FORECASTING MODEL OF ENERGY CONSUMPTION USING LEAST SQUARES SUPPORT VECTOR MACHINES

In the oil and gas industry, accurate measurement of gas is a very important aspect for the gas transmission operation. The outgoing gas flow during the transmission operation is monitored and maintained by a metering system. The metering system must be ensured reliable and dependable at all cost to maintain the billing integrity between distributors and customers. The main concern is products sold and returned as money worth product to seller and buyer. An existing system in Transmission Operation Division (TOD), PETRONAS Gas Berhad (PGB), Gurun is held responsible to calculate the energy consumption from the sales gas produced. The system consists of a turbine meter, measuring equipment which are pressure transmitter and temperature transmitter, gas chromatography and flow computer. However, the system is a standalone system that does not have any reference system to verify the integrity of it. Customers are billed according to the amount of energy consumption calculated and any error in calculation will cause loss of profit to the company and affect PETRONAS’s business credibility. As a solution, a Least Squares Support Vector Machines (LS-SVM) prediction model of energy consumption is proposed as a verification system of the outgoing gas flow. The model will predict the energy consumption and compare it with the results of the existing metering system to ensure the reliability and accuracy of the system. The billing integrity between PETRONAS and the customers could be maintained and in the future if the project is expanded, it will have the potential of saving of millions of dollars to Malaysian oil and gas companies.