### Abstract

The short term electric load forecasting which is generally from one hour to one week is one of the intelligent electric grid (smart grid), for control of stable load supply hour-to-hour or day-to-day. The difficulty of short time forecasting is that the trend of time series usually change, and the non-adaptive auto-regressive integrated moving average (ARIMA) could not fit accurately. To solve that problem, conventional adaptive ARIMA with constant forgetting factor that gives a larger weight to more recent train data for dealing with non-stationary change of stochastic disturbance. The forgetting factor governs the recursive least squares (RLS) algorithm. However, constant forgetting factor usually result in over-fitting that increases forecasting error. A new adaptive ARIMA is proposed in this paper to improve the accuracy with lazy learning algorithm to reduce over-fitting error.

Original language | English |
---|---|

Title of host publication | Proceedings of the International MultiConference of Engineers and Computer Scientists 2018, IMECS 2018 |

Editors | Oscar Castillo, David Dagan Feng, A.M. Korsunsky, Craig Douglas, S. I. Ao |

Publisher | Newswood Limited |

ISBN (Electronic) | 9789881404886 |

Publication status | Published - 2018 Jan 1 |

Event | 2018 International MultiConference of Engineers and Computer Scientists, IMECS 2018 - Hong Kong, Hong Kong Duration: 2018 Mar 14 → 2018 Mar 16 |

### Publication series

Name | Lecture Notes in Engineering and Computer Science |
---|---|

Volume | 2 |

ISSN (Print) | 2078-0958 |

### Other

Other | 2018 International MultiConference of Engineers and Computer Scientists, IMECS 2018 |
---|---|

Country | Hong Kong |

City | Hong Kong |

Period | 18/3/14 → 18/3/16 |

### Fingerprint

### ASJC Scopus subject areas

- Computer Science (miscellaneous)

### Cite this

*Proceedings of the International MultiConference of Engineers and Computer Scientists 2018, IMECS 2018*(Lecture Notes in Engineering and Computer Science; Vol. 2). Newswood Limited.

**Adaptive ARIMA model based on lazy learning algorithm for short period electric load forecasting.** / Li, Chengze; Murata, Tomohiro.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

*Proceedings of the International MultiConference of Engineers and Computer Scientists 2018, IMECS 2018.*Lecture Notes in Engineering and Computer Science, vol. 2, Newswood Limited, 2018 International MultiConference of Engineers and Computer Scientists, IMECS 2018, Hong Kong, Hong Kong, 18/3/14.

}

TY - GEN

T1 - Adaptive ARIMA model based on lazy learning algorithm for short period electric load forecasting

AU - Li, Chengze

AU - Murata, Tomohiro

PY - 2018/1/1

Y1 - 2018/1/1

N2 - The short term electric load forecasting which is generally from one hour to one week is one of the intelligent electric grid (smart grid), for control of stable load supply hour-to-hour or day-to-day. The difficulty of short time forecasting is that the trend of time series usually change, and the non-adaptive auto-regressive integrated moving average (ARIMA) could not fit accurately. To solve that problem, conventional adaptive ARIMA with constant forgetting factor that gives a larger weight to more recent train data for dealing with non-stationary change of stochastic disturbance. The forgetting factor governs the recursive least squares (RLS) algorithm. However, constant forgetting factor usually result in over-fitting that increases forecasting error. A new adaptive ARIMA is proposed in this paper to improve the accuracy with lazy learning algorithm to reduce over-fitting error.

AB - The short term electric load forecasting which is generally from one hour to one week is one of the intelligent electric grid (smart grid), for control of stable load supply hour-to-hour or day-to-day. The difficulty of short time forecasting is that the trend of time series usually change, and the non-adaptive auto-regressive integrated moving average (ARIMA) could not fit accurately. To solve that problem, conventional adaptive ARIMA with constant forgetting factor that gives a larger weight to more recent train data for dealing with non-stationary change of stochastic disturbance. The forgetting factor governs the recursive least squares (RLS) algorithm. However, constant forgetting factor usually result in over-fitting that increases forecasting error. A new adaptive ARIMA is proposed in this paper to improve the accuracy with lazy learning algorithm to reduce over-fitting error.

UR - http://www.scopus.com/inward/record.url?scp=85062596844&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85062596844&partnerID=8YFLogxK

M3 - Conference contribution

T3 - Lecture Notes in Engineering and Computer Science

BT - Proceedings of the International MultiConference of Engineers and Computer Scientists 2018, IMECS 2018

A2 - Castillo, Oscar

A2 - Feng, David Dagan

A2 - Korsunsky, A.M.

A2 - Douglas, Craig

A2 - Ao, S. I.

PB - Newswood Limited

ER -