National Science Center Kharkov Institute of Physics and Technology FORECAST COST ASSESSMENT FOR HTGR RPVs FOR 2020-2030 Dr. Andrii Odeychuk Technical Meeting on the Economic Analysis of High Temperature Gas Cooled Reactors and Small and Medium Sized Reactors, 25-28 August 2015, IAEA, Vienna, Austria
OVERVIEW Introduction Materials science and forecast approach to predicting the cost of nuclear power plant components Forecast cost assessment of HTGR RPVs for 2020-2030 Conclusions
INTRODUCTION AND RELEVANCE OF RESEARCH Design, construction and operation of Nuclear Power Plants in energy sector are associated with rather large capital costs. Fig. 1 Basic Gas-Cooled Reactor The main price loading in case of implementation of the specific project of nuclear power plant is born by the principal components of this installation: reactor vessel, reactor internal elements, piping, etc. which are generally made of radiationresistant metal alloys. Forecast assessment of the cost and world production volumes of specific components has a considerable influence on the justification of including those components in an alloy (represented in the time series form). Development of approach to cost estimate of design reactor vessels is current interest.
MATERIALS SCIENCE AND FORECAST APPROACH TO CHOOSING A REACTOR VESSEL A materials science and forecasting approach entails the following: Collection of basic data: Reactor vessels parameters; Reactor vessel steels composition; Historical price of the components included in the vessel steels represented as time series; Forecast of the price of vessel steels components; Cost estimate calculation of vessel steels and reactor vessels using the received forecast; Carrying-out of an analysis and drawing conclusions and recommendations. 2015 2030 Time
RPV Composition of steel, % COST ESTIMATION OF HTGR RPVs FOR 2020-2030 ON THE BASIS OF MATERIALS SCIENCE AND FORECAST APPROACH Table 1 The composition of nuclear reactor pressure vessel steels Design GT-MHR (General Atomics) ANTARES (AREVA) PBMR (Westinghouse) Weight, ton 1328 1) 825 2) 950 3) Material (ASME) A 336 Grade F22 A 336 A 336 A 508 Class 1 4) Grade F22V 4) Grade F91 4) Grade 3 Class 1 2) Al 0,04 0.025 Cr 2 2.5 2 2.5 8 9.5 0.25 Сu 0.2 0.2 Fe 95.07 96.13 94.7 95.8 87.32 89.73 95.63 96.68 Mn 0.3 0.6 0.3 0.6 0,3 0.6 1.2 1.5 Mo 0.87 1.13 0.9 1.1 0,85 1.05 0.45 0.6 Nb 0.07 0,06 0.1 0.01 Ni 0.25 0,4 0.4 1 Si 0.5 0.01 0,2 0.5 0.4 Ti 0.03 0.015 V 0.25 0.35 0,18 0.25 0.05 Residual 0.2 0.19 0,24 0.32 elements 1) Gas turbine-modular helium reactor (GT-MHR).Conceptual design description report // General Atomics. 1996. 198 p. 2) Mizia R.E. Next generation nuclear plant reactor pressure vessel acquisition strategy // Idaho National Laboratory. Idaho Falls, Idaho 83415. 2008. 65 p. 3) Pebble bed modular reactor (PBMR) - A power generation leap into the future // Mr Thinus Greyling, Pebble Bed Modular Reactor (Pty) Ltd, South Africa. 2011. 5 p. 4) Handbook of comparative world steel standards // John E. Bringas, editor. 2004. 663 p.
DATA PREPARATION 1) Federal Reserve Bank of Minneapolis http://www.minneapolisfed.org CPI P P (1) CPI 2010 k/2010 k, k where P k the cost of the normalized component k-th years, CPI inflation index in the USA 1)
TIME SERIES FORECASTING (NEURAL NETWORK) y(k) [3] [3] W w ji W w ji W w j u 1 1 o 1 u 1 1 o 1 z 1 y(k-1) u 2 u n 1 2 n 1 o 2 o n 1 u 2 u n 2 2............... z L y(k-l) n 2 o 2 o n 2 [3] u 1 [3] 1 [3] o1 y( k 1) First hidden layer ( L+1 neurons ) Second hidden layer ( 2(L+1)+1 neurons ) Output layer Fig. 2 Diagram of the multi-layer perceptron with time delay line [3] [3] y( k 1) ( W ( W ( W x))), (2) where n s x y( k), y(k1),..., y(k L) T [s] [ s] [ s] [ s] [ s] [ s1] input vector; o j j u j j wji oi ; i0 k current moment of time; s 1,2,3; j 1,2,...,n ; L value of delay; W s w weight matrix of s-layer; [ s] [s] ji ( n 1) ( n 1) s s1 n number of neurons in s-layer; u n s w o [ s] [ s 1] j ji i i0 [ s] ; sigmoid activation functions of s-layer. s
COST ESTIMATION OF HTGR RPVs FOR 2020-2030 ON THE BASIS OF MATERIALS SCIENCE AND FORECAST APPROACH (continuing) Al Cr Cu Fe Mn Mo Statistical data of National minerals information center of U.S. Geological Survey and London Metal Exchange became the information basis of research.
COST ESTIMATION OF HTGR RPVs FOR 2020-2030 ON THE BASIS OF MATERIALS SCIENCE AND FORECAST APPROACH (continuing) Nb Ni Si Ti V Mean absolute percentage error is about 19 %
U.S. dollars (2010) per ton COST ESTIMATION OF HTGR RPVs FOR 2020-2030 ON THE BASIS OF MATERIALS SCIENCE AND FORECAST APPROACH (continuing) 50000 Reactor pressure vessel steel components 40000 30000 20000 10000 2014 2020 2030 0 Al Cr Сu Fe Mn Mo Nb Ni Si Ti V Reactor pressure vessel steels A 508 Grade 3 Class 1 A 336 Grade F91 A 336 Grade F22V A 336 Grade F22 Class 1 2014 2020 2030 0 500 1000 1500 2000 2500 3000 U.S. dollars (2010) per ton
U.S. dollars (2010) per ton COST ESTIMATION OF HTGR RPVs FOR 2020-2030 ON THE BASIS OF MATERIALS SCIENCE AND FORECAST APPROACH (continuing) 3000000 2500000 2000000 1500000 1000000 500000 0 GT-MHR (A 336 F22) Reactor pressure vessels GT-MHR (A 336 F22V) ANTARES PBMR 2014 1879405 2016779 1687247 1165717 2020 1797675 1976828 1531381 1191385 2030 2276805 2407639 2241622 1312815 General view of RPV
CONCLUSION Materials science and forecast approach to choosing a reactor vessel based on neural networks forecasting of the cost of vessel steel components has been proposed and realized: The forecast of the cost of the most promising steels for application in HTGR: A 336 Grade F91, Grade F22, Grade F22V and A 508 Grade 3 Class 1 for 2020 2030 with an error of 19% was executed. The error is less by 12 % as compared with the forecasting based not on the neural networks; The prime cost of vessel steel in 2020 will fall on average by 3 %. But in 2030 it will be 20 % higher compared with the present.
CONCLUSION The forecast of the cost of reactor pressure vessel for the HTGR conceptual projects: GT-MHR, ANTARES and PBMR was executed. The PBMR vessel manufactured from steel A 508 has the lowest prime cost by the results of comparative analysis. Materials science and forecast approach based on neural networks can be applied to forecasting of the cost of other HTGR constructional materials and components and other nuclear power systems of Generation IV.
Thank you for your attention!