On Energy Efficiency in Industry

(harder than it need be) [Nov 2023]

This post covers my personal views on developing energy efficiency improvements in industry. It is purposefully light on real data as I want to keep it high-level. I will be writing more detailed research posts in the future to expand in this. Reach me at dm@duncanmcmillan.xyz for comments.

TLDR: I believe that improving the energy efficiency of industrial processes is hard, and harder than it need be. I believe that some relatively simple software tools can transform the way industry finds and evaluates process improvement opportunities.

Intro:

Today on site*, it is rare that a person - or an effective group of people - simultaneously has the knowledge and decision-making power to enact optimal process improvements. At the site level, engineers do not have the time or technological expertise to ideate and select improvement opportunities1. Beyond the site team there is a lack of expert knowledge and context of what can really be implemented on site2. Lastly, outside experts struggle to quickly and reliably understand the sites process enough to truly deliver optimal improvement projects3.

*site: a factory, plant or production asset.

I believe that the process of discovering, quantifying, and verifying improvement ideas is needlessly laborious. Relatedly, I believe that industry under rates the value of improving knowledge management internally, and knowledge sharing with third parties. I (and ~everyone) also believe that getting industry to its sustainability targets (~Net Zero) will be very hard, as this requires transformational changes, not just incremental improvements to low level systems. Industry lacks software tooling to tackle these problems.

Today, there is no robust, effective and easily shareable processing modelling tool focused on delivering energy saving opportunities. Today, many sites have reasonable submetering and a data analysis platform without advanced anomaly detection tailored to industrial processes. Lastly, there is no AI assistant tailored to helping engineering teams on industrial sites.

The Energy Saving process and its challenges:

Energy Saving Process4:

Typical challenges:

Gathering documents: Often diagrams are out of date, contradictory, or non-existent. The same goes for client made models (mental, excel, advanced software). Different teams and personnel have different responsibilities/focuses and often they not only have different levels of understanding – as expected – but they often disagree. Finding or creating reliable diagrams and models of the current process is laborious and error prone.

Gathering data: Metering coverage is typically less than ideal, as there are valuable metering points (they would pay for themselves via insights) that are not metered. Often the site's metering data is stored and only accessible across multiple systems. Occasionally sites have untrustworthy meters. Occasionally there are important systems that rely on manual readings and/or do not have telemetry and software making the readings easily available. There is less ambiguity on what good data points exist or don’t exist, but there is ambiguity where the client engineering team are required to assume a figure or make an estimate. Sites and third parties struggle to gather and share good data points whilst understanding the errors and uncertainties present in the data.

Gather non-formalised data: I have painted a picture of there being a lack of formalised documents and data. This is surprising to people outside of industry, but a fact of life inside it. A huge amount of the insight generation process relies on information solely held in the minds of key personnel. For a given subsystem it really is often true that the domain expert fully understands and can fully explain the nuances of the system, but the docs, diagrams and data relating to that system get nowhere near that level of insight. Unfortunately however, there is often a genuine lack of understanding or contradictory opinion. Sites should have improved knowledge capturing tools and processes. The process of formalising knowledge held inside people’s minds is laborious and error prone.

Get an understanding of the existing process, by building a model: A direct consequence of the points above is that this task is laborious and error prone. Note that by ‘build a model’ I am including mental models and the most basic models possible “gas and water in, steam and hot air out”. I am referring to a model good enough for an improvement to be selected, not some grand atomic scale simulation of the system.

Evaluate room for improvement: For most opportunities the qualitative improvement is self-evident: temperatures are fluctuating, air is leaking, valves are hunting, control loops are off etc. For a subset of these the quantified savings benefit is easy to calculate, but often the calculation is in a grey zone where the uncertainty bound sits across ‘attractive investment’ and ‘unattractive investment’. For operational improvements, control improvements, and repairs – improvements requiring no new kit – the improvement at the specific system can easily be modelled but tying that process improvement to a financial saving involves modelling the larger system, at least to the point where utilities are consumed. Quantifying savings is hard because the model of the current process is unreliable, and often the model of best practice is also unreliable. Quantifying the savings from new kit installs relies on the estimates and guarantees of third parties (OEMs, vendors, consultants). These are largely reliable within the systems boundaries and assumptions stipulated by the third party. The issue faced here is dealing with the subtleties of the given process, i.e. what is the exact solution required and can the third parties kit truly meet the needs of the system. Matching third party solutions to the given system is hard. This usually involves an ongoing discussion where the third party requests data and discusses options. Some third parties do offer online calculators and product selection tools that massively streamline this process, but for big ticket items the process is slow and arduous. I believe it can be streamlined. Because the process is time intensive a client site never fully compares all relevant designs, they often stick with preferred OEMs and vendors. The most effective way of reducing a large projects cost is picking a better project to begin with! There is huge downstream value to nailing project identification and FEED.

Create Cost Benefit Analysis: This is a challenge due directly to the problems above. Modelling the expected savings is hard, picking the right solution (to give a cost and expected savings) is hard.

Identify how to enact best solution: Again, finding the best solution is hard, most install problems come down to not having the right solution in the first place, other than that it is ‘merely’ a project management issue. As noted the largest impact on a project's success, cost and delivery come from its inception, not the later phases where large amounts of money starts moving hands.

Footnotes:

1 Added nuance on site teams: In general, I am assuming that site engineers are an effective group that are experts in their process. They lack knowledge of outside technologies, relying on 3rd parties (OEMs, vendors, consultants) to bring in that outside knowledge. Unfortunately, they also often do not have the time and/or incentive to really give energy efficiency improvements the attention they deserve. This is all of course a generalisation. Many sites do have dedicated Energy Managers that hold real sway in the org. Some orgs do have a strong sustainability culture where even the lowest level technician can discover and drive process improvement. Some engineers (or engineering teams) do have outside knowledge and expertise in energy efficiency.

2 Added nuance on exec level: Out with the site team executives have at best a high-level understanding of sites processes, and can only really discuss large energy supply projects, which to some extent disregard to complexities of downstream processes. They also very rarely have technical knowledge. At this level there is almost always a deep desire (and incentive) to pursue energy saving projects.

3 Added nuance on 3rd parties: Consultants (with no ties to particular products) have limited knowledge of the products available and struggle to get the context required to select the best solution. OEMs and vendors do have deep technical knowledge on their solutions, but not others, and even if they did they would ~never recommend them. Outside experts do exist who could really provide insight, but its very time intensive to give them the context (knowledge) about the specific site process.

4 This is a high-level summary that relates to both a third-party process and an internal one. Note that a client team would typically start at evaluating plants systems and evaluating room for improvement, as they already have access to docs and data, and a deep knowledge of site. Of course, they often go back to docs and data to improve understanding, spot errors and update them. In general, this is more of a continuous process for client sites doing this themselves. Those familiar will note a similarity to ISO50002 and other energy audit processes.