A predictive demand model for systems planning, using noisy realization theory

✍ Scribed by Lov Kumar Kher; Soroosh Sorooshian

• Publisher: Elsevier Science
• Year: 1988
• Tongue: English
• Weight: 620 KB
• Volume: 24
• Category: Article
• ISSN: 0005-1098
• DOI: 10.1016/0005-1098(88)90114-8

✦ Synopsis

Identification of a demand model, when all the independent and dependent variables are noisy, is one of the major problems in systems planning. Generally, regression or time-series models are used for such problems. It has been suggested that the existing methods to tackle noisy variables are not useful because of nonidentifiability and unboundedness problems. We present an alternative procedure to formulate a multiple input-single output dynamic noise-invariable model (NVM). The NVM explicitly considers noise in the data for all the variables of the system. A mathematical programming-based solution algorithm is developed to identify the NVM. This algorithm gives maximum and minimum bounds (a range) on both the model parameters and the noise covariance matrix. The unboundedness problem is addressed by proposing a procedure to use prior information about the noise. Finally, a prediction scheme is presented to generate demand scenarios for the future. We also implement these procedures; and the energy demand for the United States is modeled as an example.