摘要 |
An object is to predict an amount of electric power that is demanded in a future prediction period with good accuracy. A demand prediction system (10) includes a receiver (11), a memory (12), an extractor (13), a model generator (14) and a predictor (15). The extractor (13) is configured to extract consumption periods for branch circuits (22), during each of which power is consumed once. The model generator (14) is configured to generate models. Each model includes: an order in which the consumption periods have occurred; amounts of electric power respectively consumed during the consumption periods; and start time points and duration times, of the consumption periods. The model generator (14) is further configured to calculate transition probabilities between the models. The predictor (15) configured to predict, as a prediction result, an amount of electric power that is demanded in a prediction period defined as a time period after a determination time point, based on: a certain model in past before the determination time point, of the models; and a transition probability corresponding to the certain model. |