Modern life would be different without electricity. Industries use electricity to power their processes. Commercial establishments use electricity to conduct their businesses. Residential areas use electricity to power their homes and appliances.
As the demand for electricity globally continue to rise the demand for thousands of new power plants using various energy resources also increases. With this, different types of power plants were needed to suffice the demand of electricity.
The Philippine electric power industry was once dominated by the National Power Corporation (NPC) in the generation sector.
All generating power plants were owned and controlled by the NPC; the Independent Power Producers (IPP) were then restricted in directly connecting to the electric distribution utility.
Under the Electric Power Crisis Act of 1993 (R.A. 7648) and the Expanded BOT Franchising Law of 1994 (R.A. 7718), IPP’s were allowed to deal directly with distribution utilities and bypass the NPC grid. And the enactment of the Electric Power Industry Reform Act (EPIRA) of 2001 (R.
A. 9136), the generation has become a competitive segment of the industry. 
 An analysis of the Philippine Electric Power Industry
WESM was established as a provision of the Electric Power Industry Reform Act of 2001 (EPIRA Law) to provide the mechanism for determining the price of electricity not covered by bilateral contracts between sellers and purchasers of electricity. It aims to create fair and transparent competition in electricity generation. Its purpose is to create reliable price signals to assist participants in weighing investment options and provide as well as maintain a fair and level playing fields for suppliers and buyers of electricity.
The Philippines Wholesale Electricity Spot Market (WESM) is the venue for trading electricity as a commodity in the spot market.
Evolution of power systems have been evident during the past century. The electric grid is getting more and more refined due to modern technologies and business requirements, such as the implementation and integration of smart grid technologies, placement of ultra-high voltage transmission systems, and assimilation of ultra-high levels of renewable resources, to name a few. All of these factors are stimulating today’s energy forecasting practice. (
Load forecasting can be defined as the prediction of future electrical demand. Since the power generated must be equal to that consumed, efficient management of electrical power generation is of importance. The term load forecasting refers to the projected load requirement determined using past knowledge to define future load in sufficient quantitative detail . Demand electric load prediction is an important aspect in electricity planning  of electric utilities. It is a means of knowing how much power is to be scheduled for the future demand, useful in drawing up a feasibility report for generating power plants and is also useful in determining whether the current capacity of the power stations or sub-stations of a service area will be surpassed.
 Singh A, Kalra P.K, Emmanuel, “Point of view of knowledge based systems for load forecasting”, Symposium on expert systems application to power systems, 1998, pp 7~1- 7~4.
 Taylor W. J, McSharry E.P, de Menezes M.L, “A comparison of univariate methods used for forecasting electricity demand up to a year ahead”, International Journal of Forecasting,2006, Vol.22, pp 1-16
Under normal circumstances, depending upon the forecasting purpose and period, load forecasting can be classified as:
Short-term load forecasting refers to the prediction of electrical demand which range from a few minutes to a week forecast. Medium-term load forecasting is usually range from a week to a year ahead while long-term load forecasting is for a longer duration which can last up to 20 years. 
The short-term load forecast period is generally 24 or 48 hours to a week. This is used to predict the load capacity of the region or the daily or weekly consumption data. The forecasting data generally indicates daily or monthly.
Short-term load forecasting is vital for the economic generation dispatch of power systems; meaning the power generation can be planned according to the forecasted value in order to decrease the risk of equipment failures (e.g. transformer and transmission lines). The possibility of occurrence of blackouts is also reduced as well as losses in revenue of power utilities.
The following are the key objectives of this forecast:
Medium -term load forecasting is prediction of demand for a week up to a year. The prediction usually targets the load capacity on a region, or the monthly electricity consumption. The forecast data generally indicates cyclical and each month of each year consists with the similar growth pattern. The prediction is useful in the arrangement of monthly maintenance plan, operation mode, reservoir operation plans and coal transportation plans. Peak load demand is the main focus of this forecast but off-peak demand is also forecasted.
This type of forecast is for much longer duration. The forecast period may last up to twenty years or more depending upon the preferred forecast duration. The main purpose of this forecast is to determine the annual peak system load. The prediction is useful in providing the base data for the power grid planning that will help in determining the grid operation mode and annual maintenance plans.
In order to develop a load forecast model, it is important to determine all the factors which affects the electric load demand.
The factors that needed to be considered are:
The following are regarded as the most important factors.
Weather is the most important variable in load forecasting. Considering weather parameters in a specific region in order to develop an accurate forecast. The effect of weather is most prominent for domestic and agricultural consumers, but can also alter the load profile of industrial consumers. Weather data is usually incorporated in short-term load forecasting.
Electricity Load Forecasting for Urban Area Using Weather Forecast Information, 2016 IEEE International Conference on Power and Renewable Energy, Vasudev Dehalwara, Akhtar Kalama, Mohan Lal Kolheb Aladin Zayegha, pp. 355-359.
The following are the most important weather factors:
Temperature is defined as the measure of the average kinetic energy of atoms or molecules of an object. It can also be defined as the degree of hotness or coldness of the body. Temperature is said to have a direct effect on the load. In the Philippines, two seasons are experienced; summer and rainy seasons. During summer, increase in the temperature will lead to increase in the load. This is due to the increased usage of air conditioning units and electric fans. And the opposite will be true during rainy seasons.
Humidity is the term used for the amount of water vapor in air. Humans are sensitive to humidity because the mechanism used to regulate the body temperature is evaporative cooling. At high humid atmosphere the rate of evaporation through skin (perspiration) is lower than it would be under normal conditions. Since human perceives rate of transfer of heat rather than temperature itself, so we feel warmer at high humid conditions.
Thus humidity can increase the feeling of the severity of temperature and make people to use more cooling appliances therefore due to this fact daily load curve will show high value during humid day .
 Muhammad Usman Fahad and Naeem Arbab, Factor Affecting Short Term Load Forecasting, Journal of Clean Energy Technologies, Vol. 2, No. 4, October 2014.
Precipitation is defined as the amount of rain, snow or hail fallen at a specific place within a specific period of time.
Precipitation can affect load consumption directly and indirectly.
Heavy rain or snow can make people to stay home and it can cause darkness. So due to the fact that people will be forced to stay indoors, they will consume more electricity for lighting and entertainment purposes.
By indirect effect we mean that heavy rain or snow can decrease the temperature thus may have positive or negative effect on load consumption. By negative we mean load consumption will increase and by positive we mean decrease in load consumption .
 A. Azadeh, M. Saberi, and O. Seraj, “An integrated fuzzy regression algorithm for energy consumption estimation with non-stationary data: A case study of Iran,” Energy, vol. 35, pp. 2351-2366, Jun 2010.
Population also plays an important role in electric load forecasting. Since everyone is using electricity, the demand increases as the population increases. At present, the population of the Philippines has reached to 107,392,179 which accounts to about 1.4 percent of the world’s total population.
Department of Economic and Social Affairs, Population Division. World Population Prospects: The 2017 Revision.
Electricity pricing also have a significant effect on the usage of electricity. When the cost of electricity is high, the consumers tend to limit the usage of appliances, thus the price of electricity is variable. 
Long Term Hourly Peak Demand and Energy Forecast, Electric Reliability Council of Texas, Inc., Taylor, TX, 2010, pp. 9.
Artificial Neural Networks (ANN) are electronic models and software implementations  which is based on the functions of a human brain. A neural network is a non-linear circuit which is capable of performing non-linear curve fitting.
The brain consists of neurons, which are interconnected by dendrites and collect information via the connection. This provides human to think, remember and apply previous experiences to every action and decision.
An introduction to neural networks for beginners. Adventures in Machine Learning. Dr. Andy Thomas, undated
The neuron is the fundamental processing element of a neural network. This receives inputs from sources, combines them, process the information and outputs a result.
Natural neurons have four basic components which are known by their biological names – dendrites, soma, axon and synapses. Dendrites are hair-like extensions of the soma which act like input channels. These input channels receive their input through the synapses of other neurons. The soma then processes these incoming signals over time. The soma then turns that processed value into an output which is sent out to other neurons through the axon and the synapses.
A neuron can be represented by a mathematical model.
The mathematical model is in the form:
The most common simple neural network structure consists of an input layer, a hidden layer and an output layer. The input layer is where the external input data enters the network. The hidden layer are internal layers which contain many of the neurons in various interconnected structures. Neural networks can have many hidden layers. The output layer contains the desired output.
The moment a network has been fully structured for a particular application, it is now ready to be trained. There are two types of training approaches used in ANN – supervised and unsupervised.
Supervised training involves a mechanism which provided the network with the desired output either by manually grading the performance of the network or by providing the desired output with the given inputs. Unsupervised training is where the network has to make sense of the input without outside help. It is based on the provided local information and there is no particular target output.
In a supervised training, the input and the output are provided. Then the network will compare the results against the desired output. The system will adjust weight which controls the network due to the error propagated back through the system. This process will occur repeatedly as the weights are continually tweaked. The set of data which enables the training is called the “training set.” During the training of a network, the same set of data is processed many times as the connection weights are ever refined.
In this learning mode, the network is trained without the help of external factors. The data presented only consists of inputs and there is no target values for the outputs. The system will decide what features it will sue to group the data inputs. The weights of the system are adjusted by the network itself by monitoring the internal performance. The network will look for trends in the input signals and make changes according to the chosen functions.
Recurrent neural network (RNN) is a type of neural network which is designed to recognize patterns in sequences of data (handwriting, texts, spoken words, genomes, or numerical time series data). These algorithms take sequence and time into account and have a temporal dimension.
In a RNN, the information cycles through a loop allowing information to persist. They are used to model time dependent data. . It makes a decision by taking into account the current input and what has been learned from the input it previously received. The information is then fed to the network one by one and the nodes in the network store their state at one time step and use it to inform the next time step.
 Dorffner, G. Neural Networks for Time Series Processing. Neural Netw. World 1996, 6, 447-468.
 Electricity Price Forecasting Using Recurrent Neural Networks, Umut Ugurlu , Ilkay Oksuz ID and Oktay Tas 1 I, Energies 2018, 11, 1255; doi:10.3390/en11051255