In this work, trend analysis is proposed as a preliminary step for data reconciliation in linear dynamic systems. The trends of measured data along a specified time horizon are identified before they are made consistent with those of the dynamic process model. This preprocessing of data leads to a reduction in the variance estimation of dynamic data reconciliation. For the trend analysis step, denoising is applied using wavelets. The performance of the proposed strategy is compared with that of the Kalman filter. For this purpose, data from simulations of a linear dynamic case are reconciled. Further extensions that contemplate nonlinear cases are also introduced, showing promising results in terms of accuracy and computing efficiency.