Development of a reservoir tool model

Thanks to tanks that can act as buffers, the pumping of water can be postponed or advanced to avoid electricity consumption during network shortages, or conversely, to consume more when electricity is abundant on the network.

The day-ahead price of electricity serves as a proxy for network load. High-price hours indicate times when electricity is scarce on the network, while low prices indicate times when electricity is abundant. So it is considered to be a good indicator for the way the water network should be conducted to help maintain network balanced.

Since tanks typically allow for a few hours of water storage and the day-ahead prices are known a day before consumption, it is convenient to create pumping schedules based on these prices to relieve network strain.

Establishing the best pumping schedule is equivalent to solving an optimization problem in which the total cost of pumping water for 24 hours of production is minimized.

We have developed a tool for this purpose, implemented in Python and utilizing the Pyomo module. The tool facilitates the connection of different components of the water network, such as pumps, tanks, pipes (connectors), collectors, etc.

Below is a typical representation of a segment of a water network modelled using the tool.

Flexibility potential related to production

A first estimation of the displaceable electricity consumption was performed using simple data from the network (essentially, power and hours of operation of the pumps). This allowed for an initial overview of the displacable loads (see image below).

We used our tool on the site that offer the highest flexibility potential.

Here below we illustrate results obtained on the Tailfer-Callois branch shown above

• First Graph: This shows the power consumption of the "nourricières" pumps, which draw water from the Meuse to supply treatment lines. Each color represents a different pump's operation, while electricity prices are also displayed. The pumping schedule reacts to electricity price variations.
• Second Graph: This depicts the power consumption of the main pumps, which must lift water over 150 meters. The optimization's impact is more pronounced here due to the buffering effect of reservoirs. The strategy alternates between full-power operation and complete shutdown.
• Third Graph: It consolidates the total power absorbed by all pumps.
• Fourth Graph: This illustrates the water volumes stored in reservoirs, highlighting how optimization leverages the large capacity of the Callois reservoir. Notably, water pumping is reduced on Thursday when electricity prices are high and maximized on Sunday when prices drop, taking advantage of available storage space.
• Fifth Graph: This represents flow rates through treatment lines and as regulated by the valve.

The optimization’s all-or-nothing approach limits potential energy recovery at the valve. Since energy recovery is optimal at medium flow rates, future optimization could consider this aspect, although it presents a more complex challenge due to the nonlinear relationship between recovered energy and flow rate.

The results can be analysed based on two key metrics: the movable electrical power and the duration for which this power can be shifted. To achieve this, results from several characteristic weeks - considering both the evolution of day-ahead electricity prices and production/consumption patterns - are aggregated for analysis. This type of aggregation is illustrated in the following graph.

In this example, the evolution of movable power is observed as a function of the shifting duration. This is a swarm plot displaying the results of all optimized weeks, with colors representing different seasons. Boxplots are also superimposed. The red points represent the average values, while the points with black edges indicate outliers.

As expected, the movable power decreases as the shifting duration increases. However, in this case, we observe that the maximum movable power remains available for at least three hours, which is relatively significant.

Other sites within Vivaqua’s network, as well as sites from other producers, were analysed. A distinctive characteristic of the Vivaqua network is that maximum power shifting durations last several hours, whereas for other producers, these durations are generally shorter.

In all cases, we observed that the magnitude of movable power is of the same order as the nominal production power. By nominal production power, we mean the annual average power, calculated by dividing the total electricity consumption over a year by the total number of hours in a year. This value is, of course, lower than the installed power at the sites.

An extrapolation of the results at the Belgian level provides a rough estimate of approximately 50 MW of cumulative movable power across all Belgian producers. However, this is only an approximation, heavily dependent on:

• The assumption made on the amount of electrical energy required for water production, transportation, and distribution.
• Specific conditions encountered by different producers, particularly in terms of available storage time (evaluated by dividing the average flow rate through a reservoir by its capacity).
• Technical feasibility aspects, including controllable pumps, supply security, etc…

Other flexibility potential (pumping/turbining and turbining for energy recovery)

Pumping/turbining

The potential to utilize segments of the network similar to a Pumped Storage Power Station (PSPS) has been explored. Specifically, this potential was assessed at the Tailfer/Bois-de-Villers site, which forms the backbone of the Vivaqua network's production and storage.

To address the question of whether the water network can provide flexibility for the electrical network (TSO/DSO), the Tailfer-Bois-de-Villers site is the most promising within the entire Vivaqua network. Topologically, there is a 152-meter elevation difference between the Tailfer production site and the Bois de Villers storage site.

The Tailfer site consists of 2 reservoirs with a total capacity of 29,982 m³, while the Bois-de-Villers site consists of 2 reservoirs with a total capacity of 48,000 m³.

The potential flexibility that the Vivaqua network could provide to the electrical network lies in the idea of operating the Tailfer-Bois-de-Villers segment in reverse. This means that water would not be pumped from Tailfer to Bois-de-Villers, but rather turbined from Bois-de-Villers to Tailfer.

We have studied the evolution of water levels in the reservoirs of both sites over 3 years.

This study shows that there is a daily transfer of a certain volume of water from Tailfer to Bois-de-Villers. We measured that a real capacity of 20,000 m³ could be returned from Bois-de-Villers to Tailfer for turbining. Vivaqua has indicated that, for hydraulic reasons, only 10,000 m³ could be considered for turbining.
If in the Tailfer pumping station, one of the six operational pumps is replaced by 1 horizontal Francis turbine, operating at a gross head of 152 m for a net head of 145.6 m, and at the nominal flow rate of 1.389 m³/s, then a nominal power of 1823 kW could be developed, equivalent to 1.823 MW. Turbining 10,000 m³ at the nominal flow rate of 1.389 m³/s takes 7200 seconds, or 2 hours.

Therefore, the site could provide 1.823 MW for 2 hours to the electrical network.

The potential exists and would perfectly suit the following operation: turbining the waters from Bois-de-Villers to Tailfer for 2 hours during the peak hours of the electrical network, either in the morning between 6 am and 9 am, or in the evening between 7 pm and 10 pm. This represents a potential of 2 times 2 hours daily.

As mentioned, Vivaqua manages its pumping and water volume transfer between Tailfer and Bois-de-Villers on a daily basis. Therefore, the operation of turbining the waters from Bois-de-Villers to Tailfer for 2*2 hours (morning & evening) must be planned.

Statistically, we know that there are daily peak hourly consumption periods on the electrical network. And with precise planning, it would be possible for Vivaqua to fulfill its water supply mission and contribute to alleviating the load on the electrical network.

Energy recovery

The difference in altitude between the various tanks in the network provides a fixed static head difference. This variation in static head value, combined with the characteristics of the pipes (primarily diameter and length), allows for a nominal maximum flow rate. During operation, the supply and distribution segments of the network typically operate at a fraction of the nominal flow rates. Therefore, to achieve partial flow rates, some pressure must be dissipated through valves, resulting in energy dissipation in the form of heat. Utilizing turbines instead of valves would enable the conversion of the pressure drop into mechanical energy, and subsequently, electricity through the use of an alternator.

In the Vivaqua network, there are locations with valves that handle significant flow and/or pressure drops, where electricity could be generated. Additionally, there are locations within the distribution network where electricity generation is feasible but with lower power output.

The larger potential is at the Mazy station, located as an extension of the Bois-de-Villers station. It serves as a node between 2 network branches, one going to Plancenoit/Callois and the other to Daussoulx/Emine.

In this station, we can identify 3 valves at the ends of which a pressure drop occurs. By using the classical rules of hydro-electricity and choosing a turbine efficiency of 0.7 (a somewhat optimistic efficiency), we obtain the following productive forecasts.

Out of the 3 valves, V2 and V7 serve the Plancenoit/Callois section, and V25 serves the Daussoulx/Emine section.

V2 (red) and V7 (orange) have a potential energy recovery of 151 kW and 122 kW respectively. It should be noted that the readings are consistent over the 3 years of data collection.

Vivaqua has 2 operating modes concerning the Daussoulx/Emine section. This is why at the ends of valve V25 (blue) we observe either a potential of 450 kW (before December 2020) or 50 kW (after December 2020). Vivaqua explained that the classic operating mode corresponds to that after December 2020.

That being said, if we base ourselves on these 3 valves, there is energy that can be harnessed 24/7, amounting to 320 kW. It is important to note that the recoverable powers fluctuate around their average values, oscillating daily and in the form of a "square wave"

Another interesting site to consider is Plancenoit site where an average power of 21.8 kW could be harnessed