Bioprocesses require efficient monitoring and control of the respective process parameters. In order to achieve defined states to produce a product in high quantities and required quality, the respective time-dependent biomass concentration as a catalytic converter is a key variable. However, several offline measurements such as the optical density or dry weight measurements exist but repeatedly do not satisfy the requirements for modern bioprocess development. Therefore, hard sensor types based on permittivity measurements and dielectric spectroscopy were used to determine the biomass online. However, these measurements come with different challenges and disadvantages. An alternative to these hard-type sensors are soft-type sensors, which indirectly estimate the biomass produced by existing measurements with a bioreaction system in real-time. Therefore, the known inlet, outlet and feed streams were quantified to calculate rates. These rates are used to develop a balance system according the law of mass conservation for carbon, nitrogen and degree of reduction. This overdetermined model system of equations can be solved with calculated rates or reconciled rates for the best possible estimation of the biomass. The estimated biomass amount is used for different specific substrate uptake rates during the experiment. For a trustworthy estimate of the biomass, a statistical test has been introduced which can distinguish whether a systematic error exists or whether the deviations of the rates can be explained with random errors. In this context, the error propagation of the measurements was considered in real-time as a novel approach. The investigation of the soft sensor concept shows that in the case of a valid model, the degree of reduction balance with the calculated rates of the measured quantities and the C-balance with the reconciled rates best describe the bioreaction. There were achieved accuracies below 10% for the estimation for the biomass relative to reference measurements such as optical density and dry weight measurements demonstrated with a Saccharomyces cerevisiae cultivation. A workflow was developed which should guarantee a successful fermentation using this robust soft sensor concept.