This is to certify that I am responsible for the work presented in this project, that the original work is my own, except as specified in the references and acknowledgments, and that the original work contained herein was not undertaken or done by unspecified sources or persons . The continuing trend towards deregulation and unbundling of transmission services has resulted in the need to measure the flow of energy primarily for. Therefore, tracking methodology was introduced to overcome the problems associated with the marginal pricing of transmission costs.
This study is twofold: the first revolves around the validation of the method and a 3600 analysis of the proportional method leading to a redefined power tracing method; the second is to further refine the proposed prediction method in [1] by identifying trends of the learning coefficients, and use them to investigate the relationship between accuracy and the number of samples taken. The response of individual producers to changes in demand and associated losses are also presented. Perumal consistently providing helpful guidance and advice, his dedication to the weekly advisory sessions, and his 'gracious' handing out of conference papers and reference materials, all of which enabled me to conduct the most diligent literature review I have conducted to date.
Credit to the University for providing access to IEEE and a well-equipped IRC that provided information at my fingertips.
CHAPTER I INTRODUCTION
CHAPTER 2 LITERATURE REVIEW
- The Locational Marginal Pricing (LMP) and Its Shortcomings
Dependence on the location of the slack bus in the LMP method would make the fairness of the costs questionable. Due to the nature of electricity being indistinguishable, each MW leaving the node can be assumed to contain the same share of inflow as the total node flow. The proportion of the ih generator output used to meet the k`h load demand at the time of generation is indicated by the respective k-i element of the AU-1 matrix.
The element ih P= Ad'PL shows the distribution of the 1st node power among all loads in the system. The contribution of the generator k to the load demand ih at the point of consumption is shown by the corresponding element i-k of the matrix Aj'. This is extended to predict (i) the fraction of generations that meet retailer demand, (ii) retailer demand and transaction power loss, (iii) retailer demand and the fraction of in-line partial transactions, and (iv) retailer demand retail and losses relating to each of the transactions in the line.
This equation represents the relationship between the final produced power of the generator and the required power, where Pgd is the dependent variable and Pd is the independent variable. This equation represents the ratio of the transmission loss incurred in serving a given load in the system to the required power, where Pgd is the dependent variable and Pd is the independent variable.
CHAPTER 3 METHODOLOGY
- Tools
The first part of the project is aimed at knowledge of the deregulated energy market, power tracing concept, methods and application. After identifying proportional tracking methodology as the method to build on, its strengths and weaknesses are analyzed, leading to the start of the simplified algorithm, this is detailed in the Results and Discussion section. Starting from the application of the tracking method in the deregulated energy market, topics such as congestion management, ATC and prediction are investigated.
Only after establishing the reliability of M-File, the application of the prediction method through the calculation of learning coefficients is made. The challenge in implementing the learning coefficient method is handling and addressing the M-File data, due to the large amount of learning coefficients that need to be handled, as described in the Literature Review section. Next, short-term forecasting of future demand is done using learning coefficients.
This part includes an in-depth analytical and critical review of the learning coefficient and regression method to design a prediction algorithm, for which no previous references exist. Therefore, the acceptance criterion for the prediction method is set to a MAPE of less than 10%. Finally, all individual subroutines are integrated into a fully automated program that asks the user for the load ID, the week and the hour of the day to be predicted.
Overall project progress is on time, with actual progress faster than planned. Tables 2,3,4,5 were used in conjunction with Table 1 to provide the specific load demand on all loaded buses for a given time of day in a given week of the year. The forecasting program first requires the load ID, week and time to be forecast.
Forecast results with percent error are displayed, along with a graphical representation of the forecast performed. It will continue to perform load flow and tracing, using the transmit and receive algorithm for all hourly demand data used to perform forecasts in the previous section, i.e. Finally, the program will make calculations using the obtained learning coefficients, the estimated values of the generation and the associated losses in meeting the forecasted oncoming demand.
CHAPTER 4
RESULTS AND DISCUSSION
CHAPTER 5
CONCLUSION AND RECOMMENDATIONS
More studies will be done to extend the now able to perform hour-ahead forecasting with lead time of 1 hour to also perform day-ahead forecasting with lead time less than 24 hours [18]. The program should also be tested with actual load demand data to further improve its credibility. Further prospect of reducing the MAPE to a tighter 3% is also envisaged to be very beneficial.
BIALEK: "Trace-Based Unifying Framework for Transfer Pricing of Cross-Border Trades in Europe", Int. AHMAD: "Allocating transmission system usage based on traceable contributions of generators and loads to flow", IEEE Transactions on Power Systems, Vol 13, No. WALLACE: "A methodology for allocating transmission losses due to cross-border trades", IEEE Transactions on Power Systems, Volume 9, Number 3, August 2004.
Report prepared by the Reliability Testing System Working Group of the Subcommittee on the Use of Probabilistic Methods. Congestion risk management using probable congestion value and optimization of operating regimes. A case study for the western part of the Romanian electricity system. SAVAGE: "A Closed-Form Solution for Computing Generator Contributions to Water Loads and Flows in an Open Access Market", IEEE 2003.
ROSADO, KHATIB ABDEL-RAHMAN, NOUREDINE HADJSAID: "Tracing The Path of Electric Power Flow -A Study for Deregulated Power Systems", IEEE 2001.
APPENDICES
APPENDIX A
PROJECT GANTT CHART
APPENDIX B
CODING DOCUMENTATION
This program performs load flow using the Newton-Rhapson method and initial data preparation for power tracking. This subroutine is available for systems of various sizes other than the current IEEE 24 bus system. This subroutine performs the trace function and all related loss distribution and extent using line calculations for all the sample points selected from predict2.
The program only displays relevant results for the charging bus entered in flow_tracingprep_bus8_2. For example, a simulation performed with the load on bus 8 only shows the loss distribution of load 8 and the degree of usage of the line that bus 8 serves. This feature prevents a cluttered results screen and information overload, so results can be quickly analyzed and viewed.
The results are merged into a single matrix and then split into three matrices: final power generation, final power retail, loss. These results are stored to be passed to the learncoef#2 subroutine for generating the coefficients. Each set of data from one sample is incorporated into one large matrix by adding the successive data sets to the current matrix.
One data set from one sample is a 10 x 4 matrix, so the final dimension of the data matrix will be 10 x (4 x # of samples), where each column is 1+ 4n generator ID, and column 2+ 4n is the power supplied from the generator end, column 3+ 4n the power received at the end of the load and column 4+ 4n the loss incurred in this transaction. The logic for (samples - 3) is attributed to the fact that the coefficients are calculated with 4 examples as a basis, so matrix indexing must account for this.