Changing Facets of Power Systems Reliability and Resiliency

Power System reliability is a time-tested subject established over decades. With the push for renewable energy, this area appears to have slipped in its true application. This article explains the nuances.

Reliability, in everyday understanding, is “the expected quality of consistent performance”. It is a reasonable prediction on the ability of a product (or a system) to perform as intended under stated conditions. It consists of two components namely (a) Dependability (trustworthiness of stated repeatability) and (b) Security (consistent output of such repeatable performance). As a simple example, we expect a light bulb (over its stated life) to operate (as labelled) every time it is switched on. If it does not light-up on a few occasions, it is not “dependable” and if it turns on every time but its light output varies, then it is not “secure”. For any one of the two flaws, we deem it as unreliable.

Resiliency is generically understood as “the capacity to recover quickly from internal or external difficulties”. It does not mean that a system will not shutdown, but should restart (or recover), once such conditions are removed. There is a major difference here between “operational recovery” (from dynamic or transient disturbances) and “black start recovery” which includes a full post-disturbance manual restart. As a simple example, we expect our home furnace or air-conditioner to restart after a power restoration. We also expect its controls to manage limited dynamic power-quality issues and keep the operation going.  When such recovery is not quick (or much delayed), we deem the unit (or system) as being non-resilient.

Power systems (like most critical infrastructure) is built on reliability and resiliency. The design of such supply in its generation, transmission and distribution components (each with its own redundancy), is based on indices at various levels (equipment, MV, HV, network). Key Performance Indices (KPIs) are often an integral part of regulatory monitoring. For example, a Loss of Load Probability (or System Index) of 0.1 day/year would mean an average expected customer power outage of 2.4 hours over a full year of 8,760 hours (or 0.0274% unavailability or 99.973% uptime). Most advanced power systems are built to such standards with a few like Dubai, Singapore and Japan boasting a near-perfect 99.999% reliability. This means every part of the power system is made adequately robust and redundant to satisfy expected distress and maintenance conditions. Plant availability (fuel, readiness, etc.) is a subcomponent in the dependability calculation. Extreme stochastic variability in load or generation (or both) will lower overall reliability, unless addressed. Today, weather-related variability is factored by ISOs in their generation forecasts, but power systems reliability planning models in all its facets appear little changed.

During operations, redundant elements automatically remove themselves (with auto-reconnect if safe to return) in response to internal or external distress (faults), with the rest of the surviving operating elements readjusting their output levels to ensure the power system continues to operate reliably. Typically, many network disturbances take place daily (generator trips, lightning, storms, floods, tree contact, equipment problems, weather, etc.), but the customer rarely notices any power interruption. Such a reliability hallmark, has allowed for economic development, high-tech industries and lifestyle amenities. Today, we value electricity at 60-80 times its energy cost (on a “societal value” basis). In case of a total blackout, there are procedures to re-start the power system in a sequential fashion using specific generation assets. Each generator when interconnected provides “rotating inertia” (frequency stability) till the whole network is fully interconnected. Any interim overloads will cause these steps to fail.

Higher penetration of distributed solar/wind generation (VRE) is resulting in centralized fossil-fired plants to shut down or operate in a ramp mode (VRE backup). Battery energy storage (BESS) is being initiated to firm-up VRE output. All this is resulting in a shift towards a newer form of “synthetic inertia” (non-rotational inertia) replacing the traditional rotating inertia. This shift needs to be factored into reliability and resiliency models. Second, as the grid gets further subdivided into (isolatable) microgrid segments with more inverter-based systems, new approaches will be required to factor such reliability and resilience. Third, high VRE penetration requires real-time intelligent load management (ILM). Thus, grid performance is already changing with VRE and will change even more in the future. Reliability and resiliency models need to include a grid dominated by VREs, BESS, microgrids and ILM. These are outlined below:

1.    Equipment Derating: Ambient temperatures, humidity, snow and rain/flood conditions have changed dramatically over the last 75 years. Appropriate derating and equipment reliability metrics must be considered now. Per US DOE, about 383 power outages in 2020 were twice that in 2017, with an annual loss of USD 169 billion.

2.    “Fuel” Variability: Fuel availability in central fossil and hydro plants is a given in reliability models. However, it needs to be included dynamically for VRE and BESS in both time and spatial domain due to (a) daily variability; (b) seasonal variability; (c) extreme weather; (d) cascading weather patterns; (e) group area proximity; and (f) BESS state of charge. 

3.    VRE Locations: Central generation plants connect to HV grids in predetermined locations. It is different with VRE and BESS as such locations are dispersed (often customer sites) and on MV grids influenced by Feed-in-Tariffs and behind the meter incentives (Net Metering, Rooftop PV). This can significantly aid or abet the power delivery reliability (local load flows, reverse power limits).

4.    Synthetic Inertia: Major advances have been made in synthetic inertia area but they have not been included in reliability and resilience models. It is still a black-box for the most part. The vendor models will be critical going forward. Other rotating-inertia augmentation may be required.

5.    VRE Backup: Fossil generation (coal, NG) backs up VRE today. Many are on a base-load of just 30% rating, delivering multiple ramps daily. Second, while BESS significantly improves ramping performance, it has limited energy ratings. Third, reliability of grid interconnection (as backup) may be compromised if both sides are VRE intensive and impacted by the same weather.

6.    Intelligent Load Management: Realtime load control improves VRE dispatchability and accommodates high penetration (above 50%). However, very few adaptive load-control is modelled today (except as load loss) and do not include the reliability of its sub-elements such as wireless telecom, DR call-response success ratio, customer owned ILM systems, etc.

In closing, as climate change force much higher VRE penetration, complex models will be needed to ensure high reliability and resiliency of the power system. Additional capacity/redundancy will be required to maintain the same reliability. A second complexity will be modelling customer owned assets. There is a distinct possibility that future black start efforts may now be inside private microgrids (big buildings or campus). In my view, many traditional grids are not robust enough for a VRE centered reliability and resiliency.