ABSTRACT
This paper proposes an innovative strategy to optimize the energy metering in large and energy-consuming plants such as industrial flour mills. The proposed solution deals with ISO 50001 implementation which represents a critical challenge for many companies because the benefits due to improvements in energy management could be potentially canceled by the costs of an Energy Management System (EnMS). In particular, the Key Performance Indexes (KPIs) monitoring is a crucial activity for several reasons: it is one of the early activities, it affects the measurement quality of KPIs and it deeply impacts the EnMS requirements and, consequently, the investment valuations. The proposed strategy supports the energy managers in the design of the energy monitoring system suggesting the points of the electrical network to be equipped with sensors and monitored. Moreover, the paper describes the results carried out in a real-world application: the energy sensor network of a 1.2MW flour mill plant sited in Italy has been designed and implemented with the proposed innovative solution.
I. INTRODUCTION
This paper proposes an innovative decision-making solution that optimizes the measurement points to be applied to a milling facility to identify the most relevant KPIs. Their analysis makes possible to examine the energy consumption of the system aiming at implementing the energy efficiency policies by following the requirements of the ISO 50001, standard for energy management systems, namely:
• develop of objective-based policy for more efficient use of energy;
• use the data to better understand and make decisions regarding the use and energy consumption;
• measure and evaluate the results;
• review the effectiveness of the policy;
• continuously improve energy management.
The design of the monitoring network, such as the installation of the measuring equipment and the implementation of data analytics has to address the following issues:
• best distribution of network nodes derived from EnPIs (Energy Performance Indicators), such as typical power consumption profiles and reference energy baselines;
• monitoring of process energy flows organized for production lines, machine sets or single machines;
• assessment and segmentation of energy consumptions for evaluation of production lines and the cost centers under different operating conditions;
• identification of extra consumptions, anomalies, and faults;
• evaluation of significant gap between installed and real consumed power in order to find inefficiencies such as suboptimal working conditions and machine oversizing;
• detailed consumption analysis for definition of energy saving policies;
• development of decision support systems for technical and management teams;
• evaluation of the results carried out by the energy saving actions.
Those issues have to be faced by the EMS (Energy Management System) to operate the continuous improvement according to the methodology Plan-Do-Check-Act.
ISO 50001, published in 2011 by International Organization for Standardization (ISO), presented the guidelines and the requirements to the implementation of an Energy Management System [1], and then have been further developed by the same Organization in [2].
The reference guide [3] is a support for EMs to understand the core elements of the ISO 50001 Standard: how to develop skills to identify key processes and to develop controls and documentation, to understand the certification process of ISO 50001, how to evaluate readiness of implementing an ISO 50001 EnMS.
A relevant experience of ISO 50001 implementation is described in [4]. Among a thousand of Dutch companies which are running an energy management system, there are a number already certified for ISO 50001, while others are in the process of implementation. The paper describes how the energy management forms a part of the energy transition approach of this sector with an incredible result for the sector as a whole and the organizations specifically. This study analyzes the result expectations and barriers to overcome when the implementation of ISO 50001 can take place, highlighting how for the sustainability of the ISO 50001 management system management commitment from top management still is one of the crucial factors for success of the actual deliverance of the system. Moreover, the Sappi Mill is deeply analyzed as case study because it is one of the the first companies in the 2012. Sappi Fine Paper Europe is one of Europe’s largest suppliers produces reels, specifically designed rotation offset printing. Although the product of the plant is paper, and not flour, as in the proposed paper, both are large milling processes with several analogies. Sappi implemented the ISO 50001, and its formalized policy, with an energy management by all layers in the organization and not longer as a task of few human resources. The main achieved goals are the tools to control the production output in sellable tons of paper and the energy consumption of the mill with the available equipment. Moreover, a decision tree has been designed to manage the shut down of several boilers or turbines in the CHP to reduce energy consumption. The main achieved benefits are cost reduction and better priorities regarding investments.
One of the first problems in the context of energy management is the design of metering system and its implementation. As shown in [5], it is a complex problem that can potentially impact the cost-benefit analysis of the entire project. The authors proposed a strategy to solve the problem of decision, at an audit stage, can concern the choice of a sound instrumentation. This strategy has been applied to a company adixen Vacuum Products in a standard for decision aiding process. The paper suggests how to use ACUTA method to setup the model of preference of the Decision Maker. Finally, a standard procedure about decision, supported by a multicriteria decision-aid tool, has been defined. The preference disaggregation of multicriteria decision-aid systems (MCDA) is based on the idea to assess global preference models from the given preferential structures and to address decision-aiding activities. In [6] there is an extended review of preference disaggregation strategies with their relevant results.
In [7], the authors proposed a framework that integrates standards for energy data exchange, on-line energy data analysis, performance measurement and display of energy usage. This solution allows to achieve the desired efficiency improvements with the real-time measurement of the energy use to derive an awareness of the energy use patterns of every part of the production process. Also this paper proposed a framework for energy monitoring and management that allows decision support systems and enterprise services to take into consideration the energy used by each individual productive asset and related energy using processes, to facilitate both global and local energy optimization.
With regard to metering problem, best practices are reported in [8]. This work explains effective energy and water metering strategies, relevant metering technologies and communications, how to collect and use metered data, and how to develop a metering plan. In the third chapter, this work analyzes metering system costs which widely changes because of: equipment specifications and capabilities, existing infrastructure, sitespecific design conditions, local cost factors, etc. For this reason, the most expensive parts are shortlisted for a preliminary budget of the metering system. In the fourth chapter the metering technologies are investigated. Metering provides the data for analytics and decision support. The key to a successful metering program lies in the best trade off between the accuracy of the metered data and the ability to make use of them. The authors explain how to determine the measuring points according to metering objectives and should be a focus in metering plan.
The proposed paper proposes an innovative solution optimizes the design of the sensor network in order to make the EMS more robust and effective. This is a main challenge for the industries at the moment.
II. THE KPIS OF THE FLOUR MILLING PROCESS
A. The industrial process
The industrial site consists of the real plant where products are processed, and the storage facilities where raw materials and finished products are stocked.
In the considered industrial plant the production process are the following:
• grain-collecting (pit) and pre-cleaning;
• silage, mixing, cleaning, wetting;
• husking, milling and pelleting;
• delivery of the flour.
The two types of finished products are flours for human consumption and for zootechnical foods. In the former case both wheat and durum wheat are processed. The entire process needs to be slightly retuned and rearranged according to the type of processed products. Regardless of kind of processed products and for sake of energy analysis, the process can always be divided into the following main departments:
1. pit / pre-cleaning dept.;
2. cleaning and conditioning dept.;
3. pellet mill dept.;
4. decortication and grinding dept.;
5. warehouse of finished products (flour/semolina, bran, chopped, etc.);
6. services dept.
These departments work almost sequentially, and the chief production officer is in charge of scheduling and workload balancing among departments in order to achieve the best production efficiency. The annual breakdown of energy absorbed by the departments is shown in Figure 1. The data clearly show that decortication and grinding are the most energy intensive activities compared to other departments.
The wheat processing is made by a number of machines whose total installed capacity is about 1.2 MW. For our purposes this installed power is assumed constant in time. The electrical energy is used by compressors to produce compressed air, by aeraulic machines to transport the flour and the products of the plant cleaning, by extraction screws for the handling of products, etc. These machines are operated by 462 induction motors and absorb about 70% of the entire energy requirement. A number of 42 motors, about 9% of the total, are inverter fed.
B. KPI definition
KPI and EnPI definitions result effective if they properly synthetize qualitative and quantitative values of production performances related to the energy consumptions. Moreover, the same KPI can be often applied to the entire production process or to its sub-processes. For these reasons, KPI definitions and application levels need to be adapted to processes and products. As a consequence, also the sensor network for data acquisition will be affected by KPIs.
Because the literature lacks of KPI definitions for wheat processing, after interviews, audits, and bottom-up analysis, the authors have defined three main performance indexes that follow the guidelines of ISO 50001 and take into account the variables daily controlled by the experienced technicians in charge of production. The three KPIs are defined as follows.
The first KPI regards the capacity production and is defined as the following dimensionless ratio:
ηh = heff / hop (1)
Where heff is the operating time, expressed in equivalent hours the plant works to its full capacity production. hop is the operating time expressed in hours, i.e. when the plant is operating in any condition. This first KPI can be simply applied to the entire process or to the single departments. It represents the ratio between the equivalent hours at full load and the actual hours of operation. If the system worked constantly at maximum capacity, then ηh = 1.
The second KPI is related to the energy processing performance and is defined as:
ηw = qin / win (2)
Where qin is the mass of incoming row material to be processed, and win is the energy required for processing. ηw is a ratio usually expressed in quintals over kWh. This energy processing performance can be applied to the entire plant such as to each department.
The third KPI is a production efficiency index defined as:
ηpr = qout / win (3)
Where qout is the mass of finished product obtained. Also for this index: win is energy required for processing, the ratio is usually expressed in quintals over kWh, and this index can be applied to a single part of the process and is related to its efficiency in terms of outcome product per unit of energy consumed.
To calculate these KPIs, the sensor network needs of scales to measure the weight of products and materials, and modern power meters or network analyzers to record working hours and energy consumptions. In order to get better results for energy audits, it is important that the installed energy analyzers, further normal specifications, have the following technical features:
• calculation the cumulative electrical energy (active, reactive and apparent) in user-defined periods;
• remotely programing of the trigger for the cumulative energy to directly relate the outcomes to particular operating conditions.
C. The electrical network
After KPI definitions and to setup the sensor network, it is necessary to map the electrical system starting from general electric panels. The electrical system of the mill is shown in fig. 2. This system can be represented with a tree chart of a typical industrial system with 20 kV medium-voltage supply.
This voltage is lowered to 400 V by two transformers to feed all the equipment and machines (USER). TRAFO 1 is the transformer that feeds the panel QG1 in charge of the energy distribution to the real industrial process. Transformer TRAFO 2, with the panel QG2, provides energy to the auxiliary services. Level 2 is made of all the panels and lines that provide energy to the department through the panels Qx. It has to be noted that each panel Qx can provide users or subpanels of more department, and vice versa, each department can be fed by more panels Qx.
The tree nodes are considered as points of interest for analysis because they can be equipped with measuring instruments. The nodes are divided into categories according to their loads. If all the nodes derived by a root belong to the same department, the root node is classified in the same way. In other case, the parent node is a mix node.
III. OPTIMIZATION PROBLEM
The optimization problem aims at finding the best set of nodes to be equipped with sensors. In other words, among all the possible sensor network setups, the best trial is the best trade off between installation costs and gain for the EnMS (Energy Management System), i.e. the capacity to calculate more KPIs (for smaller set of users) with sufficient accuracy. The possible solutions are between the following two extreme scenarios. The former is the solution that implements only the general power meter, as a consequence only global KPIs can be calculated with minimum installation costs. Vice versa, the latter scenario considers the installation of sensors to cover all the users and to get the most accurate monitoring of energy consumptions.
Moreover, the fitness function for this optimization problem has to score all the trials taking into accounts the following constrains:
• relevance of nodes depends on the power absorbed,
• users with rated power less than 7 kW can be neglected;
• user types such as fan, auger, compressor, etc. are more relevant than the others;
• nodes with same kind of users can be seen as single grouped user;
• installation costs of sensors in case of single or grouped users.
For those reasons, the authors derived a fitness function from the model obtained in [5]. The criteria adopted to design the fitness function are:
1. Compatibility with the DCS (Distributed Control System) already installed. The EnM (Energy Manager) assigns to each node a value from 0 to 10. The higher the value is the simpler the SCADA integration is.
2. Energy use related to the node. There is evidence that the node supplying more energy-consuming users need to be better monitored. For this reason, a normalized 0-10 scale has been used. This scale is obtained considering for each user its rated power and working hours.
3. Expectation in a gain for the EnMS. This is a qualitative criterion because it regards the impact that each node provided with sensors can have on EnMS performances. As in [5], the EnMS according to his experience can assign a value (among the following five: 0, 2.5, 5, 7.5, and 10) to each node.
4. Installation cost. This criterion implements the installation costs of the sensor network in the fitness function. A node provided with sensors needs about 500 euro as total installation cost divided as follows: about 190 euro for the power meter, 40 euro accessories, 270 labor cost.
IV. EXPERIMENTAL RESULTS
After optimization runs described in previous section, the sensor network has been designed as shown in figure 3. The red nodes are those interested by sensor installation. The setup suggested by the optimization process has been deeply analyzed with the energy manager.
At first glance, it has to be noted that, except levels zero and one, all the following levels are not fully provided with sensors. The level-1 sensors on both QG allow to trace the total energy consumptions, hence to calculate the global KPIs as defined in the previous section. Moreover, all the productive zones of the plant can be fully profiled in terms of energy use because well covered by the sensor network. The distribution of the sensors can be described as follows:
• distribution panels after transformers
- general power panel, QG1,
- general services panel, QG2,
• panels of departments
- pellet mill,
- pit / pre-cleaning,
- products warehouse,
- framework cleaning,
• combination of users and panels
- decortication and grinding,
- cleaning and conditioning,
• all the users operating more than 7,500 hours per year with power more than 20kW.
The sensor network setup obtained with the proposed approach has been approved by the energy manager. The same setup has been compared with the one initially proposed by the energy manager completely based on his experience. In terms of performance, the number of monitored KPIs is the same with comparable accuracy. The difference between the two solutions is the installation cost, the proposed approach carried out a setup whose installation costs are 70% less than that of the initial solution.
V. CONCLUSIONS
This paper proposes a real-world application of optimization method to support the design of the sensor network for an energy management system. This case study regards the flour mills, but can be easily adapted to other industrial sites. A variety of practical issues, well known to industrial technicians, have been considered. This allows to carry out a sensor network setup that has been approved by the energy manager and the chief production officer replacing a first proposal based on industrial experience. Moreover, the number of monitored KPIs and their accuracy prove the effectiveness of the proposed approach.
REFERENCES
1. “ISO 50001 - Energy management,” online available: https://www.iso.org/iso/home/standards/managementstandards/iso50001.htm
2. “Win the energy challenge with iso 50001,” online availablee: https://www.iso.org/iso/iso_50001_energy.pdf
3. “Guidebook for ISO 50001 Energy Management System”, The Hong Kong Electronic Industries Association (HKEIA), www.hkeia.org, March, 2013.
4. R. Vermeeren, B. Mulder, W. Post, M. van Veen, M. Reijnders, “Implementation of ISO 50001 in industry in the Netherlands,” ECEEE 2012 Summer Study on Energy Efficiency In Industry, pp. 680 – 688, 2012.
5. Rizzon, B.; Cliville, V.; Galichet, S.; Ochalek, P.; Ratajczak, E., “Decision problem of instrumentation in a company involved in ISO
50001,” in Industrial Engineering and Systems Management (IESM), 2015 International Conference on , vol., no., pp.409-416, 21-23 Oct. 2015, doi: 10.1109/IESM.2015.7380190.
6. E. Jacquet-Lagreze and J. Siskos, “Preference disaggregation: 20 years of experience,” European Journal of Operational Research, vol. 130, no. 2, pp. 233–245, 2001.
7. K. Vikhorev, R. Greenough, and N. Brown, “An advanced energy management framework to promote energy awareness,” Journal of Cleaner Production, vol. 43, no. 0, pp. 103 – 112, 2013.
8. SA Parker, WD Hunt, KM Fowler, WF Sandusky, GP Sullivan, BK Boyd, KL McMordie Stoughton, TM Koehler, R Pugh, “Metering Best Practices: A Guide to Achieving Utility Resource Efficiency, Release 3.0”, PNNL-23892 Release 3.0, Pacific Northwest National Laboratory, March, 2015.