AIMMS has always had a tight relationship with academia, so we love to highlight professors, students and researchers who are using our software to solve problems around the world. Recently, we had the pleasure of speaking with Prof. Nnamdi Nwulu from the University of Johannesburg. Nnamdi recently published a book titled “Optimal Operation and Control of Power Systems Using an Algebraic Modelling Language.” He and his co-author, Dr. Saheed Lekan Gbadamosi, used AIMMS to model and solve many of the mathematical optimization problems found in modern power grids. His book is a great entry point for energy modeling students and practitioners. Its aim is to democratize energy system modeling. Read our conversation below.  

Let's do a quick round of introductions, would you like to start Nnamdi?

Yes. My name is Nnamdi Nwulu and I am an Associate Professor at the University of Johannesburg. I am currently the Director of the Centre for Cyber-Physical Food, Energy & Water Systems (CCP-FEWS) at the University of Johannesburg as well.

I have a background in electrical engineering. My specific field of research is energy modeling and the intersection between machine learning and optimization. I also study the nexus between food, energy and water systems. I am also exploring how to use machine learning and optimization in these three areas. I discovered AIMMS in 2012, during my PhD. When I started my studies, I looked into the mathematical optimization of energy systems and specifically the incorporation of demand response programs. Most AMLs (Advanced Modeling Languages) have paid licenses, which made them unsuitable for graduate students. The AIMMS Free Academic License was ideal for my research, so I used it throughout my PhD studies and still use it today.  

What about you, Saheed? 

My name is Saheed Lekan Gbadamosi. I am a lecturer and am presently undergoing my Postdoctoral studies under the supervision of Prof. Nwulu. I majored in energy optimization and power systems. I did my PhD under Prof. Nwulu, where I focused on preserving power quality while accounting for renewable energy integration. A lot of my prior research was centred on smart grids, microgrids, and demand response. I am currently working on computational intelligence applications in energy optimization. I got to know AIMMS through Prof. Nwulu, while undergoing my doctoral studies.

What are the top challenges you see in modeling energy systems today?

Nnamdi: The key challenges relate to the increasing complexity of energy systems. This is due to several factors. First, the power system has grown rapidly with many changes. These changes include increased penetration of renewable energy and consumers who double as producers of energy, popularly referred to as prosumers. This makes power flow in the power system bi-directional. Allied to these are emerging concepts such as peer-to-peer energy trading. This adds complexity and uncertainty to the power system, which makes planning & dispatch difficult.

Furthermore, you see increased demand for efficiency and transparency due to climate change. Consumers want to see where their energy is sourced from, how much greenhouse gas emissions they emit and so on. Lastly, we now need to model human behavior in energy systems. These issues may differ slightly in different regions, but they all contribute to increasing complexity in modeling energy systems.

Can you elaborate a bit more on the context? How have energy systems changed, say in the last 5-10 years and what does this mean for optimization?

Nnamdi: Two concurrent shifts have happened over this period. First is the ongoing energy transition, and second, advancements in emerging technologies. You have more digital technologies like the Internet of Things (IoT) and Blockchain (peer-to-peer networks), which track the flow of energy in power systems. There is much more widespread deployment of these technologies and pilot projects. Optimization will have a role to play here. For instance, in peer-to-peer trading, you can use optimization for the optimal determination of market prices.

How do you define Blockchain for those who are not familiar with it and how is it being used in power systems?

Nnamdi: There are three key terms or concepts used to define Blockchain. Decentralization, transparency and immutability. Blockchain is a distributed ledger technology, which can contribute to running the power system more smoothly. It can be used as a tamper-proof system to foster transparency, record-keeping and data harvesting. There is a lot of data available from consumers nowadays—for example, data on pricing or tariffs, energy production and consumption. This data can be used to improve operations.

Let’s talk a bit more about renewables. Solar is one of the most notable technologies on the increase. According to your book, there was a steep rise in solar energy between 2010 and 2018 (from roughly 50TWh to more than 500TWh). How does this impact energy systems modeling?

Nnamdi: For long-term planning, most countries have targets. Many are decommissioning fossil fuel-powered stations and increasing renewable energy sources. For this, you have to factor in various governments' targets. For instance, having fully renewable energy by 2030 or 2040. You have to factor in effect, where to place renewables, and when to include them considering these targets - in a bid to reduce emissions. You also have to consider how much demand you expect in the future and how you envision the growth of renewable energy-powered generators in the system. You model to be able to plan ahead. From a modeling perspective, you have to improve optimization methods and forecasting models to handle increased uncertainty. Stochastic programming or robust optimization are increasingly being used.

What about power quality and short-term planning?

Nnamdi: For shorter-term planning, or operational planning, you have to consider economic dispatch down to the day and optimal power flow in seconds or milliseconds.

Saheed: Yes. When carrying out an optimal dispatch, you have to ensure it is at a minimum operational cost and also, the customer must get value for their money. That is, the system must be reliable all the time. This is challenging with renewables because of their variability and uncertainty (for the leading technologies like solar and wind). For example, you have to deal with power quality issues such as harmonics, which are constant fluctuations propagating through the grid.

Can you explain what harmonics are about? This was not an issue with other means to generate power, can you talk about this?

Saheed: Renewable energy sources such as wind and PV (photovoltaics) power electronic devices based on their mode of operations, contributing harmonics into the grid. Harmonics are nonlinear loads, and they degrade power on-grid assets, like transformers, for example.  As a modeler, you have to minimize the effect of harmonics on the grid. The harmonics contribution on the grid has been a major concern for power system planners, and this can lead to a reduction in voltage quality, affect the lifespan of electrical components, drive power losses and add ripples to generated power. Therefore, it is necessary to monitor harmonic emissions and keep them within the specified standard limits to avoid poor quality of power supply and enhance power quality management.

Nnamdi: Yes, harmonics are a serious issue that adds complexity to regular energy models. Sometimes you need to connect nonlinear devices into the power system, and they distort the system's fundamental frequency. This distorts power quality.

And how do you integrate that in an optimization model?

Nnamdi: Actually, we have a paper on that.

Saheed: Yes, in one of our papers, we look at the harmonic contribution from renewable energy source components such as wind turbines, PV panels and the HVDC (high-voltage, direct current

) transmitting medium. We model the components as nonlinear loads. The paper focuses on developing an optimal power flow optimization model for minimizing power loss and harmonics emanating from RES (renewable energy systems) with emphasis on reducing their impact on the Locational Marginal Prices. Grid modeling and simulation were done in order to identify and quantify the harmonic quantities at buses. A MINLP (mixed-integer non-linear programming) model was developed, and we have harmonic loss as an objective function to be minimized. Therefore, the proposed model ensures harmonic minimization and reduction in marginal price reduction. Harmonics cannot be eradicated, but they can be minimized.

That’s an interesting piece of information. So, it’s not just about optimal power flow but you also talk about physical impacts. Models also help to prevent major breakdowns, like load shedding in South Africa or what we saw recently in Texas. Can you talk about that a bit?

Nnamdi: Yes, the aim is to understand the complex entity you want to operate. Energy systems require extensive modeling and simulation. In the case of Texas, power equipment was not winterized. The state gets most of its energy from natural gas-fueled plants, not renewables. These plants experienced failures due to the freezing temperatures. This type of energy crisis has enormous political, economic and social consequences. Over 5 million people in Texas were left without power for more than 3 days. Some suffered carbon monoxide poisoning to stay warm, and others got hypothermia. Not to mention the economic fallout.

Modelers need to think of effective planning, improved reliability, and energy security to prevent these issues. But they must also consider cost reduction. Consumers want reliable power, but at an affordable price.

You book covers the different types of energy modeling, but I didn’t find information about modeling uncertainty. Is this a deliberate choice?

Nnamdi: Yes, it’s a choice. I can share with you some of our papers on robust optimization, harmonics minimization and so on. This book sets out to provide an entry point for anyone coming into power systems modeling. It is also an aid for post-graduate students that are entering the field. The book breaks down the whole process. We did not cover some of the more advanced concepts that we have covered in some of our papers.

Yes, the beauty of your book is that you don’t have to be an expert in power grids to understand the trade-offs that go into modeling. We found it super interesting that your book compiles all the basic models you need to know about. Can these type of models also be applied in other areas, like water or food management?

Nnamdi: Yes, in fact, that is our next project. We are applying algebraic modeling languages to food and water systems. The models that we developed for energy systems can easily be deployed in these other systems. For instance, we have a municipal water network that needs to be expanded. There is the need to anticipate the demand, the places where you will locate pumping stations, the locations for reservoirs, etc. In many ways, it is similar to a generator and transmission planning program. We actually have some of these models already in AIMMS.

We also have a model that has to do with food distribution. So, suppose you have biomass produced from food, and you want to optimize its movement throughout the whole supply chain. In that case, you can use AIMMS to assess where to site storage locations to move the biomass from the supply side to the final end-use consumer. We are also working on game theory, incentive pricing, water demand response and other aspects of the food supply chain.

So this is actually the first book in a series of wonderful modeling books, in food, water and other areas?

Nnamdi: Yes. It is easy for us to pivot into those areas. The consensus is that it is often better to optimize food, energy & water systems as a whole as opposed to individual optimization. If you have a power system that is based on fossil fuels like in South Africa, those plants will use a lot of water. Suppose you only optimize the amount of water to be produced without considering its impact on water flow and water demand. In that case, you will have an optimal system in the energy space, but it will be sub-optimal in the water space.

This year, we tried to develop a framework for a small pilot in South Africa that encompassed food, energy and water and the various technologies that can be deployed to manage this nexus.

What about the objective function? You have to compare things that are sometimes not commensurate. Would you use multi-objective?

Nnamdi:  Yes. The argument has been whether to use a weighted function or Pareto approach. This would open up the question of bias in the model. Typically in models, you have multiple objectives – from economic to environmental to incentives. So, we use weighting functions to weigh multiple objectives. We typically try to give equal weightings but depending on the system and modeler; you can tweak the weightings depending on the system you want to give preference to.

If we go back to the biomass example, you use food crops to produce energy. But that also has an impact on food security. It reduces the amount available to feed your population. So, you have to manage those trade-offs. Do you focus on power output using all the biomass you can find, or do you also consider food security? It depends on the modeler and the area. For example, if an area has more food security issues, you should place more focus on that. But in areas where you have severe power shortages, but there is enough food, more weight should be given to power generation. It all depends on the trade-offs.

I’d like to talk about some use cases in your book. For example, you talk about retrofitting inefficient energy appliances in South Africa. Can you tell us about this?

Nnamdi: Yes. So, the problem is this: you have a lot of energy wasted on inefficient devices. For instance, you may have huge industrial facilities with incandescent light bulbs or inefficient engines. Some of them may not have the financial muscle to invest in energy-efficient replacements. Some time ago, the utility company in South Africa tried to incentivize end-users to replace these inefficient devices with energy-efficient ones. The problem is that in most cases, efficient devices are pricier. Consumer willingness to invest in these replacements was relatively low. The utility company said, “ok we will give you some money to replace them.” The question was: how much should be given? The utility is subject to constraints. It is not practical for them to replace all inefficient devices with energy-efficient ones.

We set out to find out what the program participation requirements should be. For instance, in situations where consumers are able to save 2 megawatts of power per year. We found out the cost prices for some of the technologies in the basket of the program. We determined the optimal price that would be mutually acceptable for both the utility and consumers.

Thanks for illustrating this use case. It’s a pretty good example of applying energy modeling, but including some economic concerns. We can see the limits of that, of course, because we are trying to model human behavior and sometimes that’s not possible. However, testing strategies is possible. What are your plans for the future and what’s your dream for this field going forward?

Nnamdi: I presume in the next 5 years, there will be much more computational capacity and so more powerful modeling algorithms and techniques. We will see increased deployment of what we call digital imaging technologies, and this will impact modeling. More data will be generated from the power system as well due to IoT. We are trying to be at the forefront of this space and have two books on the horizon.

The first one is on Energy 4.0 - how the 4th industrial revolution will change the face of the energy industry. Maybe in about a year, this book will be published. We are also aggressively pushing for an optimization book on integrated energy systems, including food and water. We expect increased complexity due to emerging technologies. Because modeling seeks to gain insights, we need to create much more advanced models to capture the interplay and complexity of these systems.