6 CAPITULO CUATRO
6.2 DEL GOBIERNO Y LA ADMINISTRACIÓN:
The identification of typical electrical usage patterns within households is necessary as a starting point for:
• Defining the type of DSM program (e.g. peak clipping) to undertake to match the overall goals.
• Understanding the current pattern of electricity usage to allow for decisions on desired changes needed to the pattern.
• Assessing the impact of any initiatives to reduce overall energy usage to discover the amount of overall reduction which occurs during different times of the day.
• Allowing accurate aggregation to provide a pattern of total demand over the day that is to be met by supply side generation and trans- mission.
The definition of load profiles has been a long standing activity within the electricity industry but the current and forthcoming avalanche of data provides for alternative ways of approaching the definition of the load profiles.
Current industry practice has focussed on commercial electricity users. Electricity Association [51] identifies a process for defining the details of eight different standard usage profiles for the UK. Of these eight, only two refer to domestic properties although the profiles take into account the season and the day of the week. As an example of the standard pro- files, Figure2.4shows the winter profiles for Saturday and Sundays, both for Economy7 customers and non Economy7 customers, plotted as 48 half hourly readings across the day. Economy 7 is a tariff offer that provides much cheaper night time electricity (typically between 11pm and 6am) at the expense of increased day time charges.
The focus of most of the load profiling work was to develop load profiles for archetypical customers in the absence of regular customer meter read- ings. However, with the roll-out of smart meters and other monitoring devices, the focus of the load profiling work can move from building up a profile to exploring the details of the profiles captured by the monitoring devices. To apply future DSM techniques, a more precise splitting of the users into similar groupings is needed thereby allowing for targeting of appropriate groups.
(a)Standard users (b)Economy 7 users
In a report to the DistribuTECH Europe DA/DSM Conference, the Load Research Group of the Electricity Association Services Ltd. [52], set forth a set of general guidelines for load profiles:
• Each profile should represent a relatively homogeneous group of customers.
• Each profile should be distinctly different from the others.
• The identifying characteristics for assigning customer load to a pro- file should be readily determined.
• The number of load profiles should be relatively low.
• The accuracy of estimated load profiles should be judged primarily on how well they perform over a trading period (typically one year) .
Bailey [53] provides a good summary of how load profiles can be built up and distinguishes between methods that produce different kinds of profiles:
• Dynamic profiling from collecting meter data from a subset of cus- tomers.
• Dynamic modelling where knowledge of an external factor (e.g. temperature) is used to modify the standard profile using some formula.
• Same day profiling where a similar day from history is selected to represent the current load profile for a given day (based on weather, overall loading or other criteria).
• Static profiling where a profile is derived for a given season and type of day (e.g. weekend).
• Deemed profiles which are built up from assumptions made about the detail of the load used. For example, a typical household could be assumed to use certain appliances at certain times of the day and the aggregate load profile is built up from the profiles for each appliance.
Swan and Ugursal [54] provides a review of the modelling techniques in use in load profiling. The paper distinguishes between a top down approach which is not concerned with individual household appliances but which apportions calculated usage for a geographic region to indi- vidual households, and a bottom up approach which calculates the usage of individual houses and then aggregates this information to a regional level. The top down approach is criticised as being reliant on historical data and being unable to model disruptive advances in technology (such as the take-up of electric cars). However, the data collection cost of the top down approach is much lower than the bottom up approach with its need to monitor at a much finer level.
Baker and Rylatt [55] investigated the differences between households using questionnaires and then used this information to form clusters. Analysis of the criteria for membership of particular clusters showed the importance of home working as a major influence on the overall load profiles.
Figueiredo et al. [56] worked with data from Portugal for a small number of households for which appliance level measurements had been taken. From this data, the work builds up an aggregation of usage of appliances to create aggregated load profiles. These are then used to assess the effect of simulating altering the time of appliance usage on the overall load profile for a collection of households. This is close to the requirements for effective demand side management as it allows utilities to assess the impact on their supply requirements from changing consumer’s demand patterns. However it relies on detailed monitoring of a few households and would not scale well for reasons of cost and time.
Load profiles form an important part of the requirement within the electri- city industry to perform short term (a few hours) load forecasting. Gross and Galiana [57] provides a review of the current approaches and, in relation to load profiles, concludes that the load generated by a given household is a function of three principal time factors: the season, the weekly/daily cycle and the occurrence of public holidays. Other factors come into play relating to school holidays, daylight saving and weather conditions which can have large short term effects on the longer term patterns.
Cancino [58] provides a review of the literature on load profiling which is defined as the application of methods that deal with customers’ load diagrams with the goal of grouping customers with similar load profiles into coherent clusters.
Most of the published load profile work assumes that the load profile for a given household on a given day of the week and season is relatively constant. In practice, the actual usage on the day will be influenced by ex- ternal factors such as the weather. Lin et al. [59] has studied this in China and used the similarities in how substations react to external factors as the basis for clustering. This work uses information on the peak temper- ature and peak relative humidity. However, this work is less applicable to the UK due to the lack of air conditioning in domestic properties and the wide usage of gas for space heating.
Zakaria and Lo [60] distinguishes between static load profiling, dynamic modelling and dynamic load profiling. Static load profiling relies on the collection of data from a large number of customers for a period of over a year. This data can then be categorised by season and type of day (e.g. weekend) and then averaged to create a load profile for the customer. The paper makes the point that climate is not included in the categorisation and this can be a significant shortcoming. Dynamic modelling makes use of the historic load shapes (as with static load profiling) but also includes a climate adjustment mechanism. The dynamic load profiling method relies on load data being read regularly (daily) with “new” load profiles being produced daily. However this approach relies on the installation of monitoring equipment in each of the households requiring large com- mitments of time and money.
A good example of a bottom up approach has been taken by Ihbal et al. [61]. Using the results of questionnaires and making assumptions about the occupancy of a household at different times of the day and the prob- ability of using a particular appliance, it is possible to build up an overall load profile for the given household. This generated profile can then be aggregated over a population of different types of households (e.g. single person, retired) to create an overall usage profile for a community. The approach is valid but depends on various assumptions on parameters such as occupancy rate, or probability of using a particular appliance, and
relies on extensive survey work to build up realistic estimates for these parameters. The workload (in time and money) means that this approach cannot be scaled to large populations and shows some of the drawbacks of the bottom up approach.
Capasso et al. [62] is a similar example of a bottom up approach to gen- erating a household load profile. The paper uses information on the electricity usage profile of individual appliances and then uses probab- ilistic methods to build up a profile of the usage of particular kinds of appliances and thence to an overall aggregation for the whole household. The modelling includes information on household occupancy and draws on psychological theories. The aggregation is further extended over the many households in an area to generate an aggregation for the district. The model used and the predictions resulting have been validated against the aggregation of 95 households in Milan, Italy. Paatero and Lund [63] also uses a bottom up approach by defining a set of appliances that each monitored household is assumed to use. Each appliance is then assigned a usage profile (i.e. how often and for how long the appliance is used) which are then aggregated to create an overall household load profile. Dominguez-Navarro et al. [64] uses a top down approach to take a load profile for a group of consumers and to use Tabu search to disaggreg- ate this overall demand into that for individual households based on a number of assumptions about the type of electricity usage at different times of the day. This approach attempts to overcome the absence of de- tailed meter data from each household but is an approach that will be superseded by the roll-out of smart meters.