AMI & SMART METERING Sensor network metrology for electrical grids The need for optimisation and traceability By Paul Clarkson, Senior Research Scientist, National Physical Laboratory, UK Electricity networks are evolving to incorporate more smart grid technologies and integrate renewable generation, leading to active distribution grids where energy is fed in at the LV and MV levels. Knowledge of the topology and power flow of such grids is limited, but is vital to enable further integration of renewable and smart grid technologies, effective smart grid control and to ensure continuity of supply. To provide the necessary information for grid observation and control, networks of sensors placed at multiple points in the grid are required. One option would be to place sensors at every conceivable branch of the distribution network. However, such instrumentation is costly to install and maintain and operators would be overwhelmed with data. The alternative to this mass- coverage approach is to mathematically specify the intelligent placement of sensors such that a correct balance can be found between instrumentation cost and providing information of sufficient quality for effective grid observation and control. Technologists at National Measurement Institutes (NMIs) throughout Europe have been collaborating on a research programme 1 to support the development of Smart Grids through metrology (measurement science) and part of this work is about understanding sensor networks for Smart Grids and optimising their use at the distribution level. This article looks at why such research is needed and what the challenges are of future research into sensor networks for Smart Grids. Understanding sensor networks It is often the case that very few measurements are available at Medium Voltage (MV) and Low Voltage (LV) levels in distribution grids. It is therefore necessary to estimate the voltages and power flows at many of the intersecting (or nodal) points and branches using “pseudo” measurements, derived from estimates of consumer demand based past usage. For example, a limited number of real measurements are then combined with state-estimation (a mathematical technique used to recover the voltages and power flows at nodal points in the network) to provide a more accurate up to date estimate of the nodal voltages such that action can be taken as necessary to ensure quality and security of supply. Traditional measurement science is focussed on individual measurements and their associated uncertainties, but there are additional challenges when multiple simultaneous measurements are required. The uncertainties in the derived quantities of interest to network operators (active and reactive power, voltage magnitude and phase) at various points on the grid are dependent on the accuracy, position and number of measurements placed on the grid as part of a sensor network as well as on the performance of the employed state- estimation method. It is essential that these dependencies are better understood so that sufficient confidence in the measurement system exists. New techniques are required to fully understand the behaviour of sensor networks used for grid monitoring. METERING INTERNATIONAL ISSUE - 4 | 2013 Active network management As distributed generation increases, greater control at the distribution level becomes more important. Such control relies on reliable measurement information. A recent paper by the Union of the Electricity Industry (EURELECTRIC) 2 states that the success of active distribution network management tools, required for the effective control of emerging smart distribution grids, “will depend on DSOs’ [Distribution System Operators] ability to actively monitor their grids, notably at medium and low voltage level. Today, DSOs have no systems installed for acquiring data from DG [Distributed Generation], in particular small scale DG. As the share of DER [Distributed Energy Resource] expands, DSOs will need monitoring, simulation, control strategies and advanced protection systems that allow them to supervise and control power flows and voltage in their MV and LV networks. This includes relevant monitoring functionalities from smart meters.” Such instrumentation would require high voltage transformers and will be expensive to install and maintain. It is vital to limit the installation of such equipment to reduce the cost of Smart Grid operation to a manageable level. The potential use of Smart Meter data to fulfil all or part of this task should not be overlooked (see below) and could provide a cost effective method of monitoring the power flow in future grids. Further, the final May 2011 report of the European Committee for Standardization (CEN)/European Committee for Electrotechnical Standardization (CENELEC)/ European Telecommunications Standards Institute (ETSI) Joint Working Group on Standards for Smart Grids 3 recognises the importance of “enhanced monitoring and observability of grids down to the low voltage levels, also with the use of smart metering infrastructure.” Optimisation of measurement information The economic and practical considerations referred to above lead to a need for a sensor network that is optimised to provide the necessary information to effectively control the Smart Grid at the distribution level, while minimising the cost of the required sensors. Supervisory Control and Data Acquisition (SCADA) systems are only suited to centrally managed grids. These systems treat the MV and LV distribution networks as “black boxes” and little is known about the power flow and often even the topology of these networks. Distribution systems are more complex and dynamic, having many more nodal points and branches than transmission networks. As such, new techniques are required to ensure observability of these grids and enable their effective control. These techniques must expand on traditional sensing strategies, which often rely on a degree of redundancy in the measurement data. Recently developed state-estimation and measurement placement methods have been shown to lend themselves well to this purpose. However, to date little has been done to fully account for the system state dynamics, which can change on a minute by minute basis and state- estimation algorithms assume exact knowledge of the network topology and line impedances, which are often poorly understood (see below). 23