Proactive Decentralized Historian-Improving Legacy System in the Water Industry 4.0 Context

Abstract

The industry is in continuous evolution in the context of the Industrial Internet of Things (IIoT) and Industry 4.0 requirements and expected benefits. Some sectors allow higher reconfiguration dynamics considering the interference capabilities and process/equipment renewals, but others have considerable inertia that is many times justified. In most encountered situations, the reality confirms that the industry is struggling with new demands such as interoperation and efficiency improvements. The water industry makes no difference, being a sector with critical infrastructures and highly varied subsystems, where invasive interference in legacy solutions tends to be avoided. Following previous successful footsteps in researching a proactive decentralized historian, the current work focuses on a case study that refers to a water treatment and distribution facility that has been operated for several years, and the current operating regime was established by local operators following accumulated observations, restrictions, and response strategies. The proactive historian was tailored for the current case study, and it was applied and tested in the suboptimal functioning scenario where the water sources configuration was manually selected and used for water availability and energy efficiency, but without assuming current/future failures or different water demands. The proposed low-cost historian targeted to improve the functioning and operation of the water facility considering energy efficiency and other impacting outcomes of the current strategy and to establish an automatic functioning regime in a completely non-invasive manner towards the local legacy solution. The results were satisfactory, proving that the historian is able to adapt to a particular and suboptimal functioning real industrial scenario, establish recipes in a process-aware manner, and interoperate with the local legacy solution in order to apply improving actions.

1. Introduction

Sustainability in automation and Supervisory Control and Data Acquisition (SCADA) directly impacts the associated industrial processes and must be argued together. A sustainable system has to focus on long-lasting productivity and quality, on environmental impact, and on cost issues. These outcomes may derive from each other, or a proper strategy may impact their majority. A new solution has to foresee future sustainability, but obviously, a higher impact derives from adapting legacy systems. The pace of progress in Operational Technology (OT) and Information Technology (IT) is usually higher than in the actual industrial process structuring-related research. However, industrial processes need constant improvements, changes, and expansions. Also, the current status of the Industrial Internet of Things (IIoT) and Industry 4.0 is providing various opportunities to raise the level of improvement and sustainability [1,2,3].

Companies from various sectors are targeting IIoT/Industry 4.0 benefits and are struggling with interoperation and efficiency improvements. By promising a very high potential for improvements in productivity, cost reduction, availability, and safety, the aforementioned concepts became the main areas of focus for both academic and industrial actors. In order to achieve this high potential, efforts are channeled into obtaining better connectivity, information exchange, and interoperability between different industrial entities, with the aim of creating links between previously isolated subsystems. The new frameworks that are being created will serve as the backbone of intelligent software solutions that are expected to gain capabilities to optimize technical systems, maximizing their performances and durability. Such advances are crucial in light of growing concerns regarding industrial sustainability in the context of climate changes, the transformations started by Industry 4.0 and IIoT presenting undeniable contributions towards emission reduction, energy savings, reduced pollution, and a cleaner environment [4]. This aspect is commonly underrated because it is indirectly derived compared to the more direct, measurable, observable outcomes of the new solutions.

Numerous research directions have branched out under Industry 4.0 guidance [5], with consistent advances being noted lately in machine learning [6], artificial intelligence [7], security [8,9,10], and edge computing [11]. By focusing on superior interoperability and information exchange as core features, the Industry 4.0 concept naturally led to research results in the communication protocols area as well. Different solutions around the Open Platform Communications Unified Architecture (OPC UA) [12] have supported a general acceptance in recent years of OPC UA as the standard Industry 4.0 communication protocol, required by technical systems that need to become Industry 4.0 compliant, even though alternatives are available [13]. Those improvements in communication are supporting data accumulation, which is demanding the integration of Big Data techniques into the Industry 4.0 development directions, research such as [14,15] recently explored this path. Further, emerging directions under Industry 4.0 umbrella can be noted, briefly enumerating cloud computing, networking, plug and produce, information models, data models, fog computing, and standardization, among others.

Depending on the industrial sector, the access to reconfigure, change, interfere, and renew processes, solutions, and equipment can be more or less restricted. All industries have inertia, but some of them have a lower reconfiguration dynamic. For example, the automotive manufacturing industry is often expanding or changing its production lines due to new or changing requests from its clients, but the water industry is much more reluctant to solution change or too invasive interference in its current systems. Therefore, legacy systems are predominating in the water industry and constitute a multitude of different technical solutions and processes, maintaining a heterogeneous and chronologically dispersed perspective. Processes and process components change through time; control strategies need adjustments, and efficiency increase is necessary, so improvements are mandatory with respect to non-invasive interference and sustainability. This practical reality of struggling to adapt to Industry 4.0, in the context of water treatment and distribution facilities, means that issues like high energy consumption, equipment failures, high consumption of substances, maintenance, or water sources quality changes continue to persist on a large scale [16,17]. Although representing viable solutions at the time of their development, many technical implementations from the water industry became outdated, and more efforts are required from both academic and industrial actors in order to leverage the massive potential benefits and opportunities for improvements in this industry that is critical for human health and the environment in the same time [18]. From an energy efficiency perspective, research [19] considered 146 drinking water treatment plants (DWTP) and concluded that around 50% of the consumed energy could be saved, the main reasons being the source of raw water, the age of the DWTP, and used technology for treating water. Furthermore, regarding the water treatment technologies and methods, an analysis focused on energy impact is presented in [20].

After previous successful footsteps in researching a proactive decentralized historian, the current work focuses on a case study that refers to a water treatment and distribution facility that has been operated for several years. Regarding various process changes, observations, restrictions, behaviors, and learned response strategies, the operators established a local operating regime. The proactive historian is tailored for the current case study, and it is applied and tested in the suboptimal functioning scenario where the water sources configuration was manually selected and used for water availability and energy efficiency, but without assuming current/future failures or different water demands. This suboptimal scenario was chosen because it represented a very important challenge considering some criteria. The energy consumption was close to minimal, respectively the water demand in this period could be assured by two water sources at their close to optimal functioning points throughout each day. The proposed low-cost historian targeted to improve the functioning and operation of the water facility considering energy efficiency and other impacting outcomes (e.g., longer working hours of the personnel, inability to respond to rapidly changing water demands, process equipment fault due to heavy use, etc.) of the current strategy, and to establish an automatic functioning regime in a completely non-invasive manner towards the local legacy solution. The strong partnership and previous joint research between academia and the industry allowed important solution tailoring and real-world testing with only a few restrictions, assuring faster reach to higher technological readiness levels. The aim was to prove that the historian is able to adapt to a particular and suboptimal functioning real industrial scenario, to establish recipes in a process-aware manner, and to interoperate with the local legacy solution in order to apply improving actions.

The paper continues in the second section to present the previous work regarding the proactive historian, which is the essential starting point of the current research. The third section depicts the case-study solution tailoring in the context of the water treatment and distribution facility. The fourth section presents and discusses the obtained results, respectively the last section concludes the work.

2. Previous Work

Local data-collecting tools are crucial for offering the possibility of having insights into local processes. From a more practical perspective, the most often available data-collecting tools are placed at top supervisory levels only and are offered by well-known automation software producers, which are still preserving their own ecosystems in terms of collected data availability and means to access it. Alternatively, recently, data collection has been conceived even further away, towards the IT level or even the cloud. But the gathered data from the Operational Technology (OT) level relies only on the central control center SCADA software data, which is usually filtered and reduced in volume, adapted for the central operators’ perspective, with a lower level of knowledge regarding the local processes. A water distribution company may have hundreds or thousands of local processes dispersed in a large geographical area. The level of understanding of the local technological process functioning and the access to all variables is the highest locally, within the water plants. Water plants are typically containing redundant SCADA servers, but few SCADA software environments allow proper data storage and access. SCADA software represents the locally available data collection alternative. Usually, SCADA software solutions rely on logging services for a certain tag set and rudimentary archiving services. Backward data access is many times almost impossible (e.g., long waiting times, access only through trend graphs, few tags perspective, some with their own format databases, etc.), respectively no real data processing and manipulation possibilities are available. The classical historians are costly and general-purpose data gathering and manipulation systems, therefore being placed on central or regional control centers or larger water plants. The availability of historical data regarding equipment and processes operation proves to be the first obstacle for some potential exploratory research ideas or solutions. As researched and developed in [21], lightweight and low-cost historians with abilities to interface using emerging Industry 4.0 (e.g., OPC UA) and classic protocols are key elements for local data collection and storage. The legacy systems tend to have clear limitations in terms of data storage and accessibility, the long-term data retention and access to historical data being achievable through decentralized historian solutions, aspects that are crucial for performance evaluation, trend analysis, and regulatory compliance [21,22,23].

With the data available, the evolution of software tools is opening up the possibility of analyzing the collected data and identifying and optimizing recipes in an autonomous manner. The research towards data dependency analysis and pattern recognition within the historian was presented in [22] as a step towards even more intelligent and sophisticated solutions, capable of optimizing the monitored system without requiring human supervision or intervention. The non-invasive character of the solution was important, as the historian was foreseen to be implemented also for legacy systems.

The proactive historian concept was further advanced in [23], where the accumulated data, the dependency analysis, were added with implemented recipes to reach an objective function, considering inserted constraints. The solution was applied to a water facility in two steps. The first step was only collecting data that was applied to the plants’ model. A 9% energy consumption reduction resulted from selecting water sources according to a quality indicator and functioning hours, respectively, setting a flow requirement for each water well. The second step consisted in applying a reduced form of the solution to the real water facility. The test scenario was short-term, and the solution was not allowed to set the flow requirement exactly as calculated by the algorithm, and some actions were strictly controlled by the operators. The historian only selected the water sources; the flow set point was a fixed value for the local control loops. However, the results showed a reduction of energy consumption by 30% in the water plant.

Work [24] closed the general control loop, bridging the gaps for autonomous functioning. In the sense that the water quality indicators were automatically adapted, and steps were made towards a process-aware historian, adapted for the water sector.

3. Case-Study Solution Tailoring

3.1. Water Treatment Facility Revisited Challenges

An efficiency-increasing solution cannot be researched without real data and real industrial scenarios. The hypothesis that guided the current research relied on a water treatment facility that represented the testbed for the previous status of the proactive historian. The water treatment facility consists of 6 water wells that are foreseen with flow-based main local control loops and a level-based secondary control loop. The water enters the treatment plant on a common pipe. It is aerated and filtered with sand and charcoal filters, respectively chlorinated in 3 points before being distributed in the network. Figure 1 depicts the mentioned components within the process aware setting interface of the proactive historian, where all process components are represented, with all constraints and objective functions.

The historian is using the collected data to establish well priorities and flow set-points according to estimated water quality indicators (water quality based on the previous impact over the energy consumption) and functioning times (impacting the future availability of the water well), respectively to current and predicted water demand (output flow in the distribution network and water accumulation necessities in the reservoirs for peak consumption hours).

As mentioned in [23], the previous research focused on reducing energy consumption in the context of all process components functioning in an automatic regime, limited historian deployment (only selecting the water sources but without setting the calculated flow set-points within the control loops of the wells), respectively limited-time testing. The results were very satisfactory, reducing the energy consumption in the plant by 30%. Throughout the following years, the situation changed within the water facility. Targeting minimal energy consumption and following several observations regarding water demand and components functioning (including the inability to automatically establish flow set-points within the water wells), the automatic activation and selection of water sources based on functioning hours and levels in the reservoirs was stopped by the operators. Therefore, the current situation provided several challenges for research:

-

The close to the minimal room for energy reduction. The operators select the water sources and choose the best available options according to previous research and observations. Depending on the season and the weather, some situations allow a minimal number of good selections (e.g., the test scenario within the current work covered a situation when only two water sources were able to assure the necessary water for normally expected demand, using flow set-points that determine close to optimal frequencies for the drives);

-

The number of selected wells are functioning all day at the same flow set-point regardless of the water demand. This behavior benefits energy consumption pricing due to lower night tariffs. However, the manual water well selection regime will not cover situations of varying water demand nor situations where faults occur. The outcome would be either wasted water (usually water is wasted) or lack of water for the population (later accumulation of water can be difficult in some periods of time);

-

The manual regime cannot distribute the functioning time correctly for the water sources. This may lead to faulty behavior or defects within the well. Currently, the first water source cannot function on flow-based control loops anymore because of slower provision of water, and the fourth water well is not able to keep the flow set-point, having large fluctuations. This leads to an even more limited room for energy efficiency improvement but a considerable necessity for a complete automatic regime.

Besides the mentioned challenges, a supplementary challenge is to tailor and stress the proactive historian for longer testing periods in autonomous functioning and non-invasive interoperation with the legacy system.

Therefore, these challenges are transformed into historian tailoring and testing tasks, briefly presented as follows:

-

To assure the automatic selection of water wells according to accumulated data with the energy efficiency increase as the primary objective function but also considering functioning hours of the water wells;

-

To assure the automatic activation of the water wells according to varying current water demands, but considering the accumulation perspective for peak hours, the varying flow set-points in the local control loops of the water sources, and other constraints as the upper, lower, and optimal limits in pumps driving frequencies;

-

To interoperate with the legacy system in order to set the flow references and to start/stop the pumps. This task assumes multiple checking and protection functions according to the current manual regime and other functioning regimes at the water source level (e.g., steps in starting and stopping the pumps, level-based regime check, faulty behavior check, faults detected in the local systems, interoperation sampling periods proper setting, etc.);

-

To generate general safety procedures to deactivate properly the historian-based interference that provides the automatic regime in case of malfunction signs or on operator demand, with an option to reestablish the previous settings within the local system before decoupling;

-

To evaluate the overall performance in the current suboptimal regime and longer-term functioning.

3.2. Architecture and Solution Deployment

The proactive historian solution developed in [24] formed the basis for the practical implementation of the current research, but special tailoring for the current case study was required in order to fit the software solution to the specific particularities of the tested scenario. All five tasks identified at the end of the previous subsection were considered, and the adjustments made to the historian software solution in order to meet those aforementioned requirements are briefly described below.

Firstly, the optimizing algorithm described in [24] was adjusted in order to include a hysteresis h as the percent of the minimum flow of the water source. More specifically, if the difference between the total flow delivered towards the population and the sum of the water flows offered by the sources that are running is not greater than ½ of the minimum flow that can be offered by the next idle source (in the order of its priority), then the respective water source is not started, to avoid excessive water pump wear down. The same hysteresis limit applies to stopping a water source.

$$\begin{gathered} h=\frac{1}{2}·minimum\_next\_source\_flow\_setpoint \\ \left|flow\_required\_from\_source{s}_{actual}-flow\_required\_from\_source{s}_{previous}\right|>h \end{gathered}$$

if

$$flow\_required\_from\_source{s}_{actual}>{\displaystyle \sum }_{i=1}^{active\_sources}water\_source\_flo{w}_{maximal\_value}$$

or

$$flow\_required\_from\_source{s}_{actual}<{\displaystyle \sum }_{i=1}^{active\_sources}water\_source\_flo{w}_{minimal\_value}$$

The percent was established at a fixed value using the historian, and it proved to be more efficient than varying hysteresis correlated to the water accumulation necessities and minimal values of the flow for a specific water source. A varying h value for each water well requires a proper frequency of updating and proper correlation with the evolution of each water source. No benefit was encountered for a varying h.

Secondly, the historian was updated in order to read the minimum and maximum flows that can be offered by each water source from a configuration file so that those limits can be easily adjusted by operators if necessary. The limit values are important as it was noticed that, through time, these limits were changing for each water well. Also, the automation from the DWTP used in this research takes into account a specific OPC UA tag for determining when to start or stop the water pump of a source instead of using a 0-based convention for the reference flow tag, which led to the necessity of adjusting the algorithm from [24] in order to properly set this start/stop command tag value as well when it is required. Taking the manual selection of the water wells towards the automatic regime required more condition checks and further reaction toward the local functioning of the legacy solution. For example, the names of the OPC UA tags used for start/stop command, sources reference flows, and total flow delivered towards population had to be set (together with proper verification procedures), as well, particularly for this DWTP. Compared to [24], the magnitude of algorithmic and protection structures referring to local system interoperation following the efficiency recipes provided by the historian increased significantly.

Lastly, in the targeted DWTP, the filter washing operation takes place at approximately 24 h intervals, and it consumes around 50 m³ of water from the water tank containing the treated water, which will be sent to the network. The respective water tank has a maximum capacity of around 400 m^3, and the filter-washing operation takes just around 30 min to complete. So, in order to compensate for this significant drop in the distribution tank’s water level in a short time interval, the historian was adjusted to compute the target water flow (total flow that must be offered by the water sources) not as equal to the water flow towards the distribution network, but as:

flow_required_from_sources = p% × DWTP_output_flow.

(where p% was set for 120%, but it can change between 110–130% depending on the actual value of the treated daily volumes within the DWTP, which varies according to the season).

This way, the water level in the tank is slowly but constantly rising, compensating for the big drop during filter cleaning. In this context, all experiments were monitoring the frequency of the filter washing cycles and filter clogging status.

The interfacing between the specifically developed version of the historian application and the legacy system from the DWTP targeted by the current research is presented in Figure 2.

The historian implementation has two loops through which it communicates with the legacy system: the monitoring loop and the optimizing loop. The monitoring loop is used by the historian to collect data, reading the OPC UA tags values at 20 s intervals from the OPC UA Server that is functioning in the DWTP. The values are collected inside the legacy system from the field equipment (pumps, air blowers, filters, chlorine injectors, flowmeters, etc.) and exposed to the OPC UA Server, where only real-time values are available. Those real-time values are being read by the historian at 20 s intervals, and the values are being stored within an SQLite database, thus collecting historical data about DWTP’s functioning parameters. The values associated with OPC UA tags are components of internal DWTP control systems. For example, the reference flow for water sources represents the set-point of the flow-based control loop within each water source (primary closed-loop control system for each well), and the actuators are frequency converters that adjust the revolution speed of the pumps (see a simplified scheme of the flow-based closed-loop control within each water source of the legacy system in Figure 3).

The optimizing loop starts by reading from the historian’s database the most recent value recorded for the water flow distributed towards the population, which is fed into the optimizing algorithm described in [23,24], resulting in the optimized reference flows for each water source, which are being written alongside the necessary command tags values (for start/stop) on the OPC UA Server from the legacy systems. The consequence of writing new values on the OPC UA Server is that DWTP’s automation takes the new values into consideration; thus, the historian is able to influence the functioning of the DWTP in a non-invasive manner regarding the legacy system’s existing automation. The operations from the optimizing loop are repeated at 60 s intervals. The historian can run in two separate regimes: monitoring only (monitoring loop runs and optimizing loop does not run) or optimizing (both loops are running), the applying of the optimization over the monitored system being a decision left at the historian’s user choice, Figure 4 showing the settings area from the historian’s graphical user interface (GUI) which allows this choice to be made.

In advance of running the optimizing loop, the historian must analyze the recorded data and identify the water source’s priority indices (by running the algorithms described in [22,23]); this analysis operation is started manually by the user clicking a button in the ‘Analyzer’ tab from Figure 4. Because the quality of the water from sources changes over time, the historian’s user has the possibility of repeating this analysis periodically, when considered appropriate, so that the energy optimizing strategy is not considering some outdated priority indices.

From a more physical standpoint, the historian is installed on a Raspberry Pi 4 Model B hardware platform, which is located inside the DWTP’s command room, alongside the existing computers hosting the local SCADA software (redundant WinCC 7.2 servers) and OPC UA Server. The Raspberry Pi is connected to an uninterruptible power supply (UPS), and secure remote access to the historian’s GUI was implemented as well through SSH tunneling. The used Raspberry Pi has 4 GB of RAM, a Broadcom BCM2711, a quad-core Cortex-A72 (ARM v8) 64-bit processor, and a 128 GB MicroSD Card added for storage. The Raspberry Pi is connected to the local network through its Gigabit Ethernet port, and a display monitor is available exclusively for it, alongside a USB mouse and keyboard; the board also benefits from a specifically designed heatsink case to avoid possible overheating.

The acquisition of parameters is the first part of the proactive historian solution. Increasing the real-time anomaly detection and alarming for the legacy system is the basic non-invasive addition to the local solution. As a development direction, the authors are currently researching a predictive fault detection solution based on LSTM neural networks.

The deployed software solution represents not just a simple data-collecting tool but a step forward in the form of a process-aware historian that understands the meaning of the monitored OPC UA tags and also possesses proactive capabilities. The historian can autonomously analyze the stored data, compute an optimizing recipe and apply it to the monitored system in a fully automated and non-invasive manner. The optimizing strategy applied by the historian starts by analyzing the stored data, considering the water flows of water sources and the total energy consumption of the DWTP, the data dependencies identification algorithm detailed in [22] offering an output based on which a priority indicator regarding water quality can be computed for each water source. Afterward, a priority indicator regarding equipment wear down is computed for each water source based on the functioning hours recorded for each water source. Those two priority indicators are used to compute a global priority indicator for each water source, which is then considered in the optimizing algorithm for deciding the usage of the water sources. The algorithm attempts to match a specific total water flow that must enter the DWTP, which is computed based on the water flow distributed into the network from the station, and also keep the running water sources closest possible to a computed optimum flow for each. The target water flow that must be matched represents the sum of the flows of the water sources, the algorithm maximizing the usage of the highest priority water sources to the detriment of lower priority water sources, thus obtaining better energy efficiency and equipment wear-down balancing, depending on the weights of the two priority indexes when computing the global priority indicator for a water source. More details regarding the optimizing strategy and algorithm are available in [22,23].

Furthermore, the process of tailoring the historian solution developed in previous research supported an increase in the software application’s technological readiness level (TRL) at level 7.

The reason why the described approach was considered for the proposed work is that it represents a natural following step in the light of previous research, through which the developed historian solution meets the objectives and tasks described at the end of the previous section, thus reaching a maturity level that allows the direct benefits derived from 6 years of research to be transferred to a real-world industrial setting (DWTP). The approach represents the final step that allows research results to be applied and remain operational at an industrial partner, the transition being complete.

The historian software solution described in the current section was deployed at a DWTP located in Timiș, Romania, supplying a population of around 8000 people. The historian was operational for the current testing and validation purposes, in a monitoring-only regime since 30 August 2022, continuously gathering data from the DWTP until the current moment. Following previous testing experiments, the presented testing of the optimizing regime was realized during a 50 h long interval between 27 February 2023 at 13:30 local time and 1 March 2023 at 15:30 local time. During the aforementioned interval, the local operators did not make any adjustments to the functioning of the DWTP, which was left entirely under the control of the historian application with regard to the water source usage. The results of the test conducted in the optimizing regime are presented and discussed in the following section.

4. Results and Discussion

4.1. Energy Consumption Related Considerations

Being the primary objective of the applied optimizing strategy, the electricity consumption is presented at the beginning of the results section, considering four different points of view. All data referred to below was collected with the historian application installed in the DWTP, and the energy consumption mentioned is global per the entire DWTP.

The first analyzed aspect was the energy consumption per day, considering stored data from the three months preceding the test. The recorded energy indexes were used to compute the total energy consumption per respective month, which was then divided by the number of days in the respective month. Regarding the test period, the energy consumption was divided by 2083 (the result of dividing 50 h of test interval by 24 h in a day). The conclusions are summarized in

Table 1. Total energy consumption per day comparison during test with previous months.

	December 2022	January 2023	February 2023 1–27 February (Before Test)
Energy index start (kWh)	1,252,010.25	1,266,546.5	1,281,298.75
Energy index end (kWh)	1,266,546.25	1,281,298.625	1,293,673.25
Total energy consumed (kWh)	14,536	14,752.125	12,374.5
Energy per day (kWh)	468.90	475.875	475.942
Energy per day (kWh) during test	454.38
Comparison	−14.52 kWh/day −3.1%	−21.495 kWh/day −4.51%	−21.562 kWh/day −4.53%

below, and they highlight an energy consumption reduction during the test.

The second analysis was focused on comparing the total energy consumption during the test interval with other similar intervals. For this, the same interval inside a week (Monday 13:30–Wednesday 15:30) was considered from three of the four weeks preceding the test. Also, similar 50 h long intervals, but in Wednesday–Friday period, were considered from the week preceding the test and in the week of the test (immediately after the test finished). In all five intervals, the comparison of the energy consumption with the test denotes reductions using the optimizing strategy, as illustrated in

Table 2. Total energy consumption comparison during test with other similar 50-h long intervals.

	30 January 2023 13:30–1 February 2023 15:30 (Monday– Wednesday)	6 February 2023 13:30–8 February 2023 15:30 (Monday– Wednesday)	13 February 2023 13:30–15 February 2023 15:30 (Monday– Wednesday)	22 February 2023 13:30–24 February 2023 15:30 (Wednesday– Friday)	1 March 2023 15:30–3 March 2023 17:30 (Wednesday– Friday)
Energy index start (kWh)	1,280,569.5	1,283,990.5	1,287,399	1,291,612.875	1,294,833
Energy index end (kWh)	1,281,572.5	1,284,977	1,288,386.5	1,292,580.5	1,295,789.5
Total energy consumed (kWh)	1003	986.5	987.5	967.625	956.5
During test (27 February 2023 13:30–1 March 2023 15:30 Monday–Wednesday)
Energy index start (kWh)		1,293,886.25
Energy index end (kWh)		1,294,832.875
Total energy consumed (kWh)		946.625
Comparison	−56.375 kWh −5.95%	−39.875 kWh −4.21%	−40.875 kWh −4.31%	−21 kWh −2.22%	−9.875 kWh −1.03%

.

The third analyzed perspective takes into consideration the water volume that entered the DWTP to be treated, the chosen metric being the energy consumption relative to the water volume. In this case, two time intervals were considered for comparison with the test period: the entire week before the test period and a similar 50 h interval; the results of this analysis compose

Table 3. Total energy consumption per m³ entering DWTP comparison during test with previous intervals.

	20 February 2023 00:00:00– 27 February 2023 00:00:00 (the Week before Test)	22 February 2023 13:30– 24 February 2023 15:30
Total energy consumed (kWh)	3156.125	967.625
Total water volume entering DWTP (m3)	2813.8	816
Total energy consumed per water volume (kWh/m3)	1.121	1.186
During test (27 February 2023 13:30–1 March 2023 15:30)
Total energy consumed (kWh)	946.625
Total water volume entering DWTP (m3)	882.9
Total energy consumed per water volume (kWh/m3)	1.072
Comparison	−0.049 kWh/m3 −4.37%	−0.114 kWh/m3 −9.61%

, which shows good energy consumption optimizations during the test.

The final point of view in the data analysis regarding energy consumption also focused on water volumes, but this time, the total water volume that was offered by sources (the sum of the individual water volume that left each source towards DWTP during a considered interval). The comparison made between the test and the similar interval from the preceding week is detailed in

Table 4. Total energy consumption per m³ from water sources comparison during test with previous similar interval.

	Week before Test Interval 20 February 2023 13:30:00– 22 February 2023 15:30:00 Monday–Wednesday	During Test 27 February 2023 13:30– 1 March 2023 15:30 Monday–Wednesday
Total energy consumed (kWh)	914.75	946.625
Total water volume offered by water sources (m3)	881.25	936.25
Total energy consumed per water volume (kWh/m3)	1.038	1.011
Comparison		−0.027 kWh/m3 −2.60%

, and an improvement of energy consumption by using the historian application’s optimizing strategy was detected in this case, as well.

A graphical representation of the results presented in the previous four tables was added in Figure 5, in order to better visualize and summarize the conclusions of the energy-related comparisons.

As stated before, in all time intervals with which the test was compared, the DWTP’s water sources were operated manually by the local operators through the local SCADA software. The manual operation techniques that are being applied in this DWTP were refined during years of experience and were focused solely on energy consumption and water availability. As a consequence, the energy consumption was already very close to the minimum possible in the respective intervals, and being able to further improve that energy efficiency with the historian solution in this particular unfavorable test case represents a very good result under the given circumstances.

Although other recent research focused on the water industry from the energy optimization perspective exists as well, such as [25] where a water pumping station’s energy consumption is reduced by over 30%, [26] where a decrease from 5.09 to 3.8 kWh/m³ is obtained in the treatment of brackish groundwater or [27] where a 4.6% decrease in overall operating costs involving water desalination and distribution is reported, a solution which optimizes a DWTP’s energy consumption by adjusting the usage of its existing water sources could not be identified in order to facilitate a direct comparison of optimization results with the current paper.

4.2. Other Considerations

Besides the energy-related information from the previous subsection, there are other remarks derived from the same test using the historian application in optimizing the regime that can be brought into the discussion.

Firstly, no stability issues were identified regarding the historian software application, errors, or normal service interruption not occurring during the 50 h test period. Also, the interoperability with the monitored system did not raise any problems. On the other side, no stability issues were identified regarding the DWTP’s functioning. During the entire 50 h the DWTP operates under normal parameters considering water quality at the output and the water level in the distribution tank being well above any risk limit (including after the filters washing operation is finished). No human operator intervention was required during the test. This stability, concomitant with the energy-related results presented in the previous subsection, proves the increase of TRL level for the historian that was obtained during the current research.

Using the historian application for controlling the water sources usage instead of the manual source selections and activations by human operators-based approach presents major advantages. The water necessities can change either by varying demand from the population, local equipment failures, sudden bad quality of water causing clogged filters and increased washing cycles, or by broken pipes. This way, water can be wasted, or the reservoirs could empty. The case of a sudden change in the water flow distributed towards the population could result in a highly delayed water source flow increase command that could lead to an empty distribution tank in the DWTP and an interruption of water supply in the entire network. Using the described historian solution, the water source flows are adjusted appropriately after maximum 60 s after any sudden change, regardless of the time of its occurrence, thus, possibly avoiding a full interruption of the water supply.

Another observation is linked to equipment wear down, which is not currently taken into account with the current manual operation of the DWTP. Overusing a specific water source because it offers better water quality wears down the pump and, in the long term, could result in higher financial costs for the water company in the event of wear-down-induced failures. The optimizing strategy from the historian takes into consideration the functioning hours of the water sources as well when deciding the operation mode of the sources. However, the hypothesis of reducing the wear-down-induced equipment failures by using the historian compared to the current manual operation requires long time data (several years) of both usage cases to verify data that is not available at the moment.

5. Conclusions

The rapid pace of development that represents a prominent characteristic of Industry 4.0 is forcing various industrial actors to remain competitive by updating and adapting to new standards of industrial efficiency, performance, and flexibility. One important means through which this can be achieved is strong partnerships between academic and industrial spheres, oriented towards researching and developing solutions that possess a very significant practical applied character. This way, the perfect context is set for facilitating the transition of research results and improvements into real-world, daily-basis used systems, thus delivering research benefits to many people.

Constituting such research, the current paper presented the development and deployment of a historian software solution in an industrial environment represented by a DWTP. Besides the data collecting capabilities, proactive and process-aware intelligence of the solution was highlighted through a detailed case study, the results of which were presented and discussed from the energy efficiency and other points of view. The presented solution proved a high technological readiness level, very good stability, and interoperability with the associated technical system, posing valuable benefits for the future operation of the considered DWTP.

Also, considering the fact that approximately 50% of the electrical energy consumed by the DWTP is still produced from fossil fuels, through the energy efficiency increase provided by the usage of the historian solution, the current research indirectly contributes to a more sustainable operation of the DWTP by decreasing the associated carbon emissions.

The authors consider that the results of tailoring and applying the proactive historian in the current suboptimal scenario are highly notable. The targeted suboptimal perspective success rate was minimal from the energy-reducing perspective, respectively the increased interoperation with the legacy system running for around 10 years raised many issues. The proactive historian was able to increase energy efficiency in the worst possible context with almost no room for maneuver, considering the current energy-focusing strategy, reduced water source availability, and water demand. Also, the manual perspective was replaced with the automatic regime only through the proactive historian and its interoperation with the legacy system, raising the level of availability for the water sources, the reaction to varying water demands, and reducing human intervention.

Regarding future development directions, numerous possibilities are opened up by the presence of a local data-collecting solution that is process-aware and capable of understanding and analyzing the monitored data. Different objective functions can be chosen that would lead to researching other types of optimizations, such as reduction of substances (chlorine) consumption, for example. On a different note, large quantities of stored data can be used to investigate the feasibility of the application of Machine Learning or Artificial Intelligence techniques in the Predictive Maintenance area in an attempt to develop the ability to predict equipment failures. Furthermore, tailoring the historian solution to slightly different environments, such as wastewater treatment plants, brings new possibilities regarding development directions that have the potential to optimize the respective processes.