Load Forecasting in Deregulated Markets

Accurate forecasting of end user electricity demand is essential for successful participation in deregulated retail electricity markets. The foundation for accurate load forecasting is knowledge of end user load profiles and the dynamics that cause the profiles to change. End users need to know the sources of variation in their load requirements to evaluate the cost and market risk reducing potential of: DSM engineering projects, modifications to operations processes and process scheduling, self generation investments, and alternative direct access supply proposals. Energy service providers require the same information to plan long-term power supplies, estimate end user service costs, price retail service, plan short-term delivery schedules, assess market risk exposure, and ultimately determine the sources of imbalance charges due to forecast errors.

The need for understanding and forecasting the time patterns of individual end user electricity consumption is an important aspect of retail market deregulation. Hourly market price patterns create new cost structures and market risks for end users and motivate changes in historic consumption patterns to reduce power costs and risks. Because hourly electricity prices are highly correlated with high and low temperatures, the new market especially motivates end users to change building structures and operations to reduce the weather sensitivity of their consumption patterns. In this deregulated market environment, end user load forecasting requires more rigorous analysis and modeling than the mere averaging of historical load patterns and shifting those patterns forward in time. Explanatory models that describe the dynamics of changing electricity consumption patterns are essential to minimizing the power costs and imbalance charges due to load forecasting errors.

For retail energy service providers, the broad customer class aggregate modeling and forecasting approach that has been successfully used by regulated utilities is simply too blunt an approach to apply in competitive markets where the ESP’s customer base is likely to be changing, market prices are volatile, and end users will be actively modifying their load shapes. Similarly, naively applied black box, data mining approaches that have become so popular, lack the explanatory power needed for rational forecasting of future load requirements in this dynamic setting.

To the extent that groups of end users (such as residential and small commercial) can be identified that are sufficiently homogeneous in terms of load shape and load responses to weather conditions and price, then aggregated models for such groups will suffice. Large commercial and industrial customers, however, are typically different enough in the electric consumption patters that aggregation will not be effective. The appropriate size cutoff for individual versus group modeling and forecasting is ultimately determined by the trade-off between forecast error settlement charges and modeling costs.

ESC’s Load Modeling and Forecasting Service

Since having pioneered some of the first econometric load forecasting models for electric utilities, ESC has continued to specialize in the methodological and practical issues of electric load forecasting. We have applied an extensive array of techniques and model structures over the years to continually improve on forecasting accuracy.

To facilitate our load modeling and forecasting practice, ESC has developed the LOADCASTERTM software system. LOADCASTER has comprehensive capabilities for model development forecasting and reporting. LOADCASTER incorporates a time-series relational database that is optimized for efficient storage and super fast retrieval and manipulation of large volumes of hourly and subhourly data.

Integrated with the database is ESC’s state-of-the-art Pooled Daily Load Shape Regression (PDLS) program (and other estimation methods), powerful model simulation tools, and a flexible report writer.