- because powerful tools don't have to be difficult to use -

Load Modeling

Introduction

The biggest factor in creating an accurate model in DESS is loading. To make sure the loading in DESS is accurate you need to understand the difference between average loads, coincident peaks, local peaks, and to understand daily peaks vs. monthly and annual peaks.

Individual vs. Aggregated Loads

The load on a distribution system is varying all the time and varying across the system. At the local level, the load on each distribution transformer may vary very quickly, with a large percentage fluctuation from minute to minute. If you were to graph the load versus time, there would be no discernible pattern. However, if you were to take 5 transformers and measure the aggregated load on these transformers, the load variation would begin to average out across the total load and a rough daily load pattern would begin to emerge. If you were to consider 25 or more transformers, then you would begin to approach a smooth daily load curve.

Variations Over Time

Load varies over the course of a day due to usage patterns. Lights are used more when it's darker. Commercial customers use more electricity when they're open for business. Residential customers use more power for cooking and refrigeration when they prepare meals. These variations give distinctive usage patterns for different types of customer over the course of a day.

Load also varies according to the type of day. The schedules of residential customers are different on weekdays than on weekends. Industrial loads may vary on Mondays and Fridays due to startup and shutdown issues.

Load also varies according to season. Seasonal temperatures affect heating and cooling loads. Businesses may have seasonal variations. The number of hours of daylight varies for utilities in northern and southern countries.

So we can see that even discounting random variation of loads, there are differences in load patterns across the course of a typical day, a typical week and a typical year.

Random Variation

In addition to all the patterns we can identify, there will also be random variation in loads. These may be an accumulation of random fluctuations at individual customers or may be caused by large scale but unpredictable events such as elections, strikes, or particularly popular TV shows! The best we can hope to achieve with load modeling is to model average conditions.

Peaks and Averages

There are a number of different types of peak that can be measured. These include the following:

Local Peak - the time and magnitude of the peak load for a given feeder or station.
System Peak - the time and magnitude of the peak for the entire system. This may be different from any/all of the local peaks.
Absolute Annual Peak - the highest recorded value for an entire year. This can be either local or system wide.
Absolute Monthly Peak - the highest recorded value for a given month. This can be either local or system wide.

Peak information is important for designing for system limits, and may be required for computation of demand charges, but it is less useful for inferring or predicting load behavior. The best data for predicting load behavior is average loading. Properly averaged load data can eliminate random load variation while maintaining useful information for daily, weekly, seasonal or temperature variations. For example, if you could take hourly readings on a feeder every weekday for June and July, and then average these values by hour, you would have a very good idea of how that feeder would behave during an average weekday in August.

DESS Load Modeling

DESS uses load categories to represent a typical category of customer. Each category contains data for the daily, weekly and monthly load curves associated with the type of customer, such as residential, retail commercial, or a specific type of industrial load. This allows the analyses in DESS to model average conditions for these types of load. If the customers types have been identified correctly and the load curves are accurate, then this should allow the analyses in DESS to correctly match recorded local and system average conditions. Now that we've discussed peaks and load variation, we can discuss what this means for you and your DESS model. There are a number of ramifications.

You can't model small individual loads accurately

There is simply too much variation on a small load and you can't predictively model what it will do in the future. However, you can accurately model the behavior of loads on a feeder or a chunk of feeder. It is impossible to predictively model an individual customer (although some research has been performed on statistical modeling of individual loads). However, we can represent what happens on a typical type of load, such as a type of residential customer, or a class of commercial customer. DESS uses 'load categories' to represent these classes of customer. It is important to recognize that the load categories can only represent the load behavior under typical or average conditions. This is why it is important to understand the nature of system peaks.

Monthly absolute peaks of individual feeders are not useful for validating your DESS model

Monthly peaks may only represent conditions during a single 15 minute period during the month and do not represent average conditions, so the results of a load flow in DESS will typically produce lower loading that the absolute monthly peaks. Furthermore, the monthly peaks for different feeders may represent different times of day or even different days. This means that both the magnitude of the loads and the ratio between loads on different feeders will be different than what needs to be represented in DESS.

Coincident feeder measurement data is best

The most useful data for validating your DESS model is a set of feeder measurements taken at a single point in time. These will give a true relationship between the loading on the different feeders. Even better would be to take a series of these snapshots and average them. For example, if you were to take a snapshot of the loading on each feeder at 3pm on each weekday for two weeks and then average these values, they should match up very closely to the load values produced by a load flow in DESS.

Historical feeder data is probably not useful.

Many utilities keep a lot of historical data for loading on their feeders. Unfortunately, rather fewer people keep track of system configuration changes. Often, it is impossible to determine what the system configuration was when particular measurements were taken. Was there a load transfer between feeders? Was there maintenance activity? Has new load been added to a particular feeder since the measurements were taken? Uncertainty about these questions can limit the value of historical data.

You need to scale an analysis to simulate peak conditions

By default DESS will show average conditions on your system. However, sometimes you want to run analyses at absolute peak conditions. To do this, you will need to scale up the loads by changing the load scaling factor for the analysis. The scaling factor will represent the ratio between the absolute system peak and the average system peak.

References

1. Kundur, P., Power System Stability and Control, 1993, McGraw-Hill, Chapter 7.4, pp 306-311

2. Willis, H.L., Power Distribution Planning Reference Book, 2004, Marcel Dekker, Chapter 2,3, pp 47-102