Distributed and Online Advanced Metering Infrastructures Data Validation using Single-Board Devices
Ladda ner
Typ
Examensarbete för masterexamen
Master Thesis
Master Thesis
Program
Publicerad
2015
Författare
Sandström, Jonas
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
Traditional electrical grids are evolving to information-carrying cyber-physical electrical grids, also referred to as Smart Grids. One of the enablers of this shift is Advanced Metering Infrastructures (AMIs), networks of heterogeneous meter devices that provide information and remote control to energy suppliers. AMIs produce a flow of information carrying the health and power consumption of the infrastructure. This information flow is usually provided at predefined time intervals and concerns large amounts of data. For instance, if one hundred thousand meters are deployed in a city and energy consumption readings for each household are reported every hour, 2.4 million consumption readings per day need to be processed. Electricity suppliers can use this information stream in novel applications, such as real-time pricing and demand-based production. Unfortunately, the correctness of the energy consumption data stream cannot be taken for granted since there are many potential error sources such as faulty devices, wrongly calibrated devices, lossy communication protocols, or fraudulent users, among others. Hence, there is a need for validation before significant decisions are made based on this data. Of importance is that the validation is performed in a real-time fashion with low latency, to deliver up-to-date information. Needed validation may change with the specific AMI or with different error types. Thus, the validation need to have the possibility to consist of a set of rules and be reprogrammable e.g., by adding, removing or modifying existing validation rules. In order to be fast and scalable, with the increasing number of households and finer time granularity, the solution requires the validation to be distributed and parallel. Those specifications can be met by using data streaming. Notice, to accomplish a deployment of such a distributed system in an AMI, Single-Board Computers (SBCs) could be used with low cost and energy use. This thesis builds a prototype of such a system. It uses data streaming to validate the consumption data. Data streaming is necessary for online analysis. Stream Processing Engines (SPEs) consume the data stream immediately upon arrival by utilizing continuous queries. These continuous queries can be implemented to formulate validation rules, cleansing the consumption data. The implementations can be modelled to handle specific errors, which gives the system customizability. SPEs can process large amounts of data with low processing latency and in distributed and parallel fashion, and thus achieve high throughput. To make the system distributed and AMI deployable, a cluster of SBCs running a SPE will be used. Thus, also keeping the cost and energy usage low. This thesis show that this is possible with an almost linear increase in processing capacity with each added SBC, i.e. in a nearly perfect scalable way.
Beskrivning
Ämne/nyckelord
Data- och informationsvetenskap , Computer and Information Science