The paper analyses the accuracy of the determination of normal heights in the national spatial reference system using the PL-geoid-2011 quasi-geoid model. The heights were determined using the PL-EVRF2007-NH normal height system. The paper discusses the results of a measurement experiment consisting in measuring 8 points with the use of the Network Real Time Kinematic (NRTK) technique and the Virtual Reference Station (VRS) surface correction generated on the basis of the TPI NETpro commercial network of reference stations and with the use of two global satellite systems (GPS, GLONASS). In the experiment, three measurement schemes were evaluated in relation to the number of measurement epochs recorded above the designated point, as well as the order of recording points. The conducted measurement experiment allowed to determine which of the proposed schemes guarantees the best accuracy from the point of view of establishing a height measurement network with the use of the NRTK technique and VRS corrections. The tests showed that it is possible to determine the height difference between points using the NRTK technique and the VRS method with an accuracy of 0.01m. However, to ensure adequate accuracy, elevation differences must be defined as the differences of the measured heights and should be determined several times and based on independent measurements at the beginning and end of the levelling section. Thus, determining elevation differences with the use of the NRTK VRS technique may be particularly effective in areas with large differences in height, where the determination of elevation differences with the use of classical methods is time-consuming. The obtained elevation differences require tying to the points of the height control network with the use of classical methods