The introduction of deregulation and subsequent open access policy in electricity sector has brought competition in energy market. Allocation of transmission loss has become a contentious issue among the electricity producers and consumers. A closed form solution for transmission loss allocation does not exist due to the fact that transmission loss is a highly non-linear function of system states and it is a non-separable quantity. In absence of a closed form solution different utilities use different methods for transmission loss allocation. Most of these techniques involve complex mathematical operations and time consuming computations. A new transmission loss allocation tool based on artificial neural network has been developed and presented in this thesis. The proposed artificial neural network computes loss allocation much faster than other methods. A relatively short execution time of the proposed method makes it a suitable candidate for being a part of a real time decision making process. Most independent system variables can be used as inputs to this neural network which in turn makes the loss allocation procedure responsive to practical situations. Moreover, transmission line status (available or failed) was included in neural network inputs to make the proposed network capable of allocating loss even during the failure of a transmission line. The proposed neural networks were utilized to allocate losses in two types of energy transactions: bilateral contracts and power pool operation. Two loss allocation methods were utilized to develop training and testing patterns; the Incremental Load Flow Approach was utilized for loss allocation in the context of bilateral transaction and the Z-bus allocation was utilized in the context of pool operation. The IEEE 24-bus reliability network was utilized to conduct studies and illustrate numerical examples for bilateral transactions and the IEEE 14-bus network was utilized for pool operation. Techniques were developed to expedite the training of the neural networks and to improve the accuracy of results.

The introduction of deregulation and subsequent open access policy in electricity sector has brought competition in energy market. Allocation of transmission loss has become a contentious issue among the electricity producers and consumers. A closed form solution for transmission loss allocation does not exist due to the fact that transmission loss is a highly non-linear function of system states and it is a non-separable quantity. In absence of a closed form solution different utilities use different methods for transmission loss allocation. Most of these techniques involve complex mathematical operations and time consuming computations. A new transmission loss allocation tool based on artificial neural network has been developed and presented in this thesis. The proposed artificial neural network computes loss allocation much faster than other methods. A relatively short execution time of the proposed method makes it a suitable candidate for being a part of a real time decision making process. Most independent system variables can be used as inputs to this neural network which in turn makes the loss allocation procedure responsive to practical situations. Moreover, transmission line status (available or failed) was included in neural network inputs to make the proposed network capable of allocating loss even during the failure of a transmission line. The proposed neural networks were utilized to allocate losses in two types of energy transactions: bilateral contracts and power pool operation. Two loss allocation methods were utilized to develop training and testing patterns; the Incremental Load Flow Approach was utilized for loss allocation in the context of bilateral transaction and the Z-bus allocation was utilized in the context of pool operation. The IEEE 24-bus reliability network was utilized to conduct studies and illustrate numerical examples for bilateral transactions and the IEEE 14-bus network was utilized for pool operation. Techniques were developed to expedite the training of the neural networks and to improve the accuracy of results.