Load forecasting in the long-term planning

The delivery of electric power is a capital intensive business. Electricity distribution facilities need land for power line paths and substation sites, power equipment for distribution, protection, and control, and labour. The arrangements and plans for new or extended infrastructure may require several years. Network planning is a decision-making process that aims at identifying the best schedule of future resources and actions.

Financial aspects, in other words, minimizing cost and maximizing profit, service quality and reliability, environmental impacts, public image, and future flexibility are common objectives in network planning. The objective of the planning process is to meet the future electric demand with an acceptable level of reliability. Basically, this includes determining the sizes, locations, interconnections, and timing of future grid extensions, and commitment to DER such as distributed generation (Willis, 1996).

Energy consumption and power forecasts play a crucial role in electricity distribution.

Energy and power forecasts yield information about the development of the future electricity distribution environment. Thus, the DSOs can also prepare for the future challenges in advance. If the energy forecasts are erroneous, the effects on the distribution revenue can be detected swiftly, at least in the profit and loss account. In the long term, an incorrect energy forecast can cause problems in the management of the distribution business.

One of the main objectives of the electricity distribution planning is to make distribution networks work in the most efficient way. Load forecasting provides important information for the electricity network planning, and it essential for the electricity system development. The objective is to produce information of the required primary substation capacity, information for planning of feeders, distribution transformer areas, and preliminary information for field planning. New electricity distribution networks are being built and existing networks are renovated. Furthermore, there has to be a plan in which order and within which time period these network investments will be carried out (Willis, 1996). In practice, this means that there is a need for power and energy forecasts.

If the power forecasts fail, this will result in either over- or underdimensioning of the network. From the planning perspective, the network should not be over- or underdimensioned; overdimensioning will cause extra costs in the investment phase while underdimensioning will lead to extra costs afterwards. A need for forecasting does not disappear even if loads do not grow in a certain region. This kind of an area can yield information of the network capacity to be released. Network loads have one of the greatest impacts on network planning. Figure 2.10 presents long-term load forecasting objectives.

Figure 2.10. Simplified transmission and distribution planning process. All steps are based on forecasts of future loads (Willis, 1996).

Electricity load forecasting is dissimilar in different operating environments. For instance, national electricity consumption is typically forecasted by applying different methods compared with electricity distribution. Load forecasting can be divided into small- and large-area load forecasting in distribution systems. A small area typically refers to local distribution levels while a large region covers the whole capacity in a regional system (Willis, 1996).

Electrical load forecasts have traditionally been based on the DSO’s own historical information of the electricity consumption obtained by trending and simulating. The DSO’s external data such as land-use plans have also been taken into consideration (Rimali et al., 2011). Annual energy consumption has traditionally been the basis for the long-term load forecasting in electricity distribution. In this context, regional forecasts have been a prerequisite for the electricity distribution. A starting point for forecasting has been the information of the present building stock and customers’ electricity consumption data. In general planning, quite a rough distribution of geographical areas has been made. In rural areas, a municipal region has typically been a suitable unit for planning while town districts have served the same purpose in urban areas.

In region-specific forecasts, the sizes of customer groups and characteristic consumptions are the most essential elements. The size of the customer group can be determined, for example, by the number of customers and employees in industries or the building area.

The characteristic consumptions are typically given as MWh/dwelling/a or MWh/place of work/a. By multiplying the number of customers by the characteristic consumptions, the total consumption in a customer group can be determined. Summing up the consumptions of all customer groups, the total consumption in the area can be calculated (Lakervi and Partanen, 2008). Table 2.2 shows an example of a regional consumption forecast.

2.1 Electricity distribution business 37

Table 2.2. Example of a forecasting process where the electricity consumption has been forecasted by adopting a regional and customer-group-based electricity consumption approach (adapted from Lakervi and Partanen, 2008).

Year 0 10 20

Population 11 700 11 900 12 200

Number of the residential customers 4 135 4 540 5 080

Characteristic consumption MWh/dwelling/year 4.2 4.7 5.2 Consumption of residential customers, MWh/year 17 400 21 300 26 400 Number of customers with electric space heating 650 1 000 1 750

Characteristic consumption MWh/year 17.1 17.6 18.1

Consumption of electric space heating, MWh/year 11 100 17 600 31 700

Number of farms 415 400 370

Characteristic consumption MWh/farm/year 5.6 6.7 8.0

Consumption of farms, MWh/year 2 300 2 700 3 000

Number of employees in industries 1 350 1 400 1 475

Characteristic consumption MWh/employee/year 6.1 7.0 8.2

Consumption, MWh/year 8 200 9 800 12 100

Total consumption MWh/year 39 000 51 400 73 200

The forecasts of these volumes can be based for example on analyses by Statistics Finland. For instance, the number of building types or the level of livelihood can be estimated; municipal registers also provide valuable information for the purpose. These forecasts can include, for instance, plans for new buildings and employment (Lakervi and Partanen, 2008).

Region-specific forecasts have traditionally been made by the DSOs while national forecasts are typically used as a basis for spatial forecasts. Spatial characteristics have been taken into account in annual energy consumptions of the each customer group and by using the spatial population and employment forecasts in the case area. In network planning, the spatial energy forecasts have to correspond to the present or planned supply areas. Energy forecasts can be transformed into power forecasts by applying load models (Lakervi and Partanen, 2008). The above-presented methodology is quite widely used in the Finnish distribution environment. However, the long-term load forecasting process may vary significantly between different DSOs in Finland. To sum up, the basic concept has been to first prepare the electrical energy consumption forecasts, which have then been transformed into load forecasts by applying load models.

Forecasting typically involves many uncertainties, which make forecasts inaccurate.

Nevertheless, load forecasts are needed for network planning and operation of distribution networks. Electricity distribution networks are different from each other. Some

distribution networks may consist of a large number of customers in urban areas while others may have a small number of customers, for instance in rural areas. Therefore, forecasts and analyses have to be made for each case individually. An electricity distribution network consists of different levels: customers, low-voltage nodes, distribution substation service areas, medium-voltage nodes, distribution feeders, and primary substations.

Load forecasts are needed all the way in the distribution networks, and the long-term load forecasting has to be able to analyse different distribution network levels. The forecasting can be divided into subcategories: specific methodologies are needed for large and small distribution network areas. Large distribution network areas may cover network levels from the primary substation to feeder levels and distribution substation service areas.

Again, small distribution network areas may include network levels from the distribution substation service area to the customer level. Figure 2.11 presents an example of the distribution network levels and the large and small distribution network areas. The highest mean hourly power is the most interesting factor, because peak powers determine the dimensioning of the network in the long term.

Figure 2.11. Distribution network levels. The red circle indicates a large distribution network area and the blue circle a small distribution network area. Adapted from (Seppälä, 1996).

Substation feeder

Primary substation Distribution feeders

MV network 1 (open loops) Remotely operated disconnectors MV network 2 (radial)

Distribution feeder substation

LV fuse LV line

LV line to the customer

Main fuses

Energy metering

110/21 kV

20/0.4 kV