The main focus of research in the field of high-energy heavy-ion physics is the study of the quark-gluon plasma (QGP). Topic of the present work is the measurement of electron-positron pairs (dielectrons), which grant direct access to some of the key properties of this state of matter, since after their formation they leave the hot and dense medium without significant interaction. In particular, the measurement of the initial QGP temperature is considered a "holy grail" of heavy-ion physics. Therefore, in addition to the analysis of existing data, a feasibility study has been conducted to determine to which extent this goal would be achievable by upgrading the ALICE experiment at CERN. Dielectrons are produced during all stages of a heavy-ion collision, with their invariant mass reflecting the amount of energy available at the time of their formation. Dielectrons of highest mass are thus produced in the initial scatterings of the colliding nuclei by quark-antiquark annihilation. Correlated electron-positron pairs can also emerge from the decay chains of early-produced pairs of heavy-flavour (HF) particles. During the QGP stage and at the beginning of the hadronic phase, the system emits thermal radiation in the form of photons and dielectrons, which carry information about the medium temperature to the observer. In the final stage of the collision, decays of light-flavour (LF) hadrons produce additional contributions to the dielectron spectrum. The present work is based on early data from the ALICE experiment recorded from lead-lead collisions at a center-of-mass energy of 2.76 TeV. Due to the limited amount of data, a focus is placed on achieving high efficiencies throughout the analysis. To this end, a special electron identification strategy is developed and a custom track selection applied, together resulting in a tenfold increase in pair efficiency. The dielectron spectrum is evaluated on a statistical basis, using a pair prefilter, which is optimized based on two signal quality criteria, to reduce the fraction of electrons and positrons from unwanted sources at minimum signal loss. In addition, an artifact of the track reconstruction is exploited to suppress pairs from photon conversions and to correct the dielectron yield for a contribution from different-conversion pairs. The main signal uncertainty is extracted from the deviation between results of 20 analysis settings and amounts to 20% in most of the studied kinematic range. For comparison with the analysis results, a hadronic cocktail consisting of the LF and HF contributions is simulated, which can reasonably well describe the measured dielectron production, with a hint of an enhancement at low invariant mass. Two approaches to model the in-medium modification of the heavy-flavour are followed, resulting in up to 50% suppression, which creates some additional space for a thermal contribution at intermediate mass. For a complete comparison between experimental data and theoretical expectation, two model calculations are consulted. The Thermal Fireball Model provides predictions for thermal dielectron radiation from the QGP and hadron gas. The data tends to be better described with these additional thermal contributions. For a comparison with a prediction by the UrQMD model, the HF component of the cocktail is subtracted from the data. This results in better agreement if the HF suppression by in-medium effects is taken into account. The feasibility study in this work has served as a physical motivation for the ALICE upgrade for LHC Run 3. The precision with which the early temperature of the QGP can be determined via dielectrons is chosen as key observable. A multitude of individual contributions are merged into a fully modeled dielectron analysis. The resulting signal-to-background ratio represents some of the expected systematic uncertainties, while from the significance combined with the planned number of lead-lead collisions a realistic "measurement" with statistical fluctuations around the expected dielectron signal is generated using a Poisson sampling technique. Since the HF yield exceeds the QGP thermal radiation by about an order of magnitude, an additional analysis step exploiting the enhanced track reconstruction is introduced to reduce its contribution by up to a factor of five. The resulting reduction in pair efficiency is overcompensated by an up to hundred times higher collision rate. The entire cocktail is then subtracted from the sampled data to isolate the thermal excess yield. The final analysis of this spectrum shows that the inverse slope of the model prediction, which depends directly on the QGP temperature, can be reproduced within statistical and systematic uncertainties of about 10%. The promising results of this study have contributed on the one hand to the realization of the ALICE upgrade and to a design decision for the new Inner Tracking System, and at the same time represent exciting predictions for upcoming measurements.