Digitalization of Existing Process Plants

We bring process plants into the digital world: We combine plant engineering and process expertise with data infrastructure, models and forecasts. In this way, we gradually make existing plants usable for data-driven optimization and, for new plants, create the conditions for future digital applications as early as the planning stage.

Digitalization – more than automation

Many existing plants operate reliably – yet their data are often used only for immediate control purposes. Structured storage, a consistent time reference and the necessary plant context are frequently missing for forecasts, process models or data-driven optimization.

This is where true digitalization begins: measured values are linked with information on operating states, setpoints and limits, faults, and manual or automatic operation, and are made available over extended periods. Only then can correlations be identified, models applied reliably and concrete recommendations for plant operation derived from data. This gives operators the opportunity to continue using existing automation structures while gradually expanding their plants with new digital functions –- from operating analyses and soft sensors to forecasts and energy- and resource-efficient operational optimization.

Our solution: from plant operation to data-driven optimization

Fraunhofer IGB develops technical solutions and tools that enable process and operating data from existing plants to be captured, stored and monitored in a structured way and made available for digital applications. On this basis, we also develop and integrate suitable process, forecasting and optimization models for the respective application. In doing so, we combine plant and process expertise with data infrastructure and modeling and consider the entire chain from the real plant to the digital application.

We bring existing plants into the digital world

Existing plants can thus be digitally enhanced without completely replacing the existing automation. The existing PLC continues to perform its control and regulation tasks and is supplemented by a local digital layer close to the plant.

This approach enables step-by-step digitalization and can be adapted to different technical starting situations. It is particularly suitable for plants that have evolved over many years, heterogeneous automation structures and systems whose data have not yet been captured and stored in a standardized manner.

The principle: making existing plants AI-ready step by step

The added value does not come from AI as a standalone component, but from the controlled path from the real plant to robust digital operating logic. The edge device becomes the plant node between the PLC, local data storage, the model platform and the digital application.

1.
Capture real operating data

Flow rate, pressure, pump runtime, storage level, feeding, faults and process states are automatically captured from the plant and its operation.

2.
Combine forecasts and models

Electricity prices, weather data, plant status, gas storage and domain-specific logic are evaluated together.

3.
Generate recommendations

Operators receive transparent operating schedules, target-actual comparisons and and optimization recommendations.

The edge device connects the plant and the digital application

Digitalization with an edge device

A central component is an edge device that acts as the plant node between the PLC, local data storage, higher-level data platforms and digital applications. The edge device captures process and operating data directly at the plant, structures and stores them close to the process, and makes them available for visualizations, analyses, forecasts and models. Additional information such as electricity prices, weather forecasts, laboratory analyses or other external data sources can be linked with the plant data.

This enables models and optimization methods to work on a common, up-to-date data basis; results, recommendations or setpoints can be fed back into plant operation via defined and controlled interfaces. This creates an end-to-end data and optimization chain -– from the real plant to the operating decision.

Interface between PLC and edge device: lean and transparent

The process variables required for an application are clearly described during PLC engineering and brought together in a defined data block. For existing plants, this structure can be added selectively without changing the existing control logic. In addition to the variable name, information such as unit, acquisition interval, trigger conditions and other metadata is stored directly with the PLC variables. This information is then used to generate the configuration for the edge device.

No additional OPC UA infrastructure or proprietary intermediary system is required for data transmission. The connection is made directly between the PLC and the edge device. This keeps the interface lean, transparent and independent of additional licensing or platform components. Information on the process variables is not maintained multiple times in different systems, but is taken from one centrally defined source. This reduces configuration effort, prevents transmission errors and makes it easier to expand the plant later or integrate additional measuring points.

Ensuring the functionality of the data chain

It is not enough simply to transmit process data. For the reliable use of forecasts, models and AI applications, the entire data chain must operate completely, up to date and with consistent timing.

For this reason, the data flow from the PLC via the edge device through to storage is continuously monitored. Continuous time synchronization between the edge device and the PLC ensures that process data can be assigned unambiguously and correctly in time across systems. This creates a reliable data basis for downstream models, forecasts and optimization methods.

Monitoring the edge device

In addition to the process data, the technical condition of the edge device is also monitored. This includes, for example, storage utilization, processor temperature, accessibility and the status of running services and databases.

Automated warning messages can be triggered in the event of critical deviations. This is intended to identify errors in data provision before they remain undetected for extended periods and lead to incomplete data sets, incorrect model results or unsuitable recommendations for action. Monitoring the hardware, software and data chain together therefore forms the technical basis for reliable digital plant operation.

Edge device as a platform for forecasting models

The edge device is used not only for data acquisition and storage, but can also serve as a local platform for forecasting, process and optimization models. This gives models direct access to current and historical process data as well as the contextual information required for their evaluation. Depending on the application, models can be executed locally on the edge device or external model and forecasting services can be connected.

A key aspect is the distinction between data quality and model quality: technically functioning data transmission does not automatically mean that a model delivers sufficiently accurate results. For this reason, forecast and model results are evaluated not only in terms of their accuracy, but also in terms of their impact on actual plant operation.

Biogas plant reference: evaluating electricity price forecasts, gas storage and feeding together

Electricity price forecast

The forecast provides price windows for the coming days and forms the basis for flexible operation.

Gas storage management

Storage level, methane production and CHP demand are combined on an hourly basis to produce a plausible operating schedule.

Feeding and substrate mix

Availability, methane yield and costs are considered together with the biological stability of the digester. An ADM1-based process model helps to identify critical substrate combinations and overload conditions at an early stage and to safeguard the feeding strategy accordingly.

CHP operating schedule

The system shows when full-load operation, part-load operation, gas storage or bringing generation forward is economically and operationally sensible.

Use case: Electricity-price-driven operation of biogas plants with CHP units

Using BioCash as an example, we show how electricity price forecasts can be combined with plant-specific operating data and process engineering constraints. Factors considered include expected biogas production, the status of the gas storage system, substrate management, and technical and biological operating limits. On this basis, a CHP operating schedule is calculated that enables economically optimized plant operation while remaining within process constraints. The most economically suitable operating mode does not depend on the electricity price alone.

The interaction of the following factors is crucial:

forecast electricity prices
available and expected biogas production
status and capacity of the gas storage system
output and operating limits of the CHP unit
substrate availability, substrate costs and methane yield
substrate composition and feeding strategy
biological stability of the digester and potential process overload conditions, assessed using an ADM1-based process model
additional operational and process engineering constraints

Another focus is the evaluation of the electricity price forecast used. BioCash updates the forecast regularly and stores the respective forecast snapshots for the coming seven days. Once the published electricity exchange prices are available for the corresponding period, they can be compared with the previously stored forecast values. This makes it possible to investigate how accurately the price level, high- and low-price periods and their timing were predicted. At the same time, it is possible to evaluate how forecast quality changes as the forecast horizon increases. The resulting historical forecast and comparison data can be used for validation and, in the future, also for the further development of forecasting models.

However, it is not only the accuracy of the electricity price forecast that matters. It is also important to determine whether and how forecast deviations affect the CHP operating schedule calculated from it. For this reason, we investigate whether, for example, CHP start times and operating durations, use of the gas storage system or the achievable contribution margin change. Not every deviation between the forecast and the published electricity exchange price automatically leads to a different operating decision. If, for example, a high-price period is misjudged or forecast at the wrong time, the optimum operating time of the CHP unit may shift and the available biogas may be used less economically. The electricity price forecast can therefore be evaluated on two levels: How accurate is the forecast itself, and what actual influence do forecast errors have on plant operation and the economic result?

Tabbed contents

Expand all Close all

Unique features and advantages

Our approach combines the existing automation system with a local, modular data and model platform. This allows existing plants to be digitally enhanced step by step without having to completely replace functioning control structures.

Unique feature: edge device as the central plant node

The edge device forms the interface between plant control, data storage, models and digital applications. Existing process data can therefore be used for new applications without major intervention in the existing PLC structure. Once established, the infrastructure can be reused for different tasks – for example visualization, operating analysis, forecasts, soft sensors or AI-supported optimization methods. New applications can be added step by step.

Local data under customer control

The process data are captured directly at the plant, stored on the edge device and can be processed there. This keeps data storage and key processing steps under the control of the plant operator. Only the information required for external applications or cross-site evaluations needs to leave the plant. At the same time, models and evaluations can be operated directly at the plant site. This reduces dependence on a permanent cloud connection and enables local operation even when external communication is limited.

Open source

For the operating system, data storage and other technical components, we rely on open and established technologies. This reduces dependence on individual manufacturers, proprietary platforms and long-term licensing models. The modular architecture allows interfaces, databases, models and applications to be extended or replaced as required without having to redevelop the entire solution. This enables the infrastructure to grow with the requirements of the plant and to continue to be used for future applications.

Advantages of the plant node

Data are stored locally
Models run directly at the process
Controlled feedback to the PLC

Benefits for projects and the market

Reliable operating data
Optimization based on real plant data
Cross-site comparability

Collaboration

We make your existing plants AI-ready!

We support companies and plant operators in gradually opening up existing plants to data-driven applications, forecasting models and AI-supported optimization. Collaboration can begin with a specific existing plant or can already include the development and testing of new digital functions.

Typical questions include, for example:

What data are already available in my plant?
What additional information is required for the desired application?
How can existing PLC, sensor and control systems continue to be used?
How must data be structured and stored so that they can be used for models and AI applications?
Which models, soft sensors or forecasts can provide concrete added value for plant operation?
How can model results be safely integrated into existing operating procedures?
How can a developed solution later be transferred to additional plants or sites?

Contact us!

In addition to direct development and industrial projects, joint research projects are also possible, in which new data-driven methods are developed on pilot and demonstration plants and tested under real operating conditions. We would be pleased to explain the possibilities and our approach in an initial, no-obligation discussion. We look forward to your call or email.

We structure your data

With clean data structuring, we create the basis for the use of any AI tools, from modeling through to data- and forecast-based control. The starting point is the plant P&ID (piping and instrumentation diagram) with all components, from sensors, pumps and valves through to heating units. This creates a structured and transparent data basis. Changes to the P&ID or expansions of the plant can be reflected in the data structure. In this way, the system remains usable even when plants are retrofitted, expanded or technically modified.

We make your data AI-ready

A key focus is integrated, process-oriented data management. To this end, we analyze your measurement data and ensure that further relevant data are also collected, integrated into the structure and able to communicate with one another. For each element, we capture and store not only measured values but also contextual and status information, such as operating modes as well as setpoints and limits. Linking a wide range of data and protocol formats, automated transfer and safeguarding plant data (including offline data such as analytical data) create the basis for smart process control and automation, as well as efficient data processing (operational analysis, modeling, etc.).

We integrate the edge device as a model platform

Based on the data structure developed, we integrate the edge device into the existing plant and automation environment. It handles local data acquisition, storage and provision of process data and at the same time serves as a platform for visualizations, soft sensors, forecasting, process and optimization models. Depending on the use case, models are executed directly at the plant site or connected via defined interfaces. We also take into account monitoring of the data chain and the technical infrastructure. This creates a usable platform on which individual digital applications can initially be implemented with additional functions to be added later. If necessary, a controlled feedback loop to the PLC can also be set up, through which verified recommendations or setpoints can be gradually incorporated into operational management.

Overview of services

We support you from the analysis of your existing plant and the development of the data infrastructure through to the development and testing of data-driven applications and optimization methods.

Our services include:

Analysis of the existing plant and automation structure
Capture and assessment of existing data sources, interfaces and communication paths
Structuring plant data on the basis of the P&ID, including relevant metadata such as P&ID/tag number, unit, timestamp, storage interval, event conditions and quality information
Development of a plant-specific data structure for online and offline data
Integration and configuration of the edge device in the existing automation environment
Connection of PLCs, sensors, databases and other data sources
Local storage of process, operating and status data, including their time histories
Integration of offline data, for example laboratory and analytical data
Monitoring of data quality and the entire data chain
Visualization and remote monitoring of plant, process and status data
Development and integration of soft sensors, process models and forecasting models
Integration and training of AI models for plant-specific questions
Validation and evaluation of forecasting and model results
Development and testing of optimization strategies for plant operation
Derivation of recommendations for action, operating schedules and optimized setpoints

Applications and potential uses

Our approach is suitable for automated process plants in which existing process data are to be used for more efficient, predictive and more strongly data-driven operation. This is particularly relevant for existing plants where the current automation system is to remain in use and be gradually supplemented with digital applications.

Particular potential arises when external information and forecasts are taken into account in addition to current plant data. Examples include electricity prices, weather and precipitation forecasts, heavy rainfall events, or changing influent and load conditions. This allows a plant not only to respond to its current condition, but also to incorporate future developments into its operating strategy.