Scholarlink

This study is a researchResearchInferenceDynamicsGrid effort to extract causal relations from timeTime series data of electrical consumption obtained from a power gridGrid serving a university campus. Electrical consumption exhibits different characteristics and dynamicsDynamics depending on building types and diverse external conditionsConditions such as seasons, temperatures, and school calendars. Causal relations among such factors and electrical consumption from power gridGrid systemsSystems are expected to provide predictionPrediction and controlControl strategies to reduce campus electricity costs. Causal graphs, which should be Directed Acyclic Graphs (DAGs) constructed from data, are complicated. Certain strategies must be developed to give them sufficient explainability. The mathematical basis for causality inferenceInference must be strengthened using mathematical logic and discreteDiscrete geometry, as well as high-performance machine learningLearning techniques and cognitive sciences to enhance their explainability. Moreover, network controllability should be evaluated and analyzed using appropriate network centralities. This researchResearch was undertaken as a part of Discovery Intelligence Laboratory co-established by Fujitsu Ltd. and Tohoku University, which is intended to extend application areas to address widely diverse social issues and to construct robustRobust mathematical bases for them.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Causal Inference for Power Grid Dynamics on a University Campus

  • 章节

  • 发表日期：

    2026年

查看全文
  • Yuka Fujiki,
  • Takashi Yanase,
  • Hiroyasu Ando,
  • Yuta Fujishige,
  • Yuiko Ohta,
  • Hiroyuki Higuchi,
  • Hiroshi Suito

摘要

This study is a researchResearchInferenceDynamicsGrid effort to extract causal relations from timeTime series data of electrical consumption obtained from a power gridGrid serving a university campus. Electrical consumption exhibits different characteristics and dynamicsDynamics depending on building types and diverse external conditionsConditions such as seasons, temperatures, and school calendars. Causal relations among such factors and electrical consumption from power gridGrid systemsSystems are expected to provide predictionPrediction and controlControl strategies to reduce campus electricity costs. Causal graphs, which should be Directed Acyclic Graphs (DAGs) constructed from data, are complicated. Certain strategies must be developed to give them sufficient explainability. The mathematical basis for causality inferenceInference must be strengthened using mathematical logic and discreteDiscrete geometry, as well as high-performance machine learningLearning techniques and cognitive sciences to enhance their explainability. Moreover, network controllability should be evaluated and analyzed using appropriate network centralities. This researchResearch was undertaken as a part of Discovery Intelligence Laboratory co-established by Fujitsu Ltd. and Tohoku University, which is intended to extend application areas to address widely diverse social issues and to construct robustRobust mathematical bases for them.