Keywords

1 Introduction

Located at the heart of the European waterway network, the Walloon Region from Belgium manages about 450 km navigable waters and more than 80 hydraulic structures including 6 weir reservoirs and has to deal with the increasing complexity of waterways and its current issues. Climate change increases the probability of both floods and extended dry spells, while the growing traffic adds additional stress to the system.

Last year in the Walloon Region, river transport of goods represented 7% of the total freight transport, after shipping by road (84%) and by train (9%). Considering that truck transportation is responsible for 99% of land shipping CO2 emission, Wallonia aims to transfer good transports from trucks to shipsFootnote 1. For this purpose, the region wants to make waterways more attractive by modernizing the engineering structure of the waterways network and by improving its global management.

Alongside the current modernization and enlargement of the crossing structures, river dams allow to regulate water level for navigation. Nowadays, all those dams are controlled by local PLC - Programmable Logic Controller - based on a regulation loop. When the automatic mode is activated, those PLC can move the weir gates based on the upstream water level measurement and a specified water level setpoint. Given climate change and the complexity of the network, this local control does not prevent waves to propagate from reach to reach and is no longer considered the most optimal solution for a global management.

In the next few years, hydraulic structures of the Walloon Region will be remotely controlled from one and unique place, the recently inaugurated PEREX center in Namur. Structures will be connected to the center via optical fiber and a SCADA - Supervisory Control and Data Acquisition - system. The entire network could be visualized and controlled from PEREX. The project is currently underway. To have a global vision and an effective control of the network, Walloon Region wants to have a real-time decision support tool in order to optimize water management and to allow a harmonized, global and digital management of the waterways. The tool, working in close loop and supervised by an operator, will select the optimal action to be carried out on waterways hydraulic structures.

To develop this optimization tool - called Orhyx (Optimisation de la Régulation HYdraulique depuis pereX - Optimization of hydraulic control from perex) -, Walloon Region decided to get help from 2 Belgian compagnies with expertise in this field: International Marine and Dredging Consultants (IMDC) and Sumaqua. The different parties have therefore joined their efforts to create this tool which is currently under construction. The project started in 2019 and should be completed in 2028.

Orhyx is the start of a new intelligent and verry innovative system that optimizes the management of an entire waterway network in order to enhance security and navigability, improve environmental safety and optimize energy production and consumption.

A preliminary version was developed on two waterways of the Walloon network (Basse-Sambre and Canal Charleroi-Buxelles). Tests are planned on those two waterways on an operational point of view in order to test the system in real conditions. Next steps will consist in upgrading Orhyx to cover the whole network.

This paper aims to describe the tool and its characteristics as well as the optimization module behind it. The first available results will be presented as well as the improvement perspective for the rest of the project.

2 Description of the System

2.1 Information Flow

The optimization tool needs information as inputs to operate and produces results as outputs. Orhyx is therefore integrated in a complex flow of information. Figure 1 describes how the information loops through the global system and how the tool is integrated into the waterways manager physical and IT infrastructure.

Data coming from field measurement instruments (water levels, discharges, states of the equipments, regulation modes,) are transmitted in real-time to the back end of the tool where the conceptual model and the optimization algorithm are deployed. This calculation is triggered at regular intervals (30 min). The team of operators make sure the system is always up to date with latest known event occurring on the network (maintenance, heavy rainfall prediction) for the next 24 h. If approved by the operators, the new control strategy (setpoints for the equipments) is then sent to the network. Effects of these new states closes the information loop as they are recorded by the measurement instruments.

Fig. 1.
figure 1

Information flow throughout the physical and IT network.

Full size image
Fig. 2.
figure 2

Graphic dashboard in the Orhyx interface.

Full size image

2.2 Description of Orhyx

Orhyx is built as a modular web application (vue.js, django, influx and postgres databases) aimed to facilitate the management of such a complex system. The interface offers many functionalities:

Waterways Network Characteristics:

all the relevant static information regarding infrastructures and equipment as well as waterway network (dams and reaches dimensions, pumps and turbines unitary flow,…) are grouped by operational site and river stretch. The operators can visualize them via a GIS interface or via a generic summary and export them in PDF.

Timeseries Exploration:

for each river/channel of the network, the whole history (inputs and outputs of the scenarios) as well as the most recent data are displayed and compared on a flexible and customable dashboard. For instance, Fig. 2 shows water levels measurements for different sites on the graph above and regulation modes by site on the graph below (manual or automatic). Gates positions or water levels calculated as output by the tool can also be added on those graphs in order to visualize them.

Scenarios - System Configuration:

some events are not automatically caught by the system and must be manually implemented through the scenario editor. This component displays the input used by the optimization algorithm and are editable. It allows the operator to adapt the system according to reality on site (e.g. maintenance of gate) or to introduce a future event by modifying the relevant input and parameters. Once the configuration is adapted, all subsequent automatic simulations (running every 30 min) will take these changes into account.

Sensitivity Analysis and Remote Control:

the scenario editor can also be used to study the influence parameters and run calculation beside the operational automatic simulations, or to simply steer the equipment remotely (change a gate level).

These components combined offer a wide and detailed view of the network and are meant to evolve in the coming years.

2.3 Optimization Module

The optimization module consists of a tailor-made reduced genetic algorithm that allows to optimize the future management of the system. The general structure of the optimization module is illustrated in Fig. 3.

Fig. 3.
figure 3

Illustration the general structure of the optimization module.

Full size image

The forecasting component combines the hydrological forecasts from the HYDROMAX model (Moens et al. 2018) with a surrogate hydraulic model named “SCAN” to forecast the future system states up to 24 h in advance (Wolfs et al. 2015). SCAN is a very fast simulating model due to the mix of data-driven and physical based modelling components. The model is calibrated against a detailed 1D hydrodynamic HEC-RAS model. It can simulate 24 h in 0.004 s for a waterway network of 88 km. The integrated data assimilation component ensures that the forecasts align with the latest observations and is an improved version of the assimilation procedure described in Vermuyten et al. (2018a).

The optimization module contains a scenario generator that generates a large set of potential control strategies for all controllable structures in the network within the physical realistic boundaries of the network and the structures. The scenario generator uses a tailor-made parallel reduced genetic algorithm that is built upon earlier work by Vermuyten et al. (2018b). New control strategies can be generated as the result of (a) mutation of an existing strategy, (b) a new randomized control strategy or (c) cross-over between parallel optimized control strategies. Mutations refine the control strategy and pushes the strategy towards the optimal solution. The randomized strategies and the cross-overs ensure that the scenario generator explores the entire solution space and does not get stuck in local optima.

The scenario selector evaluates the different control strategies and chooses the most optimal solution based on a transparent set of priorities and objective functions. Priorities group similar objective functions and allow to clearly prioritize certain objectives over others. Table 1 shows the different priorities that are implemented in Orhyx. The optimization works sequentially: it will first minimize the objective functions linked to the highest priority before moving towards lower priority objectives. Each priority contains a set of objective functions that mathematically describe the objectives of the priority. Soft constraints are used within the objective function to penalize the exceedance of certain thresholds. Figure 4 gives an example of the soft constraints linked to navigation and the resulting objective function. The soft constraints provide the upper and lower level in between water levels should remain to ensure navigation. The resulting objective function increases quadratically once the water level exceeds these boundaries, restraining the optimization to a solution that remains well within these thresholds. Next to the soft constraints, the objective function also contains weights that allow to flexibly shift the importance within a priority depending on the situation.

Table 1. List of priorities and objectives implemented in the optimization module.
Full size table
Fig. 4.
figure 4

Illustration of the soft constraints for navigation on a certain location (left) and the resulting objecitve function penalizing the exceedance of these thresholds (right).

Full size image

Orhyx updates the optimal control strategy every 30 min based on the latest observations and forecasts. For each optimization 100.000 potential control strategies are simulated and evaluated. In the current setup 19 variables are simultaneously being controlled by the optimization module. The objective functions incorporate the model results of 87 variables along the study area.

3 Validation

The optimization module was evaluated on a broad range of historical events including extended dry spells in summer, high flow periods in winter, short high-intensity thunderstorms in summer and events with observed waves propagation through the system.

The optimization module produced a global control strategy for all structures for each of these events. These results were compared with a local control strategy where each of the structures respond individually to the disturbance. These local control strategies are used operationally and are based on years of experience with controlling the system manually. The results are evaluated for different indicators that summarize the results over the entire network. Figure 5 compares the navigability of the local current control strategy to the global control strategy (Orhyx tool). The navigability indicator gives the percentage of time where navigation on the channel can occur without problems. A value lower than 100% indicates a potential risk for navigation somewhere in the network for a certain period of time due to the undershoot of certain water level thresholds. Figure 5 shows that the global control strategy results in higher percentages of navigability for almost all events. Only for the first high flow event the percentage of navigability is lower in the global control strategy. This is linked to the fact that floods occur during this period and the global control strategy tries to avoid these floods at all costs given the general security has a higher priority than navigation (Table 1).

Figure 5 shows the comparison for the indicator linked to the variation of the water level. This indicator quantifies the short-term water level variations which could be linked to waves propagating through the network. The optimization module tries to minimize the water level fluctuations in the fourth priority. For most events the global control strategy results in lower variations of the water level. Only for the high flow events the global control strategy allows higher variations of the water levels to reduce flooding, which is a higher priority in the optimization.

Fig. 5.
figure 5

Resulting time in which navigability is ensured (higher is better) and variations in water levels (lower is better) for different storms for the local and global control.

Full size image

The results of the validation demonstrate the benefits of using a global control strategy compared to the existing local control strategy. This benefit is related to the increased temporal and spatial view of the global control strategy. The forecasting module allows the optimization to look into the future and anticipate on future disturbances with pro-active measures. Water levels in a reservoir (section of the river defined as a storage cell in the SCAN model) can e.g. be lowered to increase the available storage volume for future storm events thereby potentially reducing floods. The global control strategy also benefits from evaluating the entire system as a whole, allowing to take real time and reactive measures to e.g. reduce short term water level fluctuations.

4 Conclusions

Starting from scratch and developing an innovative tool like Orhyx is very challenging and rewarding. The framework of Orhyx and its operating principle have been studied and defined on a reduced area of the Walloon waterways network. Eighty-eight km of Walloon waterways have been therefore modeled and 14 hydraulic structures optimized with 48 optimization runs every day. The first off-line validation of Orhyx shows encouraging results and outlines the benefits of using a global, centralized and optimized management compared to the current local management. Next steps will consist in testing the outputs of the tool by implementing them in real conditions on site. For this purpose, water level setpoints and gates positions instructions will be sent to the sites. After this, Orhyx will be enlarge to the whole Walloon Region network. Challenges that arise in the next months will probably be the operational settlement and the interaction and communication with the hydraulic structure thanks to the SCADA system, also under construction.