Energieonderzoek

Monitoring Energy Efficiency in the Food Industry

C.A. Ramirez

Summary and Conclusions

Introduction

Prior to the industrial revolution, people depended primarily on renewable sources of energy: animal power, human labor, flowing water, solar energy, wind and biomass combustion. With the development of the steam engine at the birth of the industrial revolution, the use of coal and eventually other fossil fuels contributed to profound changes in production processes, farming and domestic activities. The use of fossil fuels has, however, originated environmental problems. At the local and regional level, fossil fuel energy consumption has caused air and water pollution, but it is the role of fossil fuel combustion in global climate change which has raised worldwide concern. Fossil fuel combustion is the biggest source of anthropogenic greenhouse gas emissions that are changing the composition of the atmosphere and increasing the global mean surface temperature. This thesis departs from the recognition that reducing the environmental effects of the energy cycle is a priority and that energy efficiency plays a crucial role in the transition towards a sustainable energy system.

Defining energy efficiency, measuring it, and devising specific programs to encourage it are challenging tasks. In this thesis, energy efficiency improvement refers to using less energy for producing the same amount of services or useful output. We analyse changes in energy efficiency by using the ratio of energy used to useful output. Output can be measured in economic (e.g. value added) or physical terms (e.g. tonnes of product). In the first case, the indicator is referred to as economic energy intensity, in the second case as physical energy intensity. For processes that generate one single product, the physical energy intensity indicator is defined as the ratio of the energy used to the physical amount of the product. When more than one product is generated (e.g. by an industrial sector), the physical energy intensity indicator is calculated as the ratio of the energy used to a weighted summation of the different products. The weights are based on the amount of energy needed to produce one physical un it of each product (e.g. Megajoule per tonne of product). By keeping the weights constant to a reference year, the weighted summation provides the amount of energy that would have been used if energy efficiency in the period studied would have remained equal to a reference year (frozen energy efficiency development). Hence, the physical energy intensity indicator is indeed a comparison of the realized energy use and a frozen energy efficiency development. The frozen efficiency development does account for yearly changes in the structure of production. Physical energy intensity is hence an indicator that is corrected for structural changes.

The use of physical energy intensity, although recognized as a better indicator of changes in energy efficiency than economic energy intensity, has been generally limited to energy intensive sectors which are characterized by having a limited number of key products, technologies and processes. A key feature of non-energy intensive sectors is their heterogeneity. It is, therefore, not a straightforward task to identify from an extensive list of products, processes and technologies, the ones that have enough explanatory power with regard to energy efficiency. In this thesis, we evaluate whether physical energy intensity indicators provide a feasible way of analyzing changes in energy efficiency in non-energy intensive sectors at different levels of aggregation.

Historically, non-energy intensive sectors have received a low degree of attention from policy-makers and the scientific community. Attention has, however, slowly begun to increase since it has been realized that i) taken together they make up for a sizeable portion of energy demand and ii) the saving potentials are significant. However, if policy makers are to develop and implement strategies that effectively promote energy efficiency, a thorough understanding of the economic, technical and behavioral drivers underlying energy demand and energy efficiency in non-energy intensive sectors is needed. Due to the low attention paid to these sectors, this understanding is limited. This thesis focuses on analyzing historical developments of energy use and energy efficiency as well as understanding the key underlying drivers in the non-energy intensive sector. This information is important because it will provide modelers and policy makers with a good analytical basis from which to extrapolate trends in energy us e and energy efficiency as well as a historical analysis on how various factors such as level of activity, changes in production mix and efficiency affect energy use. We use the food sector as a case study of the nonenergy intensive sector.

Scope of this thesis

The overall aim of this thesis is to examine the role that energy efficiency and other factors have played in the development of energy use of non-energy intensive sectors, with special emphasis on the food industry. Specific goals are:

  1. To study the developments in energy use, energy efficiency and sector structure in non-energy intensive industries of the Dutch manufacturing sector.
  2. To develop physical energy efficiency indicators for monitoring changes in energy efficiency in the food and tobacco industry at different levels of aggregation.
  3. To analyse the historical relations of fossil fuel demand and food production in the European food supply chain.

In each case, we identify and analyze the activity and structural drivers behind the development of energy use. This thesis is composed of three parts, one for each of the specific goals mentioned above. The results of the various parts and chapters are summarized below.

Summary of results

Part I of this thesis (Chapter 2) focuses on examining patterns of energy consumption and economic energy intensity in the non-energy intensive part of the manufacturing sector. The study is conducted using empirical data for the Netherlands for the time period 1988-1999.

A main question behind the research in Chapter 2 was to analyze if the lack of attention paid to the non-energy intensive sector is justified. The answer is no. On the one hand, we found that between 1988-1999, energy consumption in the nonenergy intensive sectors increased by 3% per year on average. Furthermore, in absolute terms, the non-energy intensive sector has been the sector driving the increase in total energy consumption of the Dutch manufacturing sector. On the other hand, aggregate economic energy intensity of the non-energy intensive sector increased between 1988 and 1999 (6% in the case of energy use per unit of value added, or by 2% for energy use per unit of production value).

A decomposition methodology (the Multiplicative Log-Mean Divisa Method) was applied to single out the effect of changes of structure (industrial mix) and increased production from changes in economic energy intensity of the 55 individual sectors. The results show that: i) the increase of aggregate economic energy intensity was primarily caused by an increase in the economic energy intensity and not by changes in structure; ii) structural effects had only a major role for fuel intensity, and only if value added is used as the measure of economic output. In all the other cases (electricity and primary energy for value added and all types of energy for production value), shifts in industrial structure had a minor role in limiting increases in energy intensity; iii) output growth has added further energy requirements to those induced by energy intensity, and iv) the use of value added as economic measure of output tended to amplify structural effects.

Finally, an analysis of production value and energy consumption growth rates pointed out no signs of decoupling of energy and output. Due to the observed strong link between manufacturing output and energy consumption, and if no changes occur in the current trends, it is expected that without additional policies, the nonenergy intensive sector will in the future contribute to an increase of energy consumption in the Netherlands. Given the trends found in this chapter, the nonenergy intensive sector should be considered a key target area for energy efficiency improvement and the reduction of carbon dioxide emissions.

Based on the results found in Chapter 2, a closer look to the non-energy intensive manufacturing sectors was taken. Using as a case study the food industry, we analyze historical patterns of energy use and energy efficiency at different levels of aggregation. Part II (Chapters 3, 4 and 5) examines whether it is possible to develop physical energy intensity indicators that provide a reliable estimate of changes in energy efficiency in the food industry.

The first chapter of Part II, Chapter 3, has as subject of research the dairy industry. The main goals of this chapter are twofold. First, to analyze the trends in energy use by the dairy industry in four European countries: France, Germany, the Netherlands and the United Kingdom. Second, to develop and apply indicators that can be used to monitor trends in energy efficiency. We carry out the analysis for the time period 1986-2000.

In the year 2000, the dairy industry consumed about 52, 34, 16 and 14 PJ of primary energy in France, Germany, the Netherlands and the United Kingdom respectively. The dairy industry of these four countries was responsible for the emission of about 6 Mt CO2 (39% of which are related to the indirect emissions caused by electricity consumption). Changes in energy efficiency were monitored in two different ways. First, by looking at the energy use per tonne of milk processed by dairies (EEIp1). Secondly, by comparing the actual energy use with a frozen energy efficiency development (EEIp2). The latter indicator corrects for differences in product mix among countries and in time. The first indicator, EEIp1, has several appealing characteristics. It is easy to calculate, requires few data, can be understood by nonspecialists audiences and easily communicated. It has, however, a main drawback: it does not reflect important changes in product mix which were found to be substantial . EEIp2, on the other hand, is a much more complex indicator to calculate and the data burden is higher. However, it accounts for differences in structures among countries and changes in product mix. It also allows for refinement as better information and data becomes available. We found that changes in production mix are important in three of the four countries and therefore EEIp2 should be preferred when comparing levels of energy efficiency in the dairy industry.

Once changes in product mix have been taken into account, our results show that while in Germany, the Netherlands and the United Kingdom the dairy industry has reduced their EEIp2 values by more than 1% p.a. (2.1%, 1.2%, and 3.8% respectively) , EEIp2 values for the French dairy industry have declined at a significantly lower rate (0.4% p.a.). We found that throughout the period studied, EEIp2 values for the French dairy industry were larger than for the other countries (e.g. in 2000, the EEIp2 values were 30-40% larger).

An analysis of the possible causes behind the differences between countries, especially between France and the other three countries, shows that higher EEIp2 values calculated for the French dairy industry can be related to the fact that the French dairy industry works at a smaller scale, is highly fragmented and has shown a relatively slow pace of concentration. Although there is a lack of detailed information, our results suggest that the French dairy industry is using different technologies (which are more energy intensive) to produce dairy products. A more detailed analysis of the French dairy industry revealed that while the cheese and butter making sub-sectors have decreased substantially in their EEIp2 values, the decrease has been offset by an increase of the EEIp2 in the ‘other milk products’ category (milk powders and condensate products). Consequently, values for the whole French dairy sector appears to remain nearly constant. Finally, our results also show tha t the British, German and Dutch dairy industry have converged towards similar (lower) values in their energy efficiency indicators and that the French dairy industry would save about 30% if it were to converge to similar values of EEIp2 as the ones reached by Germany or the United Kingdom.

Following a similar structure to Chapter 3, Chapter 4 examines patterns of production, energy use and energy efficiency in the meat sector (production and preservation of meat plus the further processing of meat products) of France, Germany, the Netherlands and the United Kingdom. The analysis is carried out for the time period 1986-2001. In the year 2001, the meat industry demanded about 39, 35, 10 and 32 PJ of primary energy in France, Germany, the Netherlands and the United Kingdom respectively. 40-60% of this amount was used in the further processing of meat products. The meat industry of the four countries studied was responsible for the emission of 4.5 Mt CO2 (58% of which are related to the indirect emissions caused by electricity consumption). Our results also show significant increases in primary energy use per tonne of product: France (3.2% p.a.), Germany (3.4% p.a.), Netherlands (1.4% p.a.), and the United Kingdom (1.6% p.a.). We found a trend in all countries towards higher electricity use due to increasing demands for refrigeration and motor drive power generation.

In order to understand the drivers behind the trends, factors such as the share of frozen products, the share of cut-up products and increasing food hygiene measures were analysed. We found that correcting the indicator for changes in the shares of frozen, cut-up and deboned meat products i) decreased the absolute values of the energy efficiency indicator, ii) decreased the gap between countries, and iii) explained some of the fluctuations showed by the trends at a disaggregate level. Nevertheless, these changes in product mix cannot explain the increase of the energy efficiency indicator displayed by the whole meat sector. The impact of stringent hygiene regulations on energy demand was stronger. It explains between one and two thirds of the increase in the energy efficiency indicator.

Chapters 3 and 4 show that patterns of energy consumption and energy efficiency can be monitored and analysed at lower levels of aggregation. Chapter 5 builds on the data and results found in Chapter 3 and 4 and expands the analysis to a higher level of aggregation. In Chapter 5 we develop indicators to monitor energy efficiency in the whole food, drink and tobacco industry (hereafter food industry). The analysis is based on physical production data at the firm level provided by the Statistics Netherlands on a confidential basis. The analysis was carried out for the time period 1993-2001.

We measure energy efficiency by using as indicator the ratio of the current energy used and a frozen energy efficiency development. We selected 49 product categories which account for 51% of the total food categories. The coverage obtained for the base year (1995) was about 81% for fuels/heat and 60% for electricity. Our results show that the Dutch food industry has improved its energy efficiency indicator in primary terms by about 1% p.a. (uncertainty range between 0.9 and 1.3) . In terms of final energy, there has been a decrease of the indicator for final demand of fuels by about 1.8% p.a. while there has been no improvement in the indicator for final demand of electricity. Furthermore, we estimate that increased penetration of combined heat and power (CHP) in the food industry since 1993 has saved about 3 PJ primary energy in the Netherlands.

In order to assess whether or not the product mix studied is representative, we compare the average annual change in physical production of the products take into account in this chapter against those left out. We found that both groups exhibit similar behaviours. Hence, we conclude that the products selected reflect important structural changes in the food industry. Furthermore, we found that the development in energy efficiency found in this chapter is coherent with the reported implementation rate of energy conservation projects and with developments reported by the Long Term Agreements, which confirms the reliability of the approach and the results.

We conclude that the type and the quality of the data compiled by Statistics Netherlands for the food industry are sufficient to develop indicators as required by energy and climate policy. This is an important finding since it means that energy efficiency in the food sector can be monitored by an energy agency without needing to implement a task force that depends on company reporting which is done with the sole purpose of monitoring developments in energy efficiency (as done by the Agency for Energy and Environment NOVEM in the Netherlands). This is a very promising outcome not only because it is rather likely that similar analyses can also be conducted for other non-energy intensive industries in the Netherlands, it also gives rise to hopes that similar analyses for non-energy intensive sectors can be conducted for other countries. The sole condition is that detailed production data can be made available (for example on a confidential basis as in the Netherlands).

In Part III (Chapters 6 and 7), we expand the system boundaries by including into the analysis agriculture, fertilizers and transport. In this way, Part III takes a system approach when examining the dynamics and interrelations of energy and food production.

The first chapter of Part III, Chapter 6, assesses energy demand due to world fertilizer consumption in the time period 1961-2002. Energy embedded in fertilizer consumption is considered the most energy intensive part of the food chain. In the first part of Chapter 6 we develop historical trends of specific energy consumption and gross energy requirements by kind of fertilizer and assess the energy embedded in world fertilizer consumption. The trends obtained are later used in Chapter 7 as part of the inputs needed to calculate total energy demand in the food supply chain. In the second part, we explore whether technological development in the fertilizer industry can be analysed by using the concept of learning or experience curve. According to our analysis, in the year 2001 the primary energy embedded in world fertilizer consumption was about 3660 PJ (1% of the world total energy demand in 2001) of which 72% was for the production of nitrogen fertilizers, 10% for phospha te fertilizers, 16% for complex fertilizers and only 2% for potassium fertilizers. Total energy demand increased by about 3.8% p.a. between 1961 and 2001. The highest average annual growth rate was shown for nitrogen fertilizers (4.5 % p.a.) followed by compound fertilizers (3.9% p.a.).

In order to understand the development in embedded energy, we apply a decomposition methodology which allow us to single out the effects of increasing fertilizer consumption and changes in fertilizer mix. There are three main findings. The first is not surprising: growth in fertilizer consumption has been the main driver of increasing energy consumption. Second, fertilizer mix has moved towards fertilizers that are more energy intensive per kilogram of nutrient, which has further increased energy demand. And third, significant improvements in energy efficiency have occurred, but they have not been able to offset the increasing effect of other factors. A comparison with best available technologies reveal a saving potential for the year 2001 of about 19% (687 PJ). This potential is mainly allocated to the nitrogen fertilizer industry.

In the second part of this chapter, we look at technological development in energy efficiency as a learning process. As far as we know no attempts have been previously done to use the concept of learning curves for the development of industrial energy efficiency. Most published material on experience curves relates prices to the cumulative production of a technology. We relate the historical trends in specific energy consumption of various nitrogenous fertilizers to cumulative production. Our results show that specific energy consumption of ammonia and urea developed in close concordance with the learning curve model, showing progress ratios of 71% for ammonia production (R2=0.997) and 88% for urea (R2=0.856). This is an important result since middle and long-term models of energy consumption and CO2 emissions face the difficulty of how to consider technological change. The use of progress ratios can provide an alternative approach to include technological change into scenario developments. Another consequ ence of our findings is that for energy intensive industries for which classical learning curves (i.e. based on prices) cannot be developed due to high dependences on market prices and strong fluctuations of raw material prices, the analysis of specific energy consumption as a main indicator of technological development can provide a way out to analyse rates of technological change.

The final chapter of Part III, Chapter 7, provides an extended historical analysis of the relationships between fossil fuel demand and food production. The system studied includes the energy used by the agricultural, food processing, transport and fertilizer sectors. The physical flows taken into account are output of the agricultural sector, import of agricultural/semi-processed products, export of agricultural products, waste due to transport and storage, waste during processing and output used for sowing. The analysis was performed for thirteen European countries (Europe-13) in the period 1970-2002.

Our results show that in the year 2002 the food supply chain in Europe-13 required about 3960 PJ of primary energy. This corresponds to about 7% of the energy used by Europe-13 in the same year. In total primary energy use of the system studied increased by 1.6% p.a. We found that the only part of the system showing decreasing energy demand (about 2% p.a.) is the energy needed for the manufacture of fertilizers. Agriculture, food processing and transport have increased their energy consumption at a rate of 1.6%, 1.8% and 2.3% per annum respectively.

In this chapter, we choose as functional unit the output of food and fodder expressed in value added and calorie content. We estimated that the physical output of the total food supply chain in the year 2002 was about 924 petacalories (or 3868 PJ), while the economic output of agriculture and food processing was about €342 billion of value added (5% of total GDP). Physical output of the total system has grown at a rate of 1.8% p.a. while economic output has grown at a rate of 3.6% p.a. Our analysis also shows that economic energy intensity decreased at a significantly higher rate (2.7% p.a.) than physical energy intensity (0.2% p.a.).

One consequence of the system boundaries chosen is that we consider both food and fodder as outputs of the system. We found that if animal feed is excluded from the system, the rate at which the physical energy intensity of the system decreased in the last three decades is significantly lower (0.2% versus 0.04% per annum). We also assessed the influence of using calorie content as the functional unit of output by comparing the output trends both in terms of calories and protein content, and found no significant differences (less than 2% variation per year). Another consequence of the chosen system boundaries is that the system is not closed: the energy use for the production of imported agricultural/semi processed products is not taken into account. However, an assessment of the impact that energy embedded in imported products could have on the energy demand trend of the food supply chain indicates that our conclusions are not affected by the choice of system limits. In this chapter we also examine the dri vers behind the energy consumption developments. We found that between 1970 and 2002, increased demand for feed production, increased calorie consumption per capita, and increased tonne-kilometer transported have been the main drivers of energy demand of the system. Output growth has increased primary energy demand by 1.8% p.a. on average while changes in physical energy intensity offset this demand by only 0.2% p.a. Thus, with exception of fertilizer manufacture, changes in physical energy intensity have not played a significant role in decreasing total energy demand in the European food supply chain.

The limited role that physical energy intensity has played so far is an important result. In the ongoing debate on dealing with greenhouse emissions, expectations are high that improvement in energy efficiency can help to significantly decrease total energy consumption without hampering the future economic growth of the nations. Because of the increasing production of processed food, the increasing amount of tonne kilometres of food transported and given the findings of this study, unless there is a significant change in the rates at which physical efficiency improves, we are far away from curbing the net energy demand in the European food supply chain.

Lessons learned

There are several conclusions that need to be stressed. First, in this thesis we have shown that, from a methodological point of view:

  • It is possible to monitor changes in energy efficiency based on physical production data in heterogeneous, non-energy intensive sectors.
  • Physical energy intensities provide a fair and feasible way of comparing energy efficiency developments among countries.
  • Decomposition methodologies, which are generally applied to monetarybased energy analysis, are an equally useful tool in physically based energy analysis. They allow large amount of information to be studied and provide a way to quantitatively analyze the impact of different factors in energy consumption/intensity.
  • Technological development in energy efficiency in energy intensive sectors can be analyzed by using the learning curve concept.
  • Provided that detailed and reliable data is available (in public databases as used in Chapter 4 and 5 or obtained on a confidential basis as in Chapter 6) and that detailed studies of energy efficiency at the industrial level for a base year exist, historical changes in physical energy intensities can be monitored without needing to implement task forces that depend exclusively on confidential reporting at the firm level.
  • The biggest limitations found in this thesis are due to data availability and data quality. A sizeable amount of the research time has been spent obtaining reliable time series with sufficient level of detail to allow the drivers behind energy efficiency changes to be analyzed (e.g. this proved to be a major constraint in understanding energy efficiency changes in the meat sector). These limitations may be a major deterrent of performing the kind of analysis made in this thesis for other sectors and/or for other geographical regions.

Second, we have shown that in the last decade, the non-energy intensive manufacturing sector in the Netherlands has increased total industrial energy demand and that it has demanded more energy per unit of output. The results highlight the need for policy-makers and scientists to increase their attention to the non-energy intensive sector and encourage industries in these sectors to adopt energy-efficient technologies and management practices.

Third, comparisons of economic and physical energy intensities in the food sector reveal large differences in the rates of decline, with economic energy intensities declining up to 3% p.a. faster than physical energy intensities. In other words, the value added of the food sector has grown at a significant faster rate than physical production in the last three decades. This difference indicates a decoupling of physical production and economic growth in the European food sector. Fourth, our results have shown that in the food sector changes in physical energy intensity have not even been close to offset energy demands imposed by growing output. In fact, there are no signs of decoupling between energy consumption and growing output in the period studied (a similar result is found in the analysis of the Dutch non-energy intensive manufacturing sector in Chapter 2). There is, of course, no contradiction between the third and fourth conclusion. Decoupling physical production and economic growth do not imply per se a reduction in the amount of energy used. Energy is only one of the factors of production. Decoupling can also be achieved by increasing labor or capital productivity. The comparison of economic and physical energy intensities not only reveals that energy efficiency has played a minor role in the decoupling of physical production and economic growth in the food sector, but also indicates that the use of economic energy intensity as an indicator of energy efficiency in the food sector fails to reflect current changes in energy efficiency.

Finally, the results found in this thesis are a source of concern because they suggest that energy policies have failed so far to make significant improvements in energy efficiency in the European food sector. Although some sectors have shown significant improvements in their energy efficiency (e.g. dairy industries), we found these improvements to have mainly been driven by concentration processes, which in most cases offer limited scope for the future. The general picture indicates that increased efforts should be made by industry and policy-makers if we want to reach energy savings that contribute significantly to the reduction of greenhouse gas emissions.

We conclude this thesis by pointing out the importance of including physical flows into energy analysis. Understanding and addressing the consequences of physical processes requires them to be dealt with not only in economic terms but also in physical terms. Ultimately, it is difficult to see how sustainability (and the ways to achieve it) can be addressed by policy makers in the absence of information regarding the time-dependency of energy and material flows within the various subsectors of economies.