수행의 방식이 변화하긴 했지만, 에너지의 안정적인 공급은 전형적인 국가의 임무로서 이를 보장하기 위한 국가의 개입이 여전히 행해지고 있다. 그간 국가는 에너지의 공급안정성을 확보하기 위하여 발전소 등 공급시설을 확충하는 공급 중심의 에너지정책을 펼쳐 왔다. 그러나 중앙집중식 대규모 에너지 공급체계는 공급시설의 설치와 관련하여 발생하는 사회적․경제적․환경적 문제로 인하여 한계에 봉착했다. 특히, 2000년대에 이르러 기후변화 대응이 부각되면서 에너지정책의 방향도 에너지 안보뿐만 아니라 효율과 환경 등 다양한 가치를 고려하게 되었다. 이러한 변화는 국가로 하여금 단기적 목표인 에너지공급보다 중장기적 목표인 에너지 수요관리의 중요성을 인식하게 하는 계기가 되었다. 본 논문은 에너지 수요관리가 공급정책의 한계를 보완할 수 있는 새로운 에너지정책으로 주목받았을 뿐, 그 자체가 규범적 논의의 대상으로 여겨지지 않고 있는 현실에 대한 문제제기에서 출발한다. 실제 에너지공급자나 에너지사용자 등 에너지수급과 관련된 이해관계인은 에너지 수요관리를 목적으로 다수의 규제를 받고 있으며, 그 범위 또한 점차 확대되어 가고 있다. 그럼에도 불구하고 에너지 수요관리에 대한 확립된 개념 정의가 없어 어느 범위까지 수요관리 목적의 국가 개입이 가능한지가 불분명한 상태이다. 논문에서는 먼저 에너지 수요관리의 개념을 연료, 열 및 전기를 포함한 에너지의 생산․전환․수송․저장 및 이용 상의 효율향상을 목적으로 하는 일련의 활동으로 정의하여, 수요관리의 직접적 목적이 에너지효율향상에 있음을 강조하였다. 다만, 에너지 수요관리는 가변적인 개념으로 수요관리의 목적이 확대되면 개념도 달라질 수 있으며, 국가의 개입 정도와 그 정당성도 달라질 수 있다. 에너지 수요관리가 에너지공급안정성 확보와 환경보호의 측면에서 헌법적 정당성이 인정된다 하더라고, 실제 수요관리제도의 구현에 있어서는 어느 하나의 가치만을 고려할 수는 없으며 다양한 목표가 상충되지 않도록 이익을 조정하고 목표를 최적화하는 과정이 필요하다. 이러한 관점에서 현행 법제상 에너지 수요관리를 분석하면, 법체계상의 문제점과 수요관리 단계상의 문제점을 도출할 수 있다. 에너지 수요관리는 에너지 공급과 수요와의 관계 하에서 상호 유기적으로 작용하기 때문에 일정한 체계 하에서 이루어질 필요가 있다. 그러므로 에너지기본법의 제정을 통하여 에너지정책의 연계성을 강화하고, 에너지계획의 수립에 있어 국민 참여 절차와 수요 예측의 정확성을 담보하기 위한 기반조성 등 수요관리를 위한 법적 기반이 마련되어야 한다. 이와 더불어 부문별로 시행되고 있는 에너지 수요관리제도의 개선도 병행되어야 한다. 이때 각 부문별 에너지소비특성을 바탕으로 개별 제도의 도입 취지나 유사 제도와의 관련성 등을 고려하여 제도 간의 유기성을 확보해야 한다. 또한 에너지 수요관리제도는 점차 본래적 기능인 에너지효율향상을 넘어 환경적 요소를 함께 고려해야 하는 결합된 영역으로 변화하고 있다. 에너지 수요관리가 사회․경제 전반에 미치는 영향을 고려하면, 장기적으로는 에너지 수요관리에 관한 통합적인 정책이 필요하다. 국가는 에너지의 안정적인 공급을 위하여 에너지 수요관리제도를 시행하고 있다. 에너지효율향상은 에너지절감에 기여하지만 국민의 자발적 의사만으로는 한계가 있기 때문에 국가의 규제가 이루어지고 있다. 그러나 기후변화대응이 인류공동의 과제로 부상하면서 현행 에너지 수요관리제도는 본래적 기능과는 별개로 점차 규제가 강화되고 있는 추세이다. 그러한 점에서 본 논문은 에너지 수요관리의 개념 정립을 통해 수요관리의 목적으로 이루어지는 규제를 체계적으로 조망하고, 그 규제가 정당화되는 범위를 명확히 하여 효과적인 수요관리를 위한 법적 기반을 마련하는 데 기여할 수 있을 것이다. Energy is an indispensable commodity for human life in modern society. The supply of energy has changed in the way of performance by government from beneficial administration to guaranteeing administration these days. The stable supply of energy is a typical national task, so the state is still intervening to ensure this. In the meantime, the state has implemented a supply-oriented energy policy expanding supply facilities such as power plants to secure energy supply stability. However, the centralized large-scale energy supply system reached the limit by the social, economic and environmental problems that arise in connection with the installation of the supply facilities. Especially, in the 2000s, as the response to climate change began to emerge, the direction of energy policy also became to include various values ​​such as efficiency and environment as well as energy security. The nations aware of the importance of energy demand management, a medium- and long-term goal, rather than a short-term goal of energy supply from these changes. This study starts from the question of the reality that energy demand management has attracted attention as a new energy policy overcoming the limitation of supply policy, but is not regarded as a subject of normative debate. As a matter of fact, Interested parties related to energy supply and demand, such as energy suppliers and energy users, are subject to a number of regulations for the purpose of managing energy demand, and the scope is also gradually increasing. Nevertheless, there is no established definition of concept for energy demand management, so it is unclear to what extent it is possible for the state to intervene for demand management purposes. This paper first defines the concept of energy demand management as a series of activities to aim at improving the efficiency of production, conversion, transportation, storage and utilization of energy, including fuel, heat and electricity, and emphasizes the fact that the direct purpose of energy demand management is an improvement of energy efficiency. However, the concept of energy demand management is a variable concept. If the purpose of demand management expands, the concept can be changed. Also, the level of intervention and legitimacy of the state can be changed. Even though the energy demand management is considered to have constitutional legitimacy in terms of ensuring energy supply stability and environmental protection, we can not consider only one value in the implementation of actual demand management system. So, It is necessary to adjust the profits and optimize the goals so that the various goals do not conflict. From this point of view, analyzing energy demand management in the current legal system can lead to problems in the legal system and problems in the demand management stage. Energy demand management needs to be done under a certain system because it interacts organically with energy supply and demand. Therefore, through the enactment of the Framework Act on Energy, a legal basis for demand management should be established by strengthening the linkage of energy policies and building the foundation for ensuring the accuracy of the public participation procedures and demand forecasts in establishing energy plans. In addition, the improvement of the energy demand management system implemented by each sector is also required. At this time, based on the energy consumption characteristics of each sector, it is necessary to secure the organicity between the systems considering the intention of introducing the individual system and the relation with similar systems. In addition, energy demand management systems, beyond the original function of improving energy efficiency, are changing into a combined area where environmental factors need to be taken into consideration. Giving consideration to the impact of energy demand management on society and the economy as a whole, an integrated policy on energy demand management is needed in the long run. The state has implemented an energy demand management system for stable supply of energy. Energy efficiency enhancement contributes to energy saving, but is regulated by state because the voluntary intention of the people has its limit. However, as the response to climate change has emerged as a common problem for mankind, current energy demand management system is gradually becoming more regulated, apart from its original functions. In this regard, this paper can contribute to give a legal basis for effective demand management by systematically reviewing the regulations for demand management through establishing the concept of energy demand management, and clarifying the extent to which the regulation is justified.

Causal Inference of Demand-Responsive Transit Implementation and Bridging Technology Adoption in Aging Societies

오다인 아주대학교 일반대학원 2026 국내석사

Aging societies worldwide face mounting challenges in maintaining accessible public transportation, particularly in rural areas where demographic decline and dispersed populations render conventional fixed-route transit (FRT) economically unsustainable. Demand-responsive transit (DRT) has emerged as a flexible solution to improve mobility in such contexts. However, two critical gaps remain: First, how DRT introduction reshapes the overall public transportation system—comprising both DRT and FRT—has not been sufficiently explored. Second, while DRT promises enhanced accessibility, its reliance on smartphone-based booking creates digital barriers that paradoxically exclude older adults with limited digital literacy, the very population most dependent on public transportation. This dissertation addresses these gaps through a sequential evaluation framework that examines two interconnected interventions in rural Cheongju, South Korea. The first study analyzes how DRT implementation reshapes system-wide Public Transit (PT) demand patterns using causal impact inference based on Bayesian Structural Time Series (BSTS) models. The analysis employs residential clustering to construct spatial units that reflect realistic transit catchment areas, addressing the modifiable areal unit problem that biases conventional spatial analyses. Results indicate that approximately half of residents in the service area experienced significant increases in total PT usage of around 30%. DRT adoption was more beneficial in areas with lower population density, lower proportions of elderly residents, lower existing bus usage rates, and higher bus stop density. These findings suggest that DRT functions primarily as a complement rather than substitute for FRT, generating new demand through induced trips and modal shifts from private vehicles. Building on this DRT-enabled transportation system, the second study evaluates a bridging technology intervention designed to overcome digital access barriers: physical call buttons installed in senior centers that enable older adults to request DRT service through direct connection with customer service operators. Employing the same BSTS-based causal inference methodology, the analysis reveals that approximately 80% of service areas experienced significant increases in DRT trips following call button introduction. This growth reflects two mechanisms: increased use of both DRT and FRT services (indicating complementary effects), and additional DRT trips generated through modal shifts from non-public transit or newly induced demand. DRT trip growth was significantly associated with senior center density and distance to medical facilities, demonstrating that bridging technology effectiveness depends on strategic placement within social infrastructure networks frequented by target populations. The integrated findings demonstrate that inclusive rural mobility requires multi-layered interventions addressing both service flexibility (DRT) and access equity (bridging technologies). The sequential analysis reveals synergistic effects: call buttons proved effective precisely because DRT had established service reliability, while DRT reached its full accessibility potential only after call buttons removed digital barriers. Methodologically, the dissertation advances transportation research by developing a causal inference framework for evaluating sequential policy interventions and demonstrating how spatial analysis methods can account for the nested structure of accessibility improvements. These findings provide actionable insights for transportation planners in aging societies worldwide, emphasizing that technology-enabled transit innovations must be coupled with inclusive access mechanisms to serve all residents, particularly those least equipped to navigate digital interfaces.

Demand Response in Smart Grid: A Stackelberg Game-Theoretic Approach : 스마트 그리드에서 스타클버그 게임 이론을 이용한 수요반응

MengmengYu 한양대학교 2015 국내박사

Smart grid is envisioned as the next generation of traditional electric power grid, which enables a two-way flow of electricity and information between power utility side and user side. Demand-response (DR) program is regarded as a promising solution for the future power grid by encouraging users to consume energy in a wise way so as to benefit both the power utility and users. In view of the requirements of devising efficient DR schemes for realizing smart grid, this dissertation develops two novel DR models under the framework of Stackelberg game theory, which are designed for the user side and power utility side respectively. First, a real-time price-based demand-response algorithm was proposed for achieving optimal load control of devices in a facility by forming a virtual electricity-trading process, where the Energy Management Center of the facility is the virtual retailer (leader) offering virtual retail prices, from which devices (followers) are supposed to purchase energy. A 1-leader, N-follower Stackelberg game is formulated to capture the interactions between them, and optimization problems are formed for each player to help select the optimal strategy. The existence of a unique Stackelberg equilibrium that provides optimal energy demands for each device was demonstrated. The simulation analysis showed that the Stackelberg game-based demand-response algorithm is effective for achieving the optimal load control of devices in response to real-time price changes with a trivial computation burden. Second, a novel demand-response model was presented for modeling the electricity trading process between one utility company (UC) and multiple users aimed at balancing supply and demand as well as flattening aggregated loads in the system. The interactions between the UC (leader) and users (followers) are formulated into a 1-leader, N-follower Stackelberg game, where minimization problems are defined for the utility company and each user with the target of minimizing the generation fluctuations and the costs of users. A pricing function is adopted for regulating real-time prices (RTP) at the utility company side, which then act as a coordinator, inducing users to join the game. An iterative algorithm is proposed to derive the Stackelberg equilibrium, through which optimal power generation and power demands are determined for the UC and users respectively. Numerical results indicate that the proposed method can efficiently reshape users’ demands, including flattening peak demands and filling the vacancy of valley demands, and significantly reduce the mismatch between supply and demand. 스마트 그리드에서 스타클버그 게임 이론을 이용한 수요반응 스마트 그리드는 전기사업소와 사용자간의 전력과 정보의 양방향 흐름을 실현하였고, 기존 전력 그리드의 차세대 기술로 발전 중이다. 수요관리 (DR) 프로그램은사용자들의 현명한 에너지 소비를 조장하며, 미래 파워 그리드의 유망한 해결책으로 간주된다. 효율적인 DR방식을 실현하기 위해, 본 논문에서는 스타클버그 게임 이론을 활용한 두 가지 DR 모델들을 소개한다. 첫째, 설비장비들을 조절하는데 사용되는 실시간 가격 기반 DR 알고리즘을 제안한다. 이 알고리즘은 가상의 전력교역과정을 구성한다. 해당 시설의 에너지관리센터는 가상의 중간상인(리더)으로써 에너지를 가상의 가격에 제공하며, 각 장비들(팔로어)은 제공되는 에너지를 구매한다. 1-리더, N-팔로어 스타클버그 게임은 이들간의 상호작용을 표현하였으며, 각 참가자들의 최적화 전략을 통해 최적문제들을 구성하였다. 각 장비에 대한 최적의 에너지 수요를 구하는 스타클버그 평형의 단일 해법을 구할 수 있음을 보였다. 시뮬레이션 분석은 스타클버그 게임 기반 DR 알고리즘의 효율성을 보여준다. 실시간 가격이 제공될 때, 각 장비들을 조작하는 최적의 수치들을 약간의 계산을 통해 구할 수 있다. 둘째, 한 개의 전기사업소(UC)와 다수의 사용자들 사이의 전력교역과정을 모델링하는 수요관리 프로그램을 제안한다. 이들의 목표는 수요와 공급의 균형을 맞추며, 전체 수요를 최소화하는 것이다. UC(리더)와 사용자들(팔로어) 사이의 상호작용은 1-리더, N-팔로어 스타클버그 게임을 통해 구성되었으며, 발전 변화량과 사용자들의 비용을 최소화 문제들은 UC와 각 사용자들을 위해 정의되었다. 전기사업소에서 가격 함수는 실시간 가격을 만들었으며, 이 함수는 사용자와의 조정자 역할을 한다. 스타클버그 평형을 찾을 수 있는 반복 알고리즘이 제안되었으며, 이를 통해 최적의 전력 생산과 수요가 결정된다. 제안하는 방식은 사용자의 수요를 변화시켜 최고수요가 낮아졌으며, 최저수요가 높아졌고, 수요와 공급 사이의 격차가 줄어들었음을 시뮬레이션 결과들을 통해 증명하였다.

(An) optimization model of on-demand mobility services with spatial heterogeneity in travel demand

박준수 Graduate School, Yonsei University 2022 국내박사

On-demand mobility (ODM) services are those that can be called when needed, such as the flexibility of personal vehicles. Smartphone booking, easy payment options, and real-time demand-supply matching have become possible due to the development of digital communication technologies, and call-based ODM services (e.g., Uber, Lift, and Via) have been introduced in many cities worldwide and have been growing rapidly. The emergence of ODM services has triggered a major change in urban transportation systems; they can provide stable and high-quality services at a more competitive cost. The ODM services have more advantages than public transportation with fixed routes, e.g., improved accessibility, lower waiting and travel times. Therefore, it is expected that the services can reduce the number of personal vehicles that spend most of their time in parking. Meanwhile, some negative impacts are also anticipated. For instance, the operation of an excessive number of vehicles increases the number of vehicles occupying the road, and may rather worsen traffic congestion. Therefore, designing level-of-service and fleet sizes is an important challenge from this perspective. Studies on free sizing of ODM services or similar services have been largely conducted by modeling and micro-simulation approaches. Studies using micro-simulation approaches provide specific and realistic results; however, it is challenging to generalize them because it has been analyzed for specific regions or cases. Moreover, they cannot predict optimal fleet sizes. Studies using network modeling methods very effectively reflect demand patterns and predict fleet sizes. However, the zone is treated as a point rather than a space; there has a limitation in providing level-of-service (e.g., customer waiting time). Daganzo and Ouyang (2019) proposed an analytical model that can be generally used to overcome these limitations. Their model predicted customer waiting time or user travel time, and expressed required fleet sizes as simple formulas, unlike the network model. The model assumes that trip origins and destinations are uniformly distributed in space. However, they are generally not uniformly distributed in the real world. Therefore, it is necessary to pay attention while applying their model. In other words, the model has limitations because it ignores the spatial heterogeneity of travel demand. This dissertation proposes an optimization model that spatially expands the general model to overcome these limitations. Specifically, we aim to develop an optimization model that can predict the required fleet size of ODM services at a given demand pattern and at various service levels. The demand-supply imbalance problem that occurs while relaxing the assumptions was solved through vehicle relocation. Vehicle relocation was reformulated into a simple linear programming problem, and the objective function was set to minimize the fleet size. Spatial heterogeneity indicators were defined to more effectively explain the effect of spatial heterogeneity of travel demand on the required fleet size. Spatial heterogeneity indicators consisted of travel distance indicator and relocation indicator . They were derived mathematically by considering that each indicator affected the required number of vehicles. The sensitivity analysis of the fleet size by the demand density and spatial heterogeneity indicators was performed, The results showed that the larger the spatial heterogeneity indicators, the larger the required fleet size, and the lower the demand, the greater the effect of improving the service level by adding vehicles. Optimization of ODM service was analyzed for fleet size minimization and cost minimization. The results showed that required fleet size could be underestimated or overestimated if the general model was applied to regions with spatial heterogeneity of travel demand. Furthermore, it showed that cost minimization requires more vehicles than fleet size minimization. If the ODM service is operated as driverless vehicles, cost minimization may be an advantageous goal. Additionally, when both goals are considered simultaneously, a reasonable fleet size appears in the range of two values. The agent-based simulation was performed to verify the model. The difference in average waiting time between the estimated and simulated values was only 8 sec; the extended model was fairly accurate. The proposed model can be applied regardless of the demand pattern and can be used by governments or agencies to design policy or operationally-optimized services. Furthermore, the model forms the basis for the practical implementation of planned services. 온디맨드 모빌리티 서비스는 필요할 때 호출할 수 있는 교통 서비스로, 개인 소유의 차량과 같이 유연한 사용이 가능하다는 장점이 있다. 디지털 통신 기술의 발달로 스마트폰 예약과 간편결제 옵션, 실시간 수요 공급 간 매칭이 가능해졌으며, 이로 인해 통신기술에 기반한 온디맨드 모빌리티 서비스(우버, 리프트, 비아 등)는 세계 여러 도시에 도입되고 빠르게 성장하고 있다. 온디맨드 모빌리티 서비스의 등장은 도시 교통시스템의 주요 변화를 촉발하고 있다. 기존의 대중교통 서비스와 비교하여 고객에게 더 높은 수준의 서비스(접근성 향상, 낮은 대기시간과 이동시간)를 제공할 수 있다는 점에서 경쟁력을 가지고 있으나, 과도한 차량 수의 운영은 도로를 점유하는 차량의 수를 증가시키며, 오히려 교통 혼잡을 악화시킬 수 있다. 실제로 뉴욕, 런던 등에서 운영되고 있는 우버 서비스는 도입 이후, 무분별한 서비스 공급자의 증가로 인해 도심의 차량을 크게 증가시켰다. 이러한 관점에서 온디맨드 모빌리티 서비스의 서비스 수준 및 적절한 차량 규모의 설계는 매우 중요한 문제이다. 온디맨드 모빌리티 서비스 또는 이와 유사한 서비스의 차량 규모 산정에 대한 연구는 주로 모델링 및 시뮬레이션을 통해 수행되었다. 시뮬레이션을 사용한 연구들은 구체적이고 현실적인 결과를 보여주지만, 특정 지역이나 사례에 대해서 분석하였기 때문에 그 결과들을 일반화하기에는 어려움이 있다. 네트워크 모델링을 사용한 연구들은 수요 패턴을 매우 효과적으로 반영하고 차량 규모를 예측할 수 있다. 하지만, 네트워크 모델 특성상 존(Zone)을 하나의 점으로 취급하고 모든 수요의 발생과 차량의 공급은 점에서 이루어지기 때문에 현실적인 서비스 수준(고객 대기시간)을 예측할 수 없다. Daganzo 와 Ouyang (2019) 은 이러한 한계를 극복하기 위해 광범위하게 사용될 수 있는 일반 모델을 제안하였다. 일반 모델은 서비스 운영 시스템을 간단한 대기열 네트워크로 표현하며, 안정된 상태에서 시스템의 평균 고객 대기시간과 차량 규모와의 관계를 간단한 공식으로 나타낸다. 이 모델은 수요의 출발지와 목적지가 공간에 균일하게 분포되어 있다고 가정하지만, 현실에서는 일반적으로 불균일하게 분포하기 때문에 모델 적용에 있어 주의가 필요하다. 본 연구에서는 이러한 한계를 극복하기 위해 일반 모델을 공간적으로 확장한 확장 모델을 제안하였다. 다시 말해, 주어진 수요 패턴과 다양한 서비스 수준에서 온디맨드 모빌리티 서비스의 필요한 차량 규모를 예측할 수 있는 최적화 모델을 개발하였다. 기존 가정을 완화하면서 발생하는 수요-공급 불균형 문제는 차량의 재배치를 통해 해결하였으며, 차량 재배치는 차량 규모를 최소화하는 목적 함수를 갖는 선형 계획법 문제로 재구성하였다. 수요의 공간적 이질성이 필요한 차량 규모에 미치는 영향을 보다 효과적으로 설명하기 위해 공간적 이질성 지표를 정의하였다. 공간적 이질성 지표는 이동 거리 지표와 재배치 지표로 구성된다. 각 지표가 필요한 차량 수에 영향을 공식으로 도출하였으며, 수요밀도와 공간적 이질성 지표에 따른 차량 규모의 민감도 분석을 수행하였다. 그 결과, 이질성 지표가 클수록 필요한 차량 규모가 크며, 수요가 적을수록 차량 추가로 인한 서비스 수준의 향상 효과가 커지는 것으로 분석되었다. 온디맨드 모빌리티 서비스의 최적화는 두 가지 목표(차량 규모 최소화 및 비용 최소화)에 따라 각각 분석하였다. 차량 규모 최소화 목표하에서 일반 모델과 확장 모델의 최적 차량 규모를 비교한 결과, 수요의 공간적 이질성이 있는 지역에 일반 모델을 적용하면 필요한 차량 규모가 과소 또는 과대 추정될 수 있음을 확인하였다. 또한, 비용 최소화를 달성하기 위해서는 차량 규모 최소화보다 항상 더 많은 차량이 필요함을 수식적으로 증명하였다. 온디맨드 모빌리티 서비스가 무인 차량으로 운영된다면 비용 최소화가 유리한 목표가 될 수 있음을 확인하였으며, 두 목표를 동시에 고려할 때 합리적인 차량 규모는 두 값의 범위로 나타내어진다. 모델 검증을 위해 에이전트 기반 시뮬레이션을 수행하였다. 모델 예측 값과 시뮬레이션 값 사이의 평균 대기시간 차이는 8초에 불과하였으며, 확장 모델이 수요 패턴을 정확히 반영하고, 강력한 예측력을 가지고 있음을 확인하였다. 제안된 모델은 서비스 지역이 어떠한 수요 패턴을 갖더라도 제약 없이 적용할 수 있는 모델로서 정부 또는 기관에서 온디맨드 모빌리티 서비스를 설계하는 데 사용할 수 있다. 또한, 계획된 서비스의 실질적인 구현을 위한 기초 모델로 사용될 수 있다.

Game Theory Based Demand Side Management to Reduce Peak-to-Average Ratio in Smart Grid : Game Theory Based Demand Side Management to Reduce Peak-to-Average Ratio in Smart Grid

Nguyen Khanh Hung 경희대학교 2012 국내석사

The smart grid, regarded the next generation power gird, uses two-way ows of electricity and information between energy consumers and providers to create a widely distributed automated energy delivery network [1]. By utilizing modern information technologies, the smart grid enables for demand side management (DSM) and provides more exibility in demand shaping. DSM is one of the key components of the future smart grid, which is implemented to control energy consumption at the customer side. Traditionally, in order to satisfy all energy demands from customers, the grid capacity needs to be designed to satisfy the peak power demand rather than the average power demand. This usually requires installation of new power generation and transmission infrastructure. To overcome this problem, DSM programs have been proposed to control the energy consumption pattern of the users with the aim of reducing peak demands [2{6]. In this thesis, rstly, we propose a novel DSM technique to reduce the peak load of the power system. We consider a smart power system with distributed users that request their energy demands from an energy provider who dynamically updates energy prices based on the load proles of the users. Each user is assumed to have a backup battery. The users try to minimize the Peak-to-Average Ratio (PAR) or Square Euclidean Distance between the instantaneous energy demand and the average demand of the power system by jointly scheduling their appliances and controlling the charging process for their backup batteries. Batteries will be charged during low-demand periods with a low energy cost and discharged during high-demand periods. We apply game theory to formulate the energy consumption scheduling game for the distributed design, in which the players are the users and their strategies are the energy consumption schedules for appliances and backup batteries. Based on the game theory setup, we also propose a distributed demand side management algorithm in which each user tries to minimize its total energy cost. The proposed distributed algorithm requires each user to exchange its load prole information with and to receive the price signal from the energy provider. Secondly, we consider a smart charging and discharging process for multiple Plug-in Hybrid Electric Vehicles (PHEVs) in a garage building to optimize the energy consumption prole of the building. the PHEVs try to minimize Square Euclidean Distance between the instantaneous energy demand and the average demand of the building by controlling the charging and discharging schedules for their batteries. We also propose a distributed algorithm in which each PHEVs tries to minimize its energy charging cost to achieve the best performance.

자산거래가 화폐수요에 미치는 영향에 관한 연구

백창현 한국해양대학교 대학원 2023 국내박사

Since the 1990’s, with the progress of globalization, financial regulations have been eased, and with the development of ICT it has quickly been combined with a financial industry. In the process, the various types of financial products with different characteristics from existing financial products emerged, and the financial policies of the governments based on stable determinants which lay foundation on the existing money demand theory, could not effectively control liquidity and credit size. Therefore, the researches were activated to establish a new money demand function by finding the determinants of stable money demand. Due to changes in the financial environment, this paper established two types of extended money demand functions. one is the asset effect estimation model which adds asset effect factors (stock prices, real estate prices, and exchange rates) to the money demand function including real income and interest rates, which are the basic factors. The other one is the uncertainty effect estimation model which adds the uncertainty effect factors (stock prices volatility and exchange rates volatility) to the basic money demand function. To confirm whether a stable relationship between these determinants and long-term and short-term demand for money is established, It was divided into the period of the Korean foreign exchange crisis (1991-2006) and the period of the international financial crisis (2007-2021), and the empirical analysis was performed using GLS method including the entire period (1991-2021), respectively. The estimation results of the asset effect estimation model and the uncertainty effect estimation model are summarized as follows. First, regardless of the monetary indicators M1, M2 and Lf in the estimation results for the entire analysis period, in most of the estimatin results by the asset effect estimation and uncertainty effect estimation models, real income had a statistically significant effect on the long and short-term money demand in the positive (+) direction and the interest rate in the negative (-) direction, as predicted by the liquidity preference theory for real income and interest rate, which are basic factors. Second, comparing the period of the Korean financial crisis and the period of the international financial crisis, the asset effect estimation model and the uncertainty effect estimation model showed greater influence on real income and interest rates during the Korean financial crisis than during the international financial crisis. However, during the international financial crisis, interest rates have some influence on short-term money demand, but they do not affect long-term money demand, showing an unstable relationship between interest rates and money demand. Third, as a result of estimation for the entire analysis period in the asset effect estimation model, stock prices had a negative (-) effect on long-term and short-term currency demand, resulting in a greater substitution effect. In general, it was found that its influence was greater during the international financial crisis. Real estate prices were found to have a greater income effect as they had a positive (+) effect on long-term and short-term money demand. In general, it can be seen that its influence increased during the international financial crisis. The exchange rate had a positive (+) effect on mid- and long-term money demand, indicating that the income effect was greater. During the Korean financial crisis, the income effect is greater in the positive (+) direction, and during the international financial crisis, the substitution effect is greater in the negative (-) direction, indicating that the fluctuation is quite large. Fourth, as a result of estimation for the entire analysis period in the uncertainty effect model, real estate prices have a positive (+) effect on all long-term and short-term money demand, resulting in a greater income effect. However, it had a greater influence on short-term money demand than long-term. It was found that real estate prices affected short-term money demand in the positive (+) direction during the Korean foreign exchange crisis, and short- and medium-term currency demand during the international financial crisis. Fifth, as a result of estimation for the entire analysis period in the uncertainty effect model, stock price volatility was found to have a negative (-) effect on mid- and long-term money demand. During the international financial crisis, it was found that it had a negative (-) effect on mid- and long-term money demand. Sixth, as a result of estimating exchange rate volatility for the entire analysis period in the uncertainty effect model, it did not have a statistically significant effect on long-term and short-term money demand. As a result of empirical analysis, Korea's long-term and short-term demand for money in 1990 is greatly affected by asset effects such as stock prices, real estate prices, and exchange rates as well as real income and interest rates, which are basic factors. On the other hand, stock price volatility affects the uncertainty effect, but it was not so big. Therefore, when implementing government policies, it is suggested that policies including asset price variables such as stock prices, real estate prices and exchange rates, and uncertainty factors such as stock price volatility should be devised, not just based on real income and interest rates. 1990년대 이후 세계화 진전에 따라 금융규제가 완화되고 정보통신기술이 발달하면서 빠르게 금융 산업과 결합되었다. 이 과정에서 기존 금융상품과 특성이 다른 다양한 형태의 금융상품들이 등장하면서 기존 화폐수요이론에 근거를 둔 안정적 결정요인에 입각한 정부의 금융정책은 유동성과 신용규모를 효과적으로 통제할 수 없었다. 따라서 안정적인 화폐수요의 결정요인을 찾아 새로운 화폐수요함수를 구축하려는 연구가 활성화되었다. 본 논문은 금융환경변화로 인해 기본요인인 실질소득과 이자율을 포함한 화폐수요함수에 자산효과 요인(주가, 부동산가격, 환율)를 추가한 추정모형(자산효과 추정모형)과 불확실성효과 요인(주가변동성, 환율변동성)을 추가한 추정모형(불확실성효과 추정모형) 등 2가지 형태의 확장 화폐수요함수를 설정하였다. 이들 결정요인과 장·단기 화폐수요 간의 안정적 관계 성립여부를 확인하기 위해 한국외환위기 기간(1993-2006)과 국제금융위기 기간(2007-2021)으로 나누고, 또한 전 기간(1993-2021)에 대해 일반최소자승추정법을 사용하여 실증 분석하였다. 자산효과 추정모형과 불확실성효과 추정모형의 추정결과를 정리하면 다음과 같다. 첫째, 전 분석기간을 대상으로 한 추정결과에서 통화지표 M1, M2, Lf와 관계없이 자산효과 추정모형과 불확실성효과 추정모형 대부분에서 기본요인인 실질소득과 이자율은 유동성선호설이 예측한 대로 실질소득은 양(+)의 방향으로, 이자율은 음(-)의 방향으로 통계적으로 유의미하게 영향을 미쳤다. 둘째, 한국외환위기 시기에 자산효과 추정모형과 불확실성효과 추정모형에서 한국외환위기 시기에 실질소득과 이자율의 영향력이 장·단기 화폐수요에 더 크게 나타났다. 그러나 이자율의 단기 화폐수요에 대한 영향력은 어느 정도 존재하나 장기 화폐수요에 대한 영향은 없었다. 셋째, 자산효과 추정모형에서 전 분석기간을 대상으로 한 추정결과 주가는 장·단기 화폐수요에 음(-)의 방향으로 영향을 미쳐 대체효과가 더 큰 것으로 나타났다. 대체적으로 국제금융위기 시기에 그 영향력이 더 큰 것으로 나타났다. 부동산가격은 장·단기 화폐수요에 양(+)의 방향으로 영향을 미쳐 소득효과가 더 큰 것으로 나타났다. 대체로 국제금융위기 시기에 그 영향력이 더 커진 것을 알 수 있다. 환율은 중·장기 화폐수요에 대해 양(+)의 방향으로 영향을 미쳐 소득효과가 더 큰 것으로 나타났다. 한국외환위기 시기에는 소득효과가 더 커서 양(+)의 방향으로, 국제금융위기 시기에는 대체효과가 음(-)의 방향으로 영향을 미쳐 그 변동 폭이 상당히 큰 것을 알 수 있다. 넷째, 불확실성 효과모형에서 전 분석기간을 대상으로 한 추정결과 부동산가격은 모든 장·단기 화폐수요에 양(+)의 방향으로 영향을 미쳐 소득효과가 더 큰 것으로 나타났다. 그러나 장기보다는 단기 화폐수요에 대한 영향력이 더 컸다. 한국외환위기 시기에 부동산가격은 단기 화폐수요에 양(+)의 방향으로 영향을, 국제금융위기 시기에는 중·단기 화폐수요에 영향을 미친 것으로 나타났다. 다섯째, 불확실성효과 모형에서 전 분석기간을 대상으로 한 추정결과 주가변동성은 중·장기 화폐수요에 대해 음(-)의 방향으로 영향을 미친 것으로 나타났다. 국제금융위기 시기에도 중·장기 화폐수요에 대해 음(-)의 방향으로 영향을 미치는 것으로 나타났다. 여섯째, 불확실성효과 모형에서 전 분석기간 대상으로 한 환율변동성을 추정결과 장·단기 화폐수요에 통계적으로 유의미한 영향을 미치지 못했다. 실증분석 결과 1990년대 한국의 장·단기 화폐수요는 기본요인인 실질소득과 이자율뿐만 아니라 주가, 부동산가격과 환율 등 자산효과에 크게 장기 안정적인 영향을 받고 있다. 반면 불확실성효과로 주가변동성이 영향을 미치지만 그렇게 크지 않았다. 따라서 정부정책을 실시할 때 기존의 기본요인인 실질소득과 이자율에만 기반을 두지 않고 주가, 부동산가격과 환율 등 자산가격변수와 주가변동성 등 불확실성 요인을 포함한 정책을 입안해야한다는 시사점을 얻는다.

발달지체유아 부모의 스트레스와 사회적 욕구에 관한 연구

조영숙 대구대학교 교육대학원 2007 국내석사

The purpose of this thesis is to prepare plans for social support through inspecting stress degree of parents nurturing children with developmental delayed To fulfill the purpose, it was studied with subjects of 210 parents of children with developmental delayed using 10 welfare center for the handicapped and the results are as follows; First, the result of difference verification about nurturing stress along financial level of parents factors is that they appeared higher stress in low than in middle as for all stress, character of children and stress between parents and children. The result of difference verification about social demands was that financial degree in the information demand appeared higher in low than in middle. Second, mental retardation appeared higher than retardation disorder and emotional disturbance appeared higher than retardation disorder, in difference verification along nurturing stress along children factors. As for character of children, mental retardation appeared higher than retardation disorder and developmental delayed appeared higher than retardation disorder. In difference verification along disorder type as for social demands along children factors, mental retardation showed higher than retardation disorder in social demand and mental retardation was higher than retardation disorder in financial demand and mental retardation appeared higher than developmental delayed and developmental delayed appeared higher than retardation disorder in total demands. As significant difference along level of disorder concerning family, social demands and total demands revealed, the more disorder was found in midium disorder, and it was found that, as severe as their disorder is, their demands were high. Third, character of children appeared as the highest as for relation among nurturing stress and demand for finance and demand for protection appeared as the highest as for relation among social demands. Also, family demand and character of children appeared as the highest as for relation between nurturing stress and social demands.

Sustainable Electricity Demand Forecasting and Risk Management under Transformative Technologies

정원영 명지대학교 대학원 2026 국내박사

전 세계 전력 수요는 대전환 기술의 확산과 함께 구조적 변화를 겪고 있다. 특히 AI 데이터센터와 전기차(EV) 도입의 가속화는 전력 소비의 규모뿐 아니라 시간적·공간적 분포까지 바꾸고 있으며, 이로 인해 기존의 안정성을 전제로 한 수요–공급 계획 체계가 점차 한계에 직면하고 있다. AI 데이터센터는 고밀도의 지속적 기반부하를 형성하고, EV 충전은 특정 시공간에 집중되는 부하를 만들어냄으로써 장기적인 예측 불확실성과 시스템 리스크를 크게 증가시키고 있다. 이러한 변화는 단순한 수요 증가를 넘어 전력 설비 포화, 국지적 정체, 대규모 정전과 같은 재난 수준의 위험으로 확장될 수 있기 때문에, 기존의 단일 예측값 중심 접근을 넘어 확률적, 구조적 관점의 수요 예측이 요구된다. 본 연구는 이러한 문제의식을 바탕으로, 대전환 기술이 야기하는 전력 수요 변화를 독립적으로 파악할 수 있는 모듈형 예측 체계를 설계하였다. 전통적 요인(GDP, 인구 등)으로 설명되는 기저 수요와 AI, EV 확산에 따른 기술기반 수요를 구조적으로 분리함으로써, 기술 확산이 전력 수요의 수준과 변동성을 어떻게 변화시키는지를 분명하게 계량화하였다. 이를 토대로 ARIMAX와 Prophet 모델을 활용하여 2025년부터 2050년까지의 미국 전력 수요를 예측하였으며, 두 모델의 추세 안정성, 외생 변수 민감도, 충격 반응 특성을 비교함으로써 기술 확산 환경에서의 모델별 활용 가능성을 규명하였다. 예측의 불확실성을 정량화하기 위해 몬테카를로 기반 확률적 예측 절차를 도입하였다. 이 접근은 단일 예측값이 아닌 잠재적 수요 범위 전체를 분석함으로써 장기 예측에서 필연적으로 증가하는 변동성과 위험도를 보다 현실성 있게 해석할 수 있도록 한다. 본 연구는 또한 생성형 AI의 대표 기법인 확산(Diffusion) 모델을 장기 전력 수요 예측에 적용하여, 비선형적 불확실성 누적과 비대칭적 위험 구조를 반영하는 확률 분포 기반의 예측 결과를 제시하였다. 이는 기존 연구들이 단기 부하 예측에 국한된 것과 달리, 대전환 기술 요인을 조건 변수로 통합하고 장기 예측에 적용한 최초의 사례라는 점에서 학술적 의의를 갖는다. ARIMAX, Prophet, Diffusion 모델을 동일한 외생변수 및 동일 기간 조건에서 비교한 분석 또한 본 연구의 중요한 특징이다. 이를 통해 각 모델이 제공하는 설명력, 추세 안정성, 확률적 적합도 등 상이한 장점을 구조적으로 파악할 수 있었고, 장기, 고불확실성 환경에서 어떤 접근법이 어떤 정책·운영 목적에 적합한지 판단할 수 있는 근거를 마련하였다. 이러한 통합적 분석은 향후 전력계획에서 특정 모델의 우월성을 주장하기보다는, 목적 기반 모델 선택과 위험 기반 의사결정이 중요하다는 점을 강조한다. 종합하면, 본 연구는 기술 주도의 구조적 전력 수요 변화를 분해하여 분석하고, 장기 불확실성을 확률 분포로 표현하며, 다양한 예측 접근법을 통합적으로 비교할 수 있는 분석 프레임워크를 제시한다. 특히 제안된 프레임워크는 모듈형 구조를 기반으로 설계되어, 향후 새로운 전력 수요원이 등장하더라도 예측 아키텍처의 근본적 변경 없이 확장이 가능하며, 기술 확산에 수반되는 불확실성을 체계적으로 관리할 수 있는 기반을 제공한다. 이는 대전환 기술이 지배하는 미래 환경에서 탄력적이고 지속 가능한 전력 시스템 계획을 지원하는 실질적인 방법론적 기여로 평가된다. The global electricity landscape is undergoing rapid structural transformation as transformative technologies such as artificial intelligence (AI) data centers and electric vehicles (EVs) expand at an unprecedented pace. These technologies not only increase overall electricity consumption but also reshape its temporal and spatial distribution, weakening long-standing assumptions of demand stability. AI data centers create persistent, high-density base-load demand, while EV charging generates locally concentrated and temporally synchronized loads. As these dynamics intensify, uncertainty in long-term electricity forecasting increases, along with risks related to supply demand imbalance, grid congestion, and, in extreme cases, system-wide disruptions or disaster-level events. These challenges highlight the need for forecasting frameworks that go beyond single point estimation and incorporate structured uncertainty assessment to support adaptive, risk-aware electricity system planning. This dissertation presents a modular forecasting framework capable of isolating and quantifying the impacts of transformative technologies on electricity demand. Baseline demand driven by traditional macroeconomic factors such as GDP, population, and labor force is separated from technology-induced demand generated by AI data center expansion and EV adoption. This structural decomposition enables the model to capture heterogeneous and nonlinear demand characteristics that are not adequately represented in conventional econometric or univariate time series models. Using ARIMAX and Prophet models, long-term electricity demand in the United States from 2025 to 2050 is forecasted under a consistent set of exogenous variables. Prophet exhibits strong trend stability and delivers the highest deterministic accuracy, while ARIMAX shows sensitivity to structural shocks, making it suitable for policy-oriented scenario analysis. To address the limitations of purely deterministic forecasts, Monte Carlo simulation is incorporated to generate probabilistic prediction intervals. This approach enables interpretation of long-horizon uncertainty and provides insight into potential system stress scenarios posed by unexpected technology adoption patterns or extreme demand outcomes. A further contribution of this dissertation is the application of a diffusion-based generative model to long-term electricity demand forecasting. While diffusion models have been used primarily in image and video generation and, more recently, in short-term load forecasting, this study applies the method to the 2025–2050 horizon and incorporates AI and EV indicators as conditional drivers. The diffusion model captures nonlinear uncertainty accumulation and asymmetric risk structures, offering a probabilistic representation of tail events such as extreme peak loads or localized grid stress that conventional models cannot adequately express. The comparative analysis of ARIMAX, Prophet, and diffusion models under identical conditions is another distinctive component of this research. Each model reveals different strengths in explanatory power, trend robustness, and probabilistic coverage. This integrated comparison provides a foundation for model selection based on risk-centered planning objectives rather than reliance on any single forecasting methodology. Overall, this dissertation offers a coherent analytical framework that decomposes technology-driven structural demand shifts, represents long-term uncertainty through probabilistic distributions, and compares multiple forecasting approaches under a unified structure. The modular design of the proposed framework further enables the integration of additional emerging electricity demand sources without fundamental changes to the forecasting architecture, thereby supporting scalable and risk-aware electricity planning under technological uncertainty. In an era shaped by transformative technologies, the proposed methodology supports the development of resilient, sustainable, and forward-looking power systems under increasingly uncertain future conditions.

(A) study on the characteristics of electricity consumption in manufacturing sector : estimation of electricity demand function and LMDI approach

박정진 Graduate School, Yonsei University 2023 국내박사

In Korea, energy consumption is high and inefficient, and the rate of improvement in energy efficiency is slow compared to major countries. In addition, the energy and electricity intensity are high, and the gap is huge compared with major countries, so it is not easy to improve to the level of major countries within a short period of time. In particular, the electricity consumption in industrial and manufacturing sectors, even after rapid changes in the industrial environment, still account for an overwhelming majority in terms of both quantities and amount, so these sectors are very important subjects in the planning and analysis of national electricity supply and demand. High electricity consumption and power consumption inefficiency in industrial and manufacturing sectors are negative factors in terms of ESG management and sustainability. Therefore, it can’t be emphasized more that the success of achieving the national carbon neutrality depends on whether Korea can improve efficiency of power consumption dramatically. In this study, to analyze the characteristics of electricity consumption in the manufacturing sector according to the recent changes in the electricity consumption environment, time series data such as electricity consumption, electricity price, and GDP were used from 2009-2021 to classify manufacturing types into 13 groups, electricity demand function and elasticity in Korea, were estimated, and LMDI decomposition analysis of the same period was performed with the same data. The results on this study are summarized as follows. As a result of estimating the electricity demand function and elasticity using the FGLS and ARDL models, price and income elasticities were significantly lower than in previous studies. By industry, price and income elasticities of industries that consume a lot of electricity were higher than those of industries that consume a few electricity. The correlation between electric tariff level and elasticity during the study period showed that price and income elasticities were high when the unit price was relatively high, and price and income elasticity were low when the electricity rate was relatively low. In some industries, economic distortion and electrification occurred due to the policy of tariff regulation, and the analysis during the COVID-19 period did not show significant statistic results, but income elasticities of industrial sector below 300kW and mass-electricity- consuming industries were estimated to be high. As a result of LMDI decomposition analysis, the total effect and activity effect were increased as a whole in the manufacturing sector, and the proportion of the activity effect to the total effect continued to expand. Contribution of the activity effect to the increase in total electricity consumption was very big. The structure effect acted as a factor in reducing electricity consumption, and the intensity effect also appeared after 2015, showing that the electricity consumption efficiency and electricity consumption intensity are improving. As a result of analysis in manufacturing sector, it was found that the total effect and activity effect were mostly increased in the industries that consume a lot of electricity. The magnitude of the activity effect also tended to increase, and the intensity effect was improved by electricity consumption intensity in industries such as electronics, basic metals, machinery equipments, non-metallic minerals, wood papers and textile leathers. During the COVID-19 period, structure effect and intensity effect are improving, indicating that industrial structure improvement, energy intensity and efficiency improvement are continuing even in a situation where economic activity has shrunk. The policy implications on this study are as follows. Firstly, since electricity price has been identified as the most important policy tool that has a direct impact on electricity consumption, the government should actively promote the policy of pricing to improve electricity consumption efficiency. Secondly, it is necessary to establish the tariff mechanism to secure economic signals and predictability of electricity prices, and the power market system must also be reorganized to reflect changes in the electric system in a timely manner based on the market principles. Thirdly, as the paradigm of electricity supply shifts from supply-oriented to demand-oriented, efforts to improve energy efficiency in manufacturing sector and to enhance electricity consumption intensity must be continued. The uniqueness and significance on this study are as follows. Firstly, the manufacturing sector was classified in detail using the data related to electricity consumption up to the present, and then the characteristics of electricity consumption in manufacturing sector was analyzed. Secondly, based on the same analysis data, it was tried to increase the reliability and usability of the research results by analyzing the electricity demand function, elasticity estimation, and LMDI decomposition together. Thirdly, not only the 2009-2021 period but also the distinctive sections were subdivided to confirm the effect of electricity consumption characteristics rate increase, and intensity improvement by period. Finally, the change in electricity consumption during the COVID-19 period was also examined, and implications from policy and economic perspectives were drawn, and it was suggested that energy efficiency improvement should be continuously promoted in manufacturing sector. In general, previous elasticity studies have found it difficult to estimate meaningful elasticity because price fluctuations are almost fixed. However, in this study, it was confirmed that the elasticity value was derived as a meaningful estimation result during the period of price fluctuation. Consequently, the price mechanism should be introduced to contribute to efficiency improvement and energy saving. It is expected that the demand for electricity will continue to increase in the future, due to the economic growth caused by industrial development and the rising in electricity consumption and electrification. In this situation, in order to improve energy efficiency, implement carbon neutrality and maintain a balanced energy system, the speedy role and collaboration of stakeholders such as the government, electricity companies, industries and power consumers are very important. Lastly, in this study, there was a limitation in estimating the electricity demand function by expanding the industry classification due to the limitations of data classification and period of data accumulation. In the calculation of elasticity and the analysis of LMDI, the analysis of manufacturing sector was performed in only 13 groups due to restrictions on the use of data by industry. In the future, it is necessary to analyze using GDP data for all industries, and if segmented data for each industry is used, microscopic characteristics of electricity consumption for all manufacturing sectors can be accurately identified.