A maximum entropy approach to the estimation of spatially and sectorally disaggregated electricity load curves. (1st September 2018)

• Record Type: Journal Article
• Authors: Többen, Johannes; Schröder, Thomas
• Is Part Of: Applied energy. Volume 225(2018)
• Page Start: 797
• Page End: 813
• Publication Date: 2018-09-01
• Subjects: Maximum entropy -- Maxent -- Spatial electricity load -- Sectoral and spatial disaggregation -- Electricity consumption -- Load estimation -- Representative load curves; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2018.04.126
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: A novel maximum entropy (Maxent) model is proposed for estimating spatially and sectorally disaggregated electricity load curves. Model reconciles measured loads, national load curves and energy balances. Model delivers information base for system planning of distributed generation. Dual of Maxent reduces computational requirements substantially. In a case study to Germany, results are found to be superior to pure Top-Down and Bottom-Up approaches. Abstract: Usually, disaggregated electricity load curves are estimated by using Top-Down or Bottom-Up approaches. The former requires estimating weightings for downscaling aggregated information, while the latter requires extrapolating micro-level information. In both cases, estimation would ideally be based on as much regional and sector specific information as possible, in order to obtain a realistic representation of the magnitude and temporal pattern of a regional sector's electricity consumption. Typically, such attempts are significantly hampered by issues of limited and possibly inconsistent data, differing levels of detail, and mismatching data classifications. This paper proposes a novel nonlinear programming model based on the maximum entropy approach. The model allows for electricity load curve estimation at arbitrary spatial, sectoral and temporal resolution, from partial and possibly inconsistent information. The proposed model integrates and systematically utilizes data usually used by either Top-Down or Bottom-Up approaches. In a case study using German data it is shown that the model combines the strength of both and, at the same time, overcomes the challenges specific to Top-Down or Bottom-up estimation.