In this study, a measurement and verification (M&V) cost minimisation model is proposed to
deal with both the M&V modelling and sampling uncertainties. In order to find the optimal
solutions in terms of the modelling accuracy level, and the sample size, the M&V cost that
includes the modelling cost, sampling cost, and overhead cost is formulated as the objective
function, in which the modelling cost is developed as a function of the modelling accuracy
in terms of the coefficient of variation of the room mean square error (CV(RMSE)), and the
sampling cost, which is directly related to the sample size.
In order to illustrate the effectiveness of the proposed model, an optimal M&V modelling
and sampling strategy is designed for a traffic intersection lamp retrofit project. In addition,
partial optimal M&V plans designed with optimal sampling but non-optimal modelling solutions,
or with optimal modelling but non-optimal sampling solutions are employed as the
benchmark. Comparisons between the optimal and non-optimal solutions show advantageous
cost savings performance in the execution of sampling and modelling activities for the case
study. More precisely, the optimal solutions reduce the sampling cost by 55% and the total
M&V cost by 14% against the solutions obtained by optimal modelling but non-optimal
To test the applicability and flexibility of the proposed model for the cost-effective design of
similar traffic light retrofit projects, simulations have been carried out to evaluate the model
performance when applying the model to M&V projects with different characteristics. The
simulation results show that the proposed model is able to offer flexible trade-offs between
the modelling and sampling uncertainties; namely, using more accurate baseline models and
smaller sample sizes or less accurate baseline models but greater sample sizes to accommodate
different practical needs in executing M&V projects with different characteristics.
The major contributions of this study can be highlighted as follows: 1) a M&V modelling
cost model is developed, which is able to offer a quantitative analysis of the M&V baseline
model uncertainty and cost; and, 2) a M&V cost minimisation model is proposed to handle
both the M&V modelling and sampling uncertainties cost-effectively. The effectiveness and
flexibility of this model are demonstrated by a case study and simulations.
Dissertation (MEng)--University of Pretoria, 2016.