Please use this identifier to cite or link to this item:
|Title:||Data Analytics of Smart Grid Environment for Efficient Management of Demand Response|
|Keywords:||Smart Grid;Data Analytics;Demand Response;smart home;electric vehicles|
|Publisher:||Thapar Institute of Engineering and Technology Patiala|
|Abstract:||The future of the power industry heavily relies on the use of modern electric grids integrated with information and communication technology (ICT). Such grids are commonly known as smart grids. The advantage of using smart grids is that they provide a better quality of service in terms of better resource and asset management, detecting faults in the system, efficient energy consumption by reducing the demand and supply gap, and peak load reduction. Data analytics has already been applied extensively in the power sector to provide various services such as-demand forecast- ing, revenue protection, and data visualization. However, there are still many areas which can be benefited by using data analytical techniques. One such area is the demand response management in the smart grid environment where data analyt- ics can be effective in order to manage the overall load on the grid. The entities involved in the smart grid comprise of power generation units, transmission and distribution units, and end-users. The end-users may belong to the different sectors such as–commercial, residential and transportation. The consumption data related to these users can be analyzed to provide many ancillary services in the smart grid and to improve the overall quality of service for the users. Keeping this in mind, the major focus of this thesis is on data analytics in the smart grid environment along with the demand response management of the connected loads. To achieve these tasks, four different schemes have been proposed in this thesis with an em- phasis on data analytics and demand response management in smart grid. In the first technique, a top-down approach is designed to detect electricity theft in the power network based on decision tree (DT) and support vector machine (SVM) Unlike the existing schemes, the proposed scheme detects and locates the real-time electricity theft at every level in power transmission and distribution (T&D). In the T&D level, the data received from various sensors is analyzed to detect the theft in the power lines by compensating for the T&D losses. At the consumer level, the DT is used to compute the value of expected load consumption in the smart homes which along with other attributes is given as an input to train the SVM classifier. Based on the training, the SVM detects the electricity theft at the consumer level using the input parameters received from various smart homes. The results obtained using the proposed scheme indicate that it detects the theft with high accuracy (i.e., 92.5%) and low false positives (i.e., 5.12%). The second scheme proposed in this thesis is designed for reducing the demand and supply gap in the grid by managing the demand response of smart homes and plug-in hybrid electric vehicles (PHEVs). For this purpose, the SVM classifiers are used to identify the users (smart homes or PHEVs), whose load profile needs to be regulated. The proposed scheme is a hierarchical scheme which manages the load profile of the grid in two phases. In the first phase, the residential loads (comprising of smart homes) are identified and managed according to the grid requirements. In the next phase, the charging rates of PHEVs are regulated when the residential loads are not sufficient to flatten the load profile of the grid. The results obtained using the case study of PJM and Open Energy Information dataset prove that the proposed scheme is effective in balancing the overall load profile of the smart grid. In the third scheme, peak load on the grid is reduced by analyzing the energy consumption patterns in the smart homes. Var- ious factors are computed from the data gathered from the smart homes; based on which, algorithms for taking the demand response decisions to manage the consumer load profiles are proposed in the peak load scenario. The proposed scheme keeps the value of load curtailment in a smart home below the user specified curtailment value. However, in the cases where load curtailment is more than the user-specified value, incentives are provided to the consumers to compensate for the violation of consumer comfort. The efficacy of the proposed scheme has been validated using two case studies of realistic and harsh scenarios of electricity usage in smart homes. The results depict that the proposed scheme efficiently brings down the peak load demand by a factor of more than 27% when the proposed algorithms are used in tandem. Moreover, the proposed scheme also increases the savings of the consumers by reducing their overall electricity bills. The fourth scheme uses the tensor-based operations to reduce the data dimensionality of the data gathered from the power network. The proposed approach includes the extraction of core data from the gathered data by using tensor operations such as-matricization, vectorization and tensorization with the help of higher-order singular value decomposition. The core data is then used for various purposes such as managing the demand response of the loads in a smart city. The obtained results give a clear indication that the tensor- based approach in reducing the data dimensionality is effective and can be used for processing the gathered data in order to manage the demand response in the smart grid environment.|
|Appears in Collections:||Doctoral Theses@CSED|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.