Philipp Wolfrum

Alternating Direction Method of Multipliers for decentralized electric vehicle charging control

The integration of Electric Vehicles (EVs) into the power grid is a challenging task. From the control perspective, one of the main challenges is the definition of a comprehensive control structure that is scalable to large EV numbers. This paper makes two key contributions: (i) It defines the EV ADMM framework for decentralized EV charging control. (ii) It evaluates EV ADMM using actual data and various EV fleet control problems. EV ADMM is a decentralized optimization algorithm based on the Alternating Direction Method of Multipliers (ADMM). It separates the centralized optimal fleet charging problem into individual optimization problems for the EVs plus one aggregator problem that optimizes fleet goals. Since the individual problems are coupled, they are solved consistently by passing incentive signals between them. The framework can be parameterized to trade-off the importance of fleet goals versus individual EV goals, such that aspects like battery lifetime can be considered. We show how EV ADMM can be applied to control an EV fleet to achieve goals such as demand valley filling and minimal-cost charging. Due to its flexibility and scalability, EV ADMM offers a practicable solution for optimal EV fleet control.

Optimal estimation and control of clock synchronization following the Precision Time Protocol

Using the Precision Time Protocol (PTP) specified by the IEEE 1588 standard, synchronization of distributed clocks is achieved by propagating the timing information of a preselected master clock throughout the entire network. Based on this directly or indirectly obtained noisy timing information, each slave clock tries to follow as closely as possible the master time. This paper formulates the PTP based clock synchronization as an estimation-control problem. An LQG controller is designed which produces an optimal reconstruction of the master time at each slave in the sense of minimizing the mean square error of the estimator and minimizing an LQR cost function for the controller. The performance of the proposed optimal controller is verified by simulation results.

Sensor and Actuator Placement for Linear Systems Based on

Sensor and actuator placement algorithms are developed for linear discrete-time systems based on H₂ and H_∞ optimization. For sensor placement, we design an observer that minimizes the H₂ norm of the error dynamics and the number of sensors at the same time. For actuator placement, we design a state feedback controller that minimizes the H_∞ norm of the closed-loop system and the number of actuators at the same time. Any other combination of actuator placement for state-feedback design or sensor placement for observer design for continuous- time or discrete-time linear systems based on H₂ or H_∞ optimization can be derived from these results. In both presented cases, the number of sensors or actuators is formulated as the ℓ₀ norm of the observer or controller gain matrix. This ℓ₀ norm is then relaxed to a weighted ℓ₁ norm in order to obtain an iterative convex optimization problem. As an application example, we use the sensor placement algorithm to place phase measurement units with maximal impact on the H₂ performance in a power grid.

H_{2}

In industrial automation networks based on the Precision Time Protocol (PTP) large temperature changes as well as mechanical shocks and vibrations may severely affect the performance of the local oscillators clocking the network nodes, thus making accurate time synchronization challenging. This problem is particularly critical in large industrial networks with long linear paths, as multiple uncertainty sources tend to accumulate while PTP event messages are forwarded towards the slave clocks. In this paper, the performance of a clock state estimator based on a special Kalman filter as well as on a detailed model of the PTP communication mechanism is described. The reported simulation results when the network nodes are subject to changeable environmental conditions provide interesting guidelines to keep synchronization accuracy in industrial networks within given boundaries.

and

Algorithms following the peer-to-peer Precision Time Protocol (PTP) specified by the IEEE 1588 standard achieve synchronization of distributed clocks by propagating the timing information of a preselected master clock throughout the entire network. Based on this noisy timing information, each slave clock tries to follow as closely as possible the master time. In this work we formulate clock synchronization as a stochastic estimation-control problem. A two dimensional LQG controller is derived which produces an optimal reconstruction of the master time at each slave in the sense of minimizing the mean square error of the estimated master counter and frequency. Owing to its specific structure, the LQG controller does not violate the transparent clock concept. The performance of the proposed controller is verified by simulations.

H_{\infty}

Abstract This paper describes a model based optimizer that allows a CHP unit to supply backup power to a Smart Grid on the one hand and minimize the cost for heat and power supply on the other hand. The model of the CHP unit is lean but nevertheless accurately represents the thermal behavior of the storage as well as the aging effect of engine starts. Thanks to the small model dimension we can solve the optimal dispatch problem efficiently using dynamic programming. Two selected soft- and hardware in the loop tests are discussed to demonstrate the performance of the approach.

Optimization

Abstract The increasing integration of renewable power generation in power systems poses new challenges to power system control. In this article, we describe a multi-layer power system control architecture that addresses these challenges and comprises economic planning, supervisory power control, as well as voltage and frequency control as main layers. Several parts of this architecture have already been developed and tested on diverse components including building energy management, micro grid management and wind park control. We provide an overview of these applications and indicate white spots in this control architecture that require further research.

A clock state estimator for PTP time synchronization in harsh environmental conditions

Robust optimal dispatch, secondary, and primary reserve for power systems is considered in this paper based on a novel robust formulation of the well-known power flow optimization. The uncertainty of power generation and load at the power system buses is modeled as nominal, expected power supply and demand, uncertain load and generation variations, as well as tunable dispatch, secondary, and primary reserves. The power transmission between the buses is modeled with algebraic, linearized power flow equations. The challenge is how to distribute dispatch as well as secondary and primary reserve at minimal cost, such that the power flow satisfies certain constraints even after unknown but bounded load and generation variations. These constraints reflect among others the maximal steady-state frequency deviation and the loading limits of the individual power lines. The resulting optimization problem is reformulated to different linear programming problems that can be solved efficiently even for very large systems. The applicability is shown for different IEEE test bus systems.