Due to the recent popularity of smart home and internet of things technologies, an increasing number of smart energy meters and sensors are being deployed, which leads to ever more data being generated. This data has the potential to deliver valuable insight and information for users and various stakeholders from business over science to politics. However, in order to fully leverage its potential, it is crucial that the generated sensor data is integrated and processed in a way which can deliver precise answers to precise questions. In particular, energy analysis based on real smart meter data usually considers a large amount of factors and can thus have a high complexity. Hence, the main challenge is finding a proper way for integration, modelling and analysis of sensor data. A possible solution might be the use of Online Analytical Processing Cubes, usually known in the context of business intelligence. Originally developed for business analytics, those analytical information systems extract data from operative and external databases and integrate them in a separate Data Warehouse for analysis. By splitting the data into and descriptive dimensions and facts to be analyzed, the data is modelled in a multidimensional way. Therefore, the data is stored in multidimensional database architectures. In this work, it was examined how well Online Analytical Processing Cubes can also be used for the integration and analysis of sensor data, in particular smart meter data. To do so, the first half of this thesis examined the basics of Smart Meters and Online Analytical Processing. Afterwards, a concept was developed to integrate energy consumption data produced by smart meters using multidimensional database architectures and data cubes and to analyze it regarding various dimensions. The main focus here were weather data and building typology and information as describing elements of the energy consumption. To do so, three hypotheses were developed and investigated using multidimensional data modelling. SQLite, R and Python were used for implementation.