Literature Review Of The Most Effective Load Forecasting Methods

Load forecasting is a unique problem in that it has many possible ways of getting a solution. There is no single approach that can be deemed to be the best. The most effective forecasting method is the one that provides a prediction that leads to the correct decisions and choices being made.

Forecasting the future load can be classified into four main categories which are: very short-term, short-term, medium-term and long-term load forecasts. Very short-term load forecasting (VSTLF) is performed in a very short time that is from one minute to a few minutes ahead. The determination of VSTLF is immensely important for the economic dispatch activities and Area Control Error (ACE) estimation. The economic operation most power systems depend largely upon is economic dispatch. Whilst trying to keep the system reliability in reasonable tolerances, the forecasted system load must be met at the lowest possible cost, this is one of the key result areas in power system operation. ‘ACE is the difference between scheduled and actual electrical generation within a control area on the power grid, taking frequency bias into account. ’

The short-term load forecasting (STLF) is performed ranging from one hour to one-week ahead. The STLF is required for improvising day-to-day optimal decisions on the economic and secure operations of a power system. The economic operations of a power system involves scheduling of generating capacity, scheduling of fuel purchases, power transfer analysis and planning of market based power transfer. On the other hand, STLF is also applied to the system security assessment that takes into account system contingency. The medium load forecasting is covering in the range from a few weeks to a few months ahead. It is required mainly for fuel allocation and maintenance scheduling.

The long-term load forecasting is performed ranging over a period of five years. The long-term load forecasting is important for the system expansion planning. The factors that influence electricity demand are the weather, socio-economic and demographic variables. Weather factors that should be considered are ambient temperature, both maximum and minimum; wind speed and rainfall or snow. Gross Domestic Product (GDP), Consumer Price Index, economic growth and population distribution and growth are the socio-economic factors that are important in demand forecasting. The historical data is of a time series nature. It is important to have good data to get a measure of how these factors have been changing over time. Some of the changes are in the form of a time line while some exhibit cyclical behaviour. The cycles may be daily, weekly, monthly and seasonal. The cyclical patterns are useful in the models for electricity demand forecasting.

A statistical approach using a regression model was done. The aim of the study was to construct an effective simplified econometric model that can be used to forecast the Peak Demand of electricity for Zimbabwe from 2011 to 2015. The objectives of the study were to measure the effect of historical socio-economic parameters from 1980 to 2011. Econometrics is the unified study of economic models, mathematical statistics and socio-economic data. Three approaches were used in the construction of the regression model. The use of three different software environments, MATLAB, SPSS and Excel confirmed that they all have sufficient statistical capability to carry out reliable modelling. Results from MATLAB were most preferable because of their simplicity. The correlations investigation revealed that there was high co-linearity with a Pearson correlation significant at the 0. 01 level (2-tailed) for: Peak Demand with Population and GDP per capita, Maximum Temperature, Minimum Temperature and Temperature Range. The results of the study in bringing about a reasonable forecast could be considered a success. The forecast could be used to evaluate Peak Demand for the period it had been carried out. There were however uncertainties as to how far the model explained the results. It was therefore safe to assume that while the forecast was within the range of other forecasts the contemporaneous nature of the variables resulted in the spurious correlation on which the model was built.

As part of continuous monitoring and improvement, Eskom went through a process to change its planning operation in 2007. Part of the business improvement was that Eskom Distribution transitioned into a new method called the Geographical Based Load Forecast (GLF). Prior to the year 2007, Eskom Distribution followed a legacy method of forecasting that was not structured. The method was based on collecting customer applications, load trending, and relied on the planner’s knowledge of the area and was based on Microsoft Excel. On seeking to improve its forecasting approach, the utility adopted a technique that was based on spatial forecasting. The technique was called a geographically based load forecasting (GLF) technique, which was performed using a special computer based tool called PowerGLF. The GLF technique is a derivative of a spatial forecast method. It combines the land-use forecast and Gompertz-curve fitting algorithms on a small area basis. As it has been especially derived for the South African utility planning environment, the GLF technique has the so-called customer based forecast algorithm, which is a manual capturing of customer applications and the conversion of these into a load forecast. The set of customised growth curves that align to different economic sectors and the constant percentage growth algorithms are used to align the electric forecast to the econometrics study. Specifically in the case of the GLF technique, the computer based tool called PowerGLF is used to perform the forecast. PowerGLF is a tool mainly used to store the forecast; the actual forecasting is performed, mostly, manually. The GLF forecast is characterised by three main components: load position forecast (where), load magnitude (how much), the anticipated period (when) and load class (what).

In the United Kingdom (UK), the National Grid is currently working with the DeepMind, a Google-owned AI team, which is used to predict the power supply and demand peaks in the UK based on the information from smart meters and by incorporating weather-related variables. This cooperation tends to maximize the use of intermittent renewable energy and reduce the UK national energy usage by 10%. Therefore, it is expected that electricity demand and supply could be predicted and managed in real time through deep learning technologies and machines, optimizing load dispatch, and reducing operation costs.

At the grid level, to minimize the energy wastage and make the power generation and distribution more efficient, the future of the power grid is moving to a new paradigm of smartgrids. Smart grids are promising, unprecedented flexibility in energy generation and distribution. In order to provide that flexibility, the power grid has to be able to dynamically adapt to the changes in demand and efficiently distribute the generated energy from the various sources such as renewables. ANN’s provide an attractive alternative tool for both forecasting researchers and practitioners because of the following features that make them attractive for a forecasting task:

• As opposed to traditional methods, ANN’s are data driven self-adaptive methods i. e. there are a few apriori assumptions about models for problems under study.
• They learn from examples and capture suitable functional relationships among data even if the underlying relationships are unknown and hard to describe.
• They are suitable for problems whose solutions require knowledge that is difficult to specify but for which there are enough data or observation.