Development of a software application utilising classical efficiency theory, regression and Data Envelopment Analysis in the evaluation of thermal power plant performance.

[摘要] ENGLISH ABSTRACT: Recent capacity constraints on the South African power grid, coupled with the economic andenvironmental implications of increasing energy requirement, has given rise to major efforts toimplement energy management initiatives in the industrial, commercial and residential load sectors.These efforts are supported by the construction of multiple new power plants, both thermal andrenewable in nature. Additionally, the Energy Efficiency (EE) of existing plants is being optimised,which requires accurate performance evaluation and benchmarking as part of plant diagnostic andMeasurement and Verification (M&V) exercises.Energy management exercises require accurate tracking of power plant efficiency. In this project aSouth African coal-fired power plant is used as a test case, and is analysed utilising both classical andData Envelopment Analysis (DEA) based EE evaluation methods in an attempt to track plantefficiency over time and in relation to similar US plants. DEA is a non-parametric linearprogramming-based benchmarking technique used to comparatively evaluate multiple peerbranches. The historical plant data used in this project is provided in monthly intervals, but is of lowquality, with measured fuel consumption values out of sync with actual fuel consumption values. Forthis reason data averaging is also considered. A software application is developed to analysehistorical plant data, supported by the development of a relational database. This database allowsfor permanent storage and access of historical plant data while the software applicationincorporates all relevant analysis methodologies and graphic user interface.The classical efficiency evaluation methods are found to provide a general overview of actual plantperformance, but do not consider plant context, often making results ambiguous. The methods arealso limited to energy datasets, and cannot incorporate additional factors that may be relevant toplant performance. Higher quality data is recommended to increase the accuracy of results.M&V interventions include an energy audit before and after an EE implementation. Preimplementationdata is referred to as the baseline and is used to evaluate the positive impact of theimplementation. Regression analysis is investigated as a means of gaining additional insight into theeffect of additional factors on overall plant efficiency, but also as a means of baseline adjustment inan M&V context. The regression analysis study does not produce significant results, but increasingthe quality of measured plant datasets may allow for more useful results.The DEA efficiency tracking methodology is found to be of use when additional factors areincorporated with energy data, and can provide a brief overview of performance between plants.When a single plant is evaluated over time the process can also easily identify inefficient periods, although additional insight is required to establish the sources of these inefficiencies. DEA is thus nota complete replacement for classical EE methods, but rather a useful supplementary tool inefficiency evaluation. The accuracy of its results is highly susceptible to the quality of data used.Evaluation of individual plant component inputs and outputs rather than overall plant inputs andoutputs would make for a useful future study.

[发布机构] Stellenbosch University