I was recently at the INFORMS 2019 conference in Seattle. It is a great conference to see the wide array of applications using optimisation. One interesting talk was the IFORS Distinguished Lecture by Andy Philpott titled Zero Carbon Analytics. The topic of this talk was the development of a model to analyse how New Zealand will go about achieving their goal of transitioning to 100% renewable electricity generation by 2035. This was a great use of optimisation in the real world to answer question of great societal importance.

The topic discussed in this blog is quite complex and I will only provide a broad overview of the work. For people interested in the work performed by Andy Philpott, I suggest exploring the Energy Power Optimization Centre website. In particular, one should read the paper “100% renewable electricity with storage” by M.C. Ferris and A.B. Philpott.

New Zealand’s Goal

The key points of New Zealand’s goal are:

  • Introduce a Zero Carbon Act and establish an independent Climate Commission.
  • Request the Climate Commission to plan the transition to 100% renewable electricity by 2035 (which includes geothermal) in a normal hydrological year.
  • Stimulate up to $1 billion of new investment in low carbon industries by 2020, kick-started by a Government-backed Green Investment fund of $100M.

We can unpack this a little bit. The main point is the second one above, where they aim to achieve 100% renewable electricity generation in a normal hydrological year. This introduces some flexibility in achieving the 100% target, since in an abnormal hydrological year, it is possible to draw upon fossil fuel based sources. Also, this requirement is quite important, since New Zealand primarily relies on hydro as a renewable source of power generation, so if they experience a dry season, then there may not be enough water in the reservoirs to satisfy the electricity demands.

The main role of operations research or optimisation in this process is to inform the government in determining the main goals or objectives, better understand the effect of uncertainty in achieving the goal and how the government can best use public investment as an incentive.


Uncertainty is a major factor that must be considered when modelling energy systems. Especially, when considering renewable energy sources. For example, solar is affected by the amount of sunlight and hydro is affected by the amount of water stored in the reservoirs.

The main source of uncertainty for hydro is river in-flow to the reservoir.

The uncertainty associated with hydro power has long range implications. Different from many other renewable sources, the water reservoirs can be used to store energy. This potential for storage can be used to “carry over” energy from one time period to another, which is not possible with solar or wind. However, this ability to store energy means that when considering forecasts of energy demand and rainfall, you may over store water. Just so you have some reserves. This is all a result of the uncertainty in the energy demand and weather.

In the modelling of Ferris and Philpott, the scenarios address the uncertainty in the wind and the run-of-river. Also, there is some consideration of solar and thermal plant electricity generation technologies.


Since uncertainty is considered in the modelling of the energy system, this requires a special class of optimisation problem—Stochastic programming. A stochastic program is an optimisation problem that minimise an objective function incorporating some expected cost over a set of scenarios. As mentioned above, the scenarios primarily look at the uncertainty affecting wind and hydro power generation, but there is also consideration of solar and thermal power generation.

The New Zealand capacity expansion model developed by Ferris and Philpott is called Gemstone. This is a two-stage stochastic program. There are many components to this model, some of the key aspects are:

  • Minimising the fixed cost associated with capital investment in different technologies and the annual maintenance costs.
  • The operating costs are uncertain, so the expected operating costs are also minimised.
  • Constraints model
    • Hydro electricity storage
    • Battery storage
    • The transmission of electricity through the grid
    • The shifting of demand, which could be modelled using the battery storage constraints.
    • Constraints on non-renewable electricity generation, such as capacity restrictions, generation restrictions and emission restrictions.

Something that I found interesting from the paper by Ferris and Philpott was the point about shifting demand. In regions where energy systems from different countries are connected, such as in Europe, then it is possible to import energy when renewables are unable to satisfy demand. In New Zealand, due to the isolation of the energy system, this is not possible. As such, a demand response that is required is for industrial users to shift their production to other countries where energy is more plentiful.

New Zealand Case Study

A number of experiments were performed using the Gemstone model. I will briefly discuss the results from Experiments 2.

In Experiment 2, the aim was to investigate the effect of increasing the constraints on i) non-renewable capacity, ii) non-renewable energy generation and iii) the average CO2 emissions. This investigation looked the change relative to 2017 levels in each of the three constraints. The results showed that modest reductions in CO2 emissions can be achieved from constraining i) and ii). Alternatively, the best reduction in CO2 emissions comes from directly constraining the average emissions. The results did show, that this latter case was also quite costly. However, a 95% reduction in emissions relative to 2017 can be achieved with only moderate cost increases.


New Zealand has taken a bold step by setting a goal of 100% renewable electricity generation by 2035. Determining the expected outcomes in achieving this goal is a problem that can be solved by operations research. This is an amazing example of optimisation being used in the real world to inform public policy.

Stephen J Maher

Author Stephen J Maher

More posts by Stephen J Maher

Leave a Reply