Loading...
Northern Powergrid (NPg)2017-11-13T22:12:42+00:00

Project Description

CLIENT

A power company covering Yorkshire and the Northeast.

PROBLEM

NPg has hundreds of thousands of energy assets – switchgear, transformers, cables, poles, conductors, fittings and others. To extend the life of the asset, maintain the overall fault rate and eliminate unsafe assets, they employ a condition and risk-based intervention strategy. This includes a well-established programme of work, based on a set budget, to replace or refurbish assets, which prevents them from failing and maintains performance. However, the assets deteriorate at different rates, over time, resulting in different failure modes. So, in order to assist the end-to-end planning process NPg has, over the last decade, developed a suite of condition-based risk models. For example, by examining individual asset and component level, NPg’s existing approach was to use a ‘health’ threshold above which the assets (or their sub components) are selected for intervention.

In reality, as these suite of decision support tools evolved over a number of years, the decision modelling was often applied inconsistently across asset categories and was performed on complex, slow-running Excel files, which were difficult to maintain and manipulate.

BRIEF

NPg needed a method to identify the list of interventions in order to effectively minimise risk in the asset portfolio, using a defined budget.  The intervention decision strategies needed to be consistent across multiple asset categories using risk as the single comparable metric.  Once the list of interventions was chosen, NPg needed a method to allocate them sensibly to a year within the delivery period.

In addition to addressing these problems, NPg wished to test and potentially introduce, innovative ideas:

  • A method to select the schemes optimally within a limited budget, by trading off schemes between asset categories using a single metric of risk
  • Using criticality (i.e. the consequence of asset failure) as part of the intervention decision-making consistently across all asset categories
  • Building an asset intervention model for asset categories where a health model does not yet exist
CHALLENGES
  • It was not possible to apply identical decision strategies across all asset categories, so intervention strategies needed to find the right mix of standardisation and realistic decision-making.
  • Within each asset category there were often multiple asset types, each requiring slightly different treatment.
  • For some asset categories, the intervention decision is taken at an asset level, and for others at a circuit level; in addition, the aggregation methods were different in most categories.
  • Some intervention rules selected parts of a circuit to be intervened on, and the application of differential intervention schemes is hard to model.
SOLUTION

decisionLab built a decision model – the Asset Risk Model (ARM) – which allows NPg to apply consistent intervention strategies across multiple asset categories to minimise the risk in the portfolio, under a range of budget scenarios.decisionLab worked with NPg to understand the details of the intervention strategies in each asset category.  Together, we developed a consistent and realistic approach to modelling the interventions that removed legacy models and implemented the latest thinking in all asset categories.  decisionLab implemented the method of Multiple Threshold Criteria to allow users to define several thresholds for intervention, including criticality.  This gives the user the flexibility to fine-tune the measures by which an asset should be selected for intervention.

Conceptual Modelling, Mathematical Modelling, Analytical Modelling, Agent Based Modelling, Multi Method Modelling

Fig 1 – Costs per asset category in each delivery year following cross-asset optimisation

decisionLab built the ARM in AIMMS – a leading commercial optimisation modelling software tool.  This has reduced the decision-making time from weeks to hours.  The data is managed safely, reducing the likelihood of spreadsheet error.

The ARM can be run in four budgetary scenarios, giving flexibility to the user to explore which thresholds and budgets provide the optimum list of interventions.  This ranges from the ‘essential’ to the ‘unconstrained budget’ list of interventions.

The ARM can be used to identify an optimal [i] set of interventions when considering multiple asset categories simultaneously.  Interventions are compared by their reduction of risk per unit cost of intervention using a single risk metric applied across categories.

data analytics, data science, optimisation, simulation, asset management, artificial intelligence

Fig 2 – Risk profile over the delivery period shown for each scenario as a % of risk as year zero

The ARM provides high level outputs for regulatory reporting and drilldowns for asset managers.  The intervention rules are user-configurable.  The ARM can either be used to find the best plan for a fixed budget, or find a budget for a given intervention strategy.

Once the best plan has been found, the user can allocate each intervention in the plan to a year in the delivery period at the click of a button.  This provides a useful list of the work required to be done and the year which it should be done in order to reduce portfolio risk as quickly as possible whilst not overloading the delivery teams in a particular year.

decisionLab built a simple version of the ARM for an asset category where no health data was available.  This uses the Weibull distribution (Fig 3) to estimate the health of the assets according to their age.  This method demonstrates how the ARM could incorporate more difficult asset bases in future, despite limited data available.

Conceptual Modelling, Mathematical Modelling, Analytical Modelling, Agent Based Modelling, Multi Method Modelling

Fig 3 – The user can adjust the shape and scale of the Weibull distribution which indicates the rate at which asset health deteriorates according to an asset’s age (no health data available)

FUTURE PLANNING

NPg can use the ARM to help establish future programmes of work and to test whether current planned delivery is optimal.  The ARM outputs can feed into the annual asset reviews which form the basis of the business plan.

decisionLab has documented the new approach to intervention modelling so that NPg have a starting place to discuss future innovations.

The next phase of this work will incorporate more asset categories and new intervention rules including linking groups of interventions between asset categories.

FEEDBACK

“The scope of this project has evolved considerably over its duration due to the emerging requirements of our business.  We are delighted with the work to date; we have amalgamated, simple, flexible and consistent models which are a testament to the flexibility, patience and technical ability of the decisionLab team.  There are some obvious next steps for us which we are excited to explore further in subsequent stages of this project.”

Gavin Howarth, Policy and Standards Engineer, Northern Powergrid (NPg)