Title:Oral: A nonlinear time series analysis of electrical consumer load.

Electrical consumer demand exhibits significant fluctuations in magnitude across a range of time scales from seconds to days. These variations pose substantial challenges for dynamic load balancing as supply must continually meet the demand. The integration of renewables such as wind and solar photovoltaics further exacerbates these challenges due to their inherent variability. Understanding the dynamics of electrical consumer load is crucial for effective energy management and optimal unit dispatch commitments by grid operators or ISOs. Moreover, accurate short-term forecasting of electricity demand remains one of the most difficult tasks for Isos, as demand dynamics are strongly influenced by highly nonlinear, external factors. Several statistical techniques have been proposed to fit highly parameterized mechanistic models to complex consumer load time-series but have fallen short of providing accurate short-term forecast. We apply a model-free forecasting method based on Takens embedding theorem and state space reconstruction, commonly referred to as Empirical Dynamic Modeling (EDM). Using standardized RMS error statistics (SRMSE), we study whether EDM can offer superior short-term forecasts over traditional linear time-series approaches such as autoregressive (AR) and autoregressive moving average (ARMA) models, when using a single time-series from a multivariate system. We present preliminary results of our EDM analysis as applied to a publicly available time-series for the California grid as a case study.