Energy-Forecast

Basic purpose of the Energy-Forecast system – the analysis and forecasting of prices of electricity. Forecasts are based on historical data and are under construction using optimal mathematical algorithms. Forecasts can be constructed with different detailing (hourly - day weeks) on different cuts of the electricity market.

Kernel of the Energy-Forecast system is the STATISTICA system that allows the user to apply powerful analytical methods within one packet.

Application of a system

The electricity market demands from all participants of drawing up forecasts, both on short-term, and for a long-term outlook today.

So, the consumer of the electric power for signing of the contract on delivery should provide the forecast of hourly consumption for the week ahead and adjust it not later than two days. The seller is faced, on the other hand, by a problem of forecasting of prices of electricity for the near-term outlook. And hourly forecasts of the prices of the next week, on days – for the next two weeks, on weeks – for the next four-five weeks are necessary.

It is also necessary to note that the price is not homogeneous across the territory of the Russian Federation and the countries of importers. Therefore the idea of trade hubs (a hub the Center – Europe, a hub the Center – the West, a hub Volga and so on) and cuts which represent uniform specialized spaces for trade in the electric power is used.

Respectively, the task comes down to creation of forecasts with different detailing (hourly-daily-weekly) for all cuts of the electricity market. It is obvious that the expert in the field of prices of electricity needs the automatic system allowing to build forecasts and to make corrections.

General description of a system

Systems capabilities of "Energy-Forecast":

• automatic search of predictive models (ARPSS, exponential smoothing), comparison of errors on a cross-test set and creation of the forecast of the prices of the set planning horizon,
• automatic aggregation of data,
• automatic filling of admissions and processing of emissions,
• calculation of characteristic profiles of behavior of a row for the selected detailing,
• visualization of dynamics of time series,
• reports generation about cleaning of data and forecasting,
• import of historical data from a DB.

The Energy-Forecast system allows the user comprehensively and to systematically analyze historical data on prices of electricity. It is based on the modern principles of the analysis and forecasting of time series and uses the ARPSS model, model of exponential smoothing, calculates average profiles of behavior of the price.

A system allows to carry out aggregation of data in a dialogue mode and to build forecasts for any combination of time series with the selected forecasting horizon.

So, the user selects type of aggregation of data: hour, day, week there is also a forecast:

• on a certain number of steps forward,
• on the set date and time,
• on a certain interval.

Besides, the user can limit the volume of the historical data used for selection of model and creation of the forecast.

Analysis results and forecasting are represented in an interactive dialog box both in tabular, and in a graphic type. In addition, a system generates reports on cleaning and data translation and also on the constructed forecasts. The program code of a system is written in language of Microsoft Visual Basic.

The STATISTICA Advanced program is necessary for system operation of "Energy-Forecast".

Initial data

For work data on prices of electricity in the form of time series on each cut of the market are required for a system. In addition to the prices of specific date and time the data file may contain additional parameters: indicator of winter/daylight saving time, type of day (worker/day off) and type of a zone.

Data can be imported from the following formats:

• MS Excel
• Lotus (1-2-3, Symphony)
• Quattro Pro
• xBase (DBase, FoxPro, etc.)
• Paradox
• ASCII
• Connection to the DB supporting the Microsoft ODBC interface:
  • MS Access
  • MS SQL Server
  • Oracle
  • Sybase
  • IBM DB2/2, DB2/6000

Results of the forecast can be exported to the same formats, or to the text report.