| dc.description.abstract | The evolution of the traditional power grid (TPG) to the digital smart grid is expected to improve the operation of the power system. The smart grids' capability to support various distributed generation technologies, self-healing capacity, and ability to minimize operational costs are some of its benefits. However, distribution networks are facing congestion challenges due to the increased number of flexible loads and Distributed Energy Resources (DER) being used. Moreover, the dependence of such technologies on uncontrollable factors, such as temperature, wind speed, solar radiation, etc., may result in potential congestion problems which were not of concern in the past within the distribution network. This is due to the high power consumption of active loads and the weakening correlation between electricity prices and demand resulting from the increased penetration level of intermittent Renewable Energy Sources (RES). Such congestion results in voltage violations and/or thermal overloading as a result of the power flow exceeding a network asset's transfer capability, possibly damaging devices such as distribution transformers and feeders. Rather than incurring a huge cost to reinforce the network assets, the Distribution Network Operator (DNO) can use Demand Side Management (DSM) to motivate consumers to shift their load from peak to off-peak times. However, congestion management methods based on DSM rely on communication between DNO, aggregators, and consumer meters to encourage customers to make some corrective actions, such as peak shifting, peak clipping, and valley filling, to relieve congestion. Cyber attacks against aggregators can compromise the operation of DSM-based congestion management methods, and hence, affect the security and reliability of electrical networks. The congestion management schemes become useless in the event of a cyber attack and can instead result in an increase in network congestion. The vulnerability of DSM-based congestion management methods to Load Altering Attacks (LAA) is hence studied to determine the cyber weaknesses in the DSM protocols. An optimization algorithm is developed to determine which aggregators a cyber attacker would compromise in order to cause congestion by minimally altering their load profiles. The impact of such attacks on congestion and consumers' electricity bill is then studied. A mitigation scheme is formulated to determine the most critical aggregators in the network. The security of these aggregators is then reinforced to mitigate such cyber attacks. Also, a detection technique is proposed using Long Short-Term Memory (LSTM) Recurrent Neural Networks (RNN). By securing the aggregators, the DSM congestion management schemes can be effective in alleviating congestion in the distribution network.
Given a cyber secure network, to minimize consumers' electricity cost and decrease the Peak to Average Ratio (PAR) of the load curve to a required level that alleviates existing congestion, a multi-objective optimization is proposed to schedule flexible loads. This results in a consumer load schedule that is economical and does not require the imposition of congestion tariffs. However, the success of the proposed congestion management scheme relies on the accuracy of the consumer load consumption. Hence, uncertainty analysis of consumers' flexible load schedule is executed to ensure the desired robustness of the power flowing in the distribution network to changes in uncertain variables. The results obtained are compared with the existing congestion management scheme demonstrating the advantage of the proposed multi-objective framework in terms of decreasing price and flattening the load curve while alleviating congestion. However, it should be noted that modeling distributed energy resources is not an easy task as they rely on uncertain factors which are hard to predict. It is very challenging to design a congestion management scheme given the uncertainty of such flexible loads consumption and electricity prices. Obtaining stochastic models for such loads may not be easily available in practice. To overcome this challenge, a Deep Deterministic Policy Gradient (DDPG) Reinforcement Learning (RL) scheme is proposed to alleviate congestion. DDPG RL is a model-free technique that does not require explicit probabilistic models of the controllable loads to determine the change in electricity prices needed, in the form of tariffs and/or subsidies. The DDPG RL technique is compared with the existing model-based congestion management scheme and the results obtained demonstrate the outperformance of the proposed technique in terms of electricity cost and PAR of the load profiles.
Aggregators load schedules utilizing the day-ahead congestion management schemes are used in the real-time market. However, this does not eliminate the possibility of having congestion in the real-time market. The electrification of the transportation system is double-edged for the smart grid. Although it is green and eco-friendly, uncontrolled charging of electric vehicles (EV) could cause not only distribution network congestion, but also long queues at the charging stations. Hence, it is necessary to ensure existing charging resources are efficiently utilized. An algorithm is hence developed that assigns EVs to charge stations such that distribution network congestion and time spent by the user from requesting a charging service to accessing it is minimized. The Lyapunov function is utilized for developing an EV assignment algorithm to manage a dynamic population of EVs ensuring queuing stability. Moreover, as the EV assignment algorithm relies on the communication network, an intrusion cyber-attack can occur resulting in an unstable queuing system. An intrusion detection technique is, hence, proposed which utilizes existing transportation network sensor data with the EV charging stations operator information to detect such attacks.
IEEE 33 bus system is used as a case study to demonstrate the effectiveness of all the proposed techniques compared to the existing models. | en |
