RESIDENTIAL AREA ENERGY MANAGEMENT USING DEMAND REGULATION AND STORAGE DEVICES

Abstract

Since last few decades power sectors at every country in this globe is going through a paradigm shift to integrate distributed green energy resources into the power system for reducing the effect of global warming. As the Sun shines at every part of the world during a certain period of the day, most of the countries agreed in 2015 at Paris agreement to install a large amount of renewable energy resources, especially solar Photovoltaic (PV) panels. In the span of last few years, different innovative solar array technologies (like Monocrystalline, Polycrystalline, Thin-Film, etc.) have been procured to increase the efficiency of the panels at low cost. This makes PV panels economical for deployment at local level for fulfilment of the energy demand. These local distributed energy resources cater the electrical energy demand of the area, and at the same time send back the surplus energy to upper distribution and/or transmission grids. Beside the advantage of network loss reduction and environment friendly clean energy supply, utilization of solar power may cause high power imbalance issues due to its intermittent nature. To mitigate variability and vulnerability in the solar power generation, storage technologies, especially batteries in case of distribution network, are collocated with them. Controlling the charging/discharging behaviours of the batteries, maximum utilization of solar power is done. According to the studies of various international agencies like International Energy Agency (IEA), World Energy Council (WEC) etc., energy requirement in all over the world is increasing exponentially with the population growth. Due to urbanization and recent development in residential appliances, domestic consumers have significant share (almost 27%) in the total energy demand of the entire world. Therefore, rooftop solar power plants are growing interest to meet the residential energy requirement with low energy cost. Now, solar power generation is available during day time but residential load profiles experience peak at evening hours. This necessitates development of new innovative strategies for optimal residential energy management by leveraging the benefits of demand regulation (such as load shedding or shifting) and deployment of battery energy storages. However, non-identical operational characteristics of the appliances and dissimilar life style of the home residents make the above process strenuous. Further, varying architecture of the residential areas (like home, network etc.) cause requirement of different optimization frameworks for each type. Day ahead deterministic strategies are well known and established conventionally considering the forecasted data. Accurate forecasting of the next day events is the main key behind the success of such strategies. However, due to afore mentioned uncertainties, accurate prediction of future events are next to impossible and under the real time uncertain conditions, deterministic approaches fail to provide optimal control decisions. This motivates the researchers ii to design novel energy management strategies to account the real time uncertainties properly. Again, success of energy management strategies pivot upon the correct identification of the residential appliances as power consumption flexibility provided by critical appliances are lesser than that of non-critical appliances. Therefore, in order to provide effective solutions to the residential energy management problem for optimal operation of the loads and different energy resources under various uncertain conditions, the research works presented in this thesis are carried out. The main objectives are to propose a simple deep learning architecture for accurate identification and classification of different residential appliances, and to design novel day ahead and real time energy management strategies for smart homes, residential apartment building and large distribution networks considering load shedding and/or load shifting operations. Demonstrating on real life and as well as standard data the efficacy and superiority of the proposed methodologies are established over the existing state of the art techniques. Initially, a deep learning based residential load identification and classification strategy is proposed as a 1-D Convolutional Neural Network (CNN) with sinusoidal kernel initializers. For efficient operation, 15 primary and 225 Multiscale Sinusoidal Kernels (MSK) are formed and included in the CNN architecture for extracting important discriminative features from the current signals. Classification accuracy of the proposed CNN module, named as MSK-CNN, is further improved by adding a new non-linear Activation Function (AF), SL-ReLU, having an adjustable logarithmic function along with softsign and ReLU (Rectified Linear Unit) functions. The efficacy of MSK-CNN is tested on a practical dataset, created by the authors, consists current signals of different real life residential appliances. It is showed that MSK-CNN can successfully distinguish 18 single household appliances and 12 appliance combinations, and can attain an accuracy of 98.61%. The experimental outcomes reveal superiority of the MSK-CNN over classical CNN modules with other kernel initializers and AFs in terms of classification accuracy and training convergence. Afterwards, a robust Conditional Value at Risk (CVaR) optimization approach for day ahead Home Energy Management Systems (HEMS) is put forward to reduce the effect of risk of real-time exposure to energy price and solar power generation uncertainties. The CVaR method is integrated with the Two-Point Estimation (2PE) analysis to approximate the solar power, modelled as beta probability distribution function, in low computation effort. Later the optimization constraints are revised to their robust counterparts by accounting a certain amount of uncertainty in the energy prices from their nominal values. The optimization problem is developed to minimize the risk value of the energy cost. Again to maximize the life of the Plug- iii in Electric Vehicle (PEV) a pseudo cost function for the PEV battery degradation is proposed. The entire optimization portfolio is developed as a Mixed Integer Linear Programming (MILP) for its easy execution. Simulation is demonstrated on a smart home, designed as an AC/DC microgrid, having practical appliance data sets, to prove the efficacy of the proposed method. To overcome the challenges associated with the real time uncertainties on smart home energy management, a multi-objective optimization portfolio is propounded for a smart home equipped with battery associated rooftop solar panels, lighting loads, air conditioners, and other smart appliances. The energy management problem is framed for simultaneous minimization of the monetary energy cost and total dissatisfaction due to regulation in power consumption. The entire optimization portfolio is designed as a time average stochastic problem, which is simplified by the combination of queueing theory and Lyapunov optimization. The revised problem takes form of a mixed integer convex nonlinear programming, which is solved using Outer Approximation (OA) approach. The proposed real-time HEMS framework needs only the current data regarding the random input parameters like renewable generation, energy price, and aggregated load demand, and does not call for their probabilistic estimation. Case study is carried out on a practical home data to proof efficacy of the proposed strategy and also the simulation outcomes are compared with one of the popular real-time energy management process named online greedy algorithm. Thereafter, a real-time energy management strategy is designed for a Smart Residential Apartment Building (SRAB) having non-identical occupants at the Dwelling Units (DUs). The aim of the present research work is to design a distributed energy management algorithm, which can optimize the real time demand of the entire building against abruptly updated rooftop solar generation and Real-Time Price (RTP) of energy. The proposed energy management strategy differentiates among the DUs by considering a new parameter named load criticality level, which is defined as the value imposed by the DU residents to their power consumption. The optimization portfolio is developed as a novel bi-level, stochastic, multi-objective optimization problem where the maximization of utility of the consumed power is considered simultaneously with the cost minimization. To this end, a virtual energy trading platform is designed in this article between central building management system and the DUs, where they interact with each other by following the directives of single-leader multi-followers Stackelberg Game (SG). The solution strategy is proposed as a Lyapunov optimization to eliminate the complexities regarding time average stochastic equations. Strenuous simulation on real-time data of four DUs, it is proved that the proposed framework can track the abrupt change in RTP and solar generation iv efficiently. Comparing with two benchmark methods viz. centralized process and greedy algorithm, the superiority of the designed energy management portfolio is established. Then a new real-time optimization framework for advanced energy efficient management of active radial distribution networks is suggested. The proposed energy management process leverages the benefits of simultaneous deployment of online Direct Load Control (DLC) and Conservation Voltage Reduction (CVR) for decreasing peak energy demand. Initially the proposed problem is designed as a time coupled stochastic Mixed Integer Non-Convex Programming (MINCP) to accommodate long-term offline beneficial aspects in real time optimization framework, which is later simplified using merger of Queueing theory and Lyapunov optimization. A successive MILP (s-MILP) solution approach is proposed for accurate and fast convergence of the revised MINCP framework. The efficacy of the developed strategy is evaluated after comparing with two-benchmark energy management models (viz. offline energy management process and online Greedy Algorithm) by demonstrating on modified IEEE 69-bus balanced and IEEE 123-bus unbalanced distribution network test systems. Simulation results show that the designed strategy can generate near global optimal solutions within very short time with the online data of solar generation, energy price and load demand, and strictly obey the constraints of DLC and CVR. Fairness in implementation of DLC proves the effectiveness of the proposed strategy. It is found that the designed s-MILP solution process for MINCP outperforms the standard second order conic relaxation and piecewise linearization based solution methods in terms of computation time and accuracy respectively. Further scalability of the proposed approach is validated by testing on large distribution networks. In view with the recent advancement in the DC loads and DC energy resources, AC/DC Hybrid Distribution Networks (HDNs) are getting attention as the futuristic architecture of the DNs. This study proposes a risk constrained energy efficient management algorithm by merging Load Shifting (LS) and CVR techniques. The optimization framework aims to simultaneously minimize both true and conditional risk or CVaR values of the expected energy cost under uncertain solar power generation, load demand and upper grid energy price. In contrast with the available stochastic optimization process, in this article two point estimation strategy is employed in place of Monte Carlo Simulation (MCS) for scenario generation from the probability density functions of the uncertain parameters to reduce computational exertion. The proposed centralized optimization framework is initially developed as MINCP problem but to avoid computation complexity, the non-linear components are replaced by their linear counterparts. Later the revised problem is solved using s-MILP approach to obtain the optimal decisions for deployment of LS and CVR through smart inverters and volt-VAR controlling devices. Efficacy of the proposed v technique is demonstrated on modified IEEE 33-bus AC/DC HDN and the most energy efficient operation is found by merging LS and CVR. Simulation outcomes prove fast and near optimal convergence of the s-MILP compared to conventional second order conic programming relaxed mixed integer convex programming and piecewise linearization based MILP. Further, to assess the impact of network size on the solution time and optimality, the proposed Advanced Distribution Network Management System (ADNMS) strategy is employed on 132 bus AC/DC HDN. The presented works in this thesis are likely to contribute significantly to the area of household load identification and residential area energy management in the presence of battery units. The different developed techniques will be particularly useful for performing demand regulation strategies in differently sized residential area for maximum utilization of the solar power and minimization of the energy cost. Some future research suggestions on observations and simulations in this research area are proposed at the end of the thesis for the benefit of potential researchers.