Avoiding "Prototype" and Average Load Data Aggregation Errors

While cost of service studies use average load profile data, using average or "prototype" load profile data for other applications can generate serious errors in analysis results. This page describes this aggregation error.

Valid segment load profiles can only be developed by starting with individual customer data and working up to the segment average using customer definitions compatible with analysis objectives.

Hourly utility customer load data are important inputs to energy-efficiency studies, demand response, marketing, distributed generation and other technology analysis, policy evaluations, forecasting and other energy-related analysis. While broad , predefined segment-average load profiles are appropriate for some evaluations (see below), the application of these average loads in most analyses limits the applications and can result in large errors. In many cases, analysis results actually provide the wrong answer to specific questions. To see an example of how aggregation error can provide the wrong answer see Assessing Utility Market "Headroom": Pitfalls of Traditional Analysis.

The problem that arises in using broad market segment category load shapes (e.g., small and large office, schools, hospitals, etc.) can be illustrated with the following table:

Commercial Sector Hourly Load Detail Analysis Possibilities From Least to Most Detailed

LEAST DETAILED 20 Predefined Segments 80 Predefined Segments ... MAISY Databases MOST DETAILED
An Average of All Commercial Customers 20 Segments by Commercial Business 80 Segments by Business, Heat and AC Info ... A Statistically Representative Sample of Individual Customers All Individual Customers

Each analysis option from left to right reflects a progression in the level of customer detail used in an hourly-loads based analysis. Hardly any analysis is conducted with the least detailed "average commercial sector load data" on the left. Similarly, analysis is usually not conducted with hourly load data on each customer in a market because those data do not exist for most applications. MAISY Hourly Load Databases and Systems provide a representative sample of individual customer hourly loads in each state and utility service area permitting detailed customer-level analysis. Rate analysis, customer profitability, equipment design issues, DG analysis, strategic pricing and many other applications have utilized the MAISY customer data.

Hourly load data are also available by broad predefined market segments (the second and third options represented above). The problem in using these predefined segments is that segment definitions are hardly ever compatible with the analysis issue at hand. For instance, an analysis of DG economics for a 60 kW system in restaurants might attempt to apply "average restaurant" electric and thermal loads to calculate the economics of the DG system. However, since "average restaurant" hourly loads include buildings that are too small to economically utilize the DG system, an analysis of average loads does not accurately reflect the economics of the market segment of interest, i.e., restaurants with day-time loads greater than 60 kW. Furthermore, since restaurants with day-time loads greater than 60 kW will have operating hours and electric and thermal load characteristics different than those of smaller restaurants, an analysis using average loads almost guarantees an inaccurate economic evaluation of the DG system.

In this example, hourly load profiles for the appropriate market segment ( restaurants with day-time loads greater than 60 kW) would, however, provide appropriate hourly load electric and thermal data for the DG economic analysis. That is, valid segment load profiles can only be developed by starting out with individual customer data and working up to the segment average using customer definitions compatible with analysis objectives. MAISY Utility Customer Databases and Analysis Systems use exactly this approach to provide appropriate segment average hourly load data. This approach is illustrated in the following table:

MAISY Commercial Sector Hourly Load Detail Analysis Possibilities

User-defined Segments by Any Customer Characteristic <= A statistically Representative Sample of Individual Customers

MAISY Database Systems automatically provide load profiles for individual customers and for customer segments based on any user-specified selection of business type, size characteristics (kW, kWh, employment, square feet), operating characteristics and many other characteristics.

The chart below illustrates masking of customer diversity that occurs with market segment load profiles. This chart shows week-day hourly loads in July for 23 medium-sized air-conditioned office buildings in Houston along with the average of all 23 buildings. To provide a meaningful comparison, each of the loads has been normalized so that the sum of kW loads over the 24 hours equals 1.0. Differences in load shapes of the individual customers from the average show the inaccuracies inherent in using average or prototype load profiles to infer results for all medium office buildings in Houston. For instance, in an evaluation of a baseload DG system, analysis with the average profile might show a unfavorable economic evaluation while in reality, loads for many of the customers might operate very economically in baseload mode.

By utilizing an individual customer hourly loads platform, MAISY Hourly Load Databases and Profiler Systems avoid the constraints and inaccuracies that result from predefined customer types and market segments.

Comparison of Hourly Load Data Sources

The following description compares MAISY Utility Customer Database and Hourly Load Systems features with characteristics of other hourly load data sources.

Traditional Utility Load Research Data. Utilities typically collect whole building loads for a sample of customers for cost of service analysis and rate structure development for individual rate classes. Since these samples must reflect a small number of rate classes (typically fewer than a dozen), load research samples often consist of no more than several hundred customers. The sample design used in this data collection as well as the small sample sizes make this data inappropriate, on statistical grounds, as accurate representations of the individual customer segments (e.g., hotels, nursing homes, hospitals, etc.). Even when load research data are combined from a large number of utilities to compile load shapes for more detailed customer segments, there is no guarantee that the resulting shapes reflect the segment average or that diversity in the sample reflects diversity in the population. An additional drawback of utility load research-based data is the absence of customer characteristics.

Predefined Prototype and Average Segment Load Shapes. Predefined prototype or segment (small office, large office, hotel, etc.) load shapes are typically developed by applying engineering heat load models to "prototype" buildings (i.e., typical hotels, hospitals, etc.) and calibrating annual energy use to average energy use estimates for the particular building type. In some cases, average hourly loads reflect the average heat load model results for a small sample of buildings within the prototype classification. While average load shapes are appropriate for some kinds of energy analyses, they are less useful for most hourly loads-based analysis because they provide segment definitions which cannot be changed (see the 50 kW restaurant DG example above) and they also ignore the diversity of customers within the "prototype" building classifications. For instance, an analysis of "average" nursing home hourly load profiles might suggest an unattractive DG application whereas an analysis of a sample of actual nursing homes might identify one-third of the nursing homes as attractive candidates for DG. In this case the two-thirds of the nursing homes who are unattractive DG candidates predominate in "average" hourly loads providing a negative economic evaluation while ignoring the fact that significant differences exist in subsets of the prototype.

MAISY Utility Customer Hourly Loads. MAISY (Market Analysis and Information System) hourly loads databases have been developed from information on more than 2,000,000 individual utility customers throughout the US and provide a representative sample of customers for individual states and utility service areas. Hourly load data are provided for each individual customer record in the MAISY State and Utility Service Area Databases. The large number of customers within the databases along with individual load data for each customer maintains the diversity of actual customer populations, providing a more accurate analysis of customers, markets and market segments compared to "average" customer information and load data based on utility load research samples.

Appropriate Applications of Segment-Average Load Profiles

Appropriate applications for broad predefined prototype and segment-average load profiles include some cost of service, strategic pricing, and certain other applications where the "aggregation errors" associated with prototypes and segment averages do not create their typical statistical inference problems. MAISY Hourly Load Databases and Systems can also provide these prototype and segment-average load profiles as outputs of the Database System or in the form of Hourly Load Profile Data.

Example of Different Load Shapes for Two Market Subsegments Within the Small Office Building Category

Different business and dwelling unit types have characteristic hourly load profiles. However, focusing on aggregate load profiles of several dozen customer segments ignores the fact that the hourly load profile variation among customers within these customer categories is typically greater than variation across the categories. This relationship is illustrated in the charts below and in the Energy Market Analysis section of the MAISY Web site.

The MAISY system permits users to select individual customers or customer segments based on dozens or even on hundreds of customer characteristics. Pick any combination of business type, floor space, operating schedules, space heating fuel, year of construction and many other variables to zero in on a specific customer type or market segment.

What about other load-profiling systems that offer 12, 36 , 75 or some other limited number of fixed customer segments? To represent 13 commercial business types, three heating fuel types (electric, gas and oil ) and three customer sizes (small, medium and large ) requires 117 prototypes or "typical" buildings. Add in three age categories and more than 350 "fixed prototypes" would be required, well beyond the scope of these "fixed" systems. With MAISY, customer and segment selections provide hundreds of possible definitions with nearly unlimited choices of customer characteristics. Only MAISY provides the detail and flexibility required to reflect the extensive customer and segment detail required in competitive markets.

Sources of load profile data which rely on fixed customer segments (e.g. large, medium and small offices) typically develop hourly load data with engineering models (e.g., DOE2) of a single "prototype" building. The aggregate nature of these representations misses the variation that exists among individual buildings within these segments, hiding important market information. For instance, a particular rate structure may provide an acceptable, competitive profit on an entire segment represented by one prototype load profile; however, analysis of subsets of the segment (which can be performed with MAISY but not with the "prototype or typical" load profile approach) may reveal significant diversity in profit levels across customer sub-segments such that some customers are provided power at a loss while profit margins on other customers result in cream-skimming targets for other suppliers.

These limitations of traditional aggregate load shapes can easily be overcome with MAISY where users can see and evaluate the whole range of load shapes within each specific building type with just a few mouse clicks. For example, the chart below shows hourly loads for a sample of California office buildings on a July week day.

This traditional-looking summer load profile reveals only part of the picture because different subsets of this office market reflect very different load profiles. For instance, drilling down to compare load profiles in two size categories shows substantial differences in July week-day load profiles.

Offices with 25-50k Square Feet

Offices with 50-100k Square Feet

While the general load shapes appear similar, the difference in the peak-to-off-peak ratios between these two building sizes is substantial. The ratio of 3:00 pm loads to 1:00 am loads is 5 for the smaller buildings and closer to 3.5 for the larger buildings. Exploring the MAISY Database reveals several reasons for these differences. The larger buildings are two times more likely to have lighting controls and three times more likely to have HVAC controls. The larger buildings also have longer operating hours and greater operation of HVAC and lighting systems in "after-hours" periods.

Comprehensive market analysis requires hourly load data for detailed market segments; furthermore, these segment definitions will be subject to change and frequent refinements. MAISY permits users to evaluate and immediately develop hourly load profiles for any market segment defined by the hundreds of customer variables in the MAISY Databases.

Another example of the benefits of using customer based-data for market analysis is presented in the Energy Insights article Assessing Utility Market "Headroom": Pitfalls of Traditional Analysis.