Pulp quality and energy economics are the key issues in any thermomechanical pulping (TMP) plant. The control in TMP plants is affected by two different types of disturbances: the slow wearing of refiner plates and the faster variations in raw material characteristics. Both these have a great effect on pulp quality and the process economy if not properly controlled. Nowadays pulp properties such as freeness, fibre length distribution, shives, coarseness and even average cell-wall thickness can be measured directly from a pulp sample, but even today there are no methods available to measure production rate or chip moisture accurately enough. The feed-back control systems that are based on pulp property measurements after latency chest can only partially solve the control problem for two reasons: it is not possible to compensate raw material variations that are faster than twenty minutes and due to the multivariable non-stationary connection between raw material variations and pulp properties it is very difficult to select the right control action. The objective of this thesis was to solve TMP control problems by developing calibration models and methods for the near infra-red (NIR) measurement unit to measure consistency and properties of pulp and a control system to compensate raw material variations. The term consistency is used for traditional reasons and is related to the moisture of pulp. We have developed a totally new control system to solve raw material variation problems. It is based on consistency measurements and an adaptive control system that manipulates the dilution water flow and the speed of the chip feeder. Variations in specific energy and refining intensity where reduced significantly and are the two most important state variables of the refiner with which to predict pulp quality. The benefits obtained is improved pulp quality, energy savings, increased line availability and production. Freeness, fibrelength, tear and tensile index, variations have decreased up to 80%. We have shown in this thesis by statistical methods that in addition to the consistency measurement, it is possible to measure with the NIR-technique other pulp properties such as freeness, size distributions, and tensile index. Due to the analyser being mounted directly in the blow line just after the refiner, there is practically no measurement delay. Pulp quality is mainly defined by the operation of the first stage refiner. The measurement of pulp properties after the first refining stage gives a possibility to develop a new type of fast quality control system that was not possible before. The main reason for the difficulties to find a good calibration model for the NIR-analyser is the lack of a theoretical model for light scattering by collections of particles, which have varying shapes and surface appearances. We have begun our theoretical model building by developing a light scattering model for wood fibre, which is the most important particle in pulp. We have modelled the irregular overall shape and the cell-wall thickness of wood fibre using state-of-the-art statistical methods. Based on the theoretical calculations we are claiming that when we are measuring pulp quality by the NIR technique, we are mostly defining the shape parameters of the average pulp particle, and not so much by their chemical composition. The agreement with theoretical and experimental results are very good and allows us to use the developed method in solving practical measurement problems. The on going development work will ultimately lead to a significant reduction of calibration samples.