The CATALYST pilots are reaching their final validation points and our colleagues at Engineering Ingegneria Informatica spa, Marilena Lazzaro and Diego Arnone have written the following article to update us with the intermediate results of the validation at the Pont Saint Martin DC which we featured earlier this year, that article can be found here.
The Engineering Data Center in Pont Saint Martin in one of the four pilots involved in the CATALYST Project.
The CATALYST solution adopted in this pilot aims to assess the flexibility capability of the DC. The goal is twofold: firstly, to improve the DC energy efficiency and secondly, to explore the DC as technological hub able to sell flexibility services in the market and contribute to increased grid stability and Smart cities improvements.
The deployment of CATALYST software in the pilot represents a prerequisite to reach these goals. Indeed, the PSM DC hosts a tailored CATALYST solution responsible for:
- the monitoring of relevant measurements made available through an OPC (Open Platform Communications) Server;
- the computation of optimization plans performed by the DC Energy Flexibility Prediction and Optimisation tool;
- flexibility optimization decision making;
- the automated participation of the DC to the CATALYST Flexibility Marketplace through the Marketplace connector.
A graphical representation of the CATALYST solution deployed in the pilot is shown in the following picture.
The CATALYST Monitoring module interfaces with the OPC Server to gather data on power consumption of Refrigerating units and the IT power demand. These data are therefore stored into the CATALYST database. Indeed, an instantiation of the data model defined in the Project has been set up in the pilot according to the available DC physical components representing exploitable flexibility resources. Static and real measurements are therefore processed by the DC Energy Flexibility Prediction and Optimisation subsystem. Flexibility actions are identified for physical components representing potential flexible resources for the DC; the optimization plans are displayed on the DC Operator Console for validation since they need the approval for being executed. In CATALYST, the optimization plans can consist of several actions that suggest to the DC an efficient use of its own resources and propose to buy or sell energy (electricity or heat), electrical flexibility or IT workload by submitting proper bids and offers on the Multi-flavour Marketplace. This means that actions suggested to the DC Operator through the Console represent not only an opportunity for improving the DC energy efficiency but also enable a new revenue mechanism based on selling flexibility services to the Grid. An instance of CATALYST Flexibility Marketplace has been set up in the pilot for testing purposes.
This pilot can exploit two flexibility resources: the refrigerating units and the diesel generators. As to the first one, even if the DC is not equipped with a physical thermal storage, it is possible to exploit the thermal inertia of the cooling pipes working as “virtual” thermal storage. Trials have proved that the dynamic usage of the cooling system, switching off the refrigerating units for around 26 minutes, represents an overall availability of 181,2kW exploitable flexibility. Regarding the usage of diesel generators, the main idea is to take advantage of diesel generators maintenance for the provision of flexibility services to the Grid. However, diesel fuel emissions are an important aspect that cannot be overlooked. Performed test proved that the DC could reduce its own power adsorption from the grid to 0 kW for about 26 minutes or longer if necessary, although usually maintenance tests last around 30min. Usually, the DC power absorption is 1MW.
Why does CATALYST represent a benefit for the DC? Because the DC can create new revenue streams through flexibility services offering with no relevant impact on normal operation. In this regard, the CATALYST solution and, in particular, the Intra DC Energy Optimizer plays a key role: the optimization actions are computed considering the technical capabilities of the DC flexibility resources and the DC constraints. In other words, CATALYST suggests to the DC Operator the action related to the electrical flexibility and the most suitable timeslots for operating in such a way as to balance financial gain and service level agreement (SLA). On top of that, metrics are computed as well for assessing the suggested flexibility actions. They are computed by the Energy Efficiency Metrics Calculator using measurements stored in the CATALYST DB.
Focusing on the trials performed in the pilot, specific KPIs have been computed in relation to each flexibility resource and the flexibility action envisioned for the service provision.
In case of the dynamic usage of the cooling system, two Key Performance Indicators are computed: PUE (Power Usage Effectiveness) and DCA (Data Center Adapt).
PUE is a ratio for DC energy efficiency, it is defined by the formula:
Indeed, this KPI defines the % of energy spent powering ancillary equipment.
and are computed using the power consumption of refrigerating units and the power consumption of UPSs respectively.
Considering the overall test, a time period of 2 hours, 37 minutes and 28 seconds the following result is obtained:
PUE = 1,25
As it is expected, the PUE decreases during the test due to the switched off of refrigerating units for about 26 minutes; the IT Equipment Energy remains unchanged. Usually the PUE of PSM DC is around 1,57. This means that the DC can improve its energy efficiency exploiting the dynamic usage of cooling system and at the same time it could gain economic advantages selling the flexibility service to the market.
DCA metric, is about the ability of a DC to change its energy consumption behaviour.
DCA is computed using the formula below:
The is the DC energy real consumption after flexibility action’s implementation; the is the DC Energy baseline consumption profile.
A good value of DCA is closer to 0, this means that the DC can adapt its energy consumption to the planned power curve provided by the grid operator or defined by the DC itself. A DCA value equal to 1 means that the curve has not changed. Considering the possibility to turn off the refrigerating units for about 26 minutes, the following results are obtained:
The time slot is the same of PUE computation (2 hours, 37 minutes and 28 seconds).
The DCA is closer to 1 than 0 since the energy consumption of the generation units is quite small in comparison with the main incomer and therefore the baseline consumption of the DC slightly changes. This value could be improved by turning off refrigerating units for longer; this could be done improving the thermal inertia of the DC through the deployment of water storage tank. Indeed, the possibility of exploiting refrigerating units to provide flexibility depends on the thermal inertia of water circulating in the system.
When the electrical energy flexibility is provided by using diesel generators, another KPI needs to be computed: CO2 savings. It is defined by the formula:
Where is the total CO2 emissions due to the usage of diesel generators and is the total CO2 emissions that would have been released due to the energy consumed by the data centre during the “same” period but without the usage of diesel generators.
Considering the overall test, i.e. a time slot of 1 hour and 18 minutes during which diesel generators were switched on only for 26 minutes, the following result is obtained:
CO2 savings = -1,170961823
As it is expected, there is no savings, the CO2 emission due to the usage of diesel generators is very high and negative. But the CATALYST software aims at providing this kind of flexibility to the DSO only by adapting the maintenance plan of diesel generators. So, diesel generators are switched on the same number of times per year as per the maintenance plan. This means that the environmental footprint of the data centre is not affected.
In conclusion, an important consideration must be done about metrics. A DC expert can say: “all present DC KPIs are static parameters, related to long-term DC performance and DC utility set-up”. And this is true: for instance, DC PUEy is an annual average value on DC energy performance, based on 24/7 registration of PUE equation values. So, when we compute a KPI in the short time slot of a test case, what are we computing? What do those metrics mean?
CATALYST does not aim at changing cooling systems, server racks, UPS, types of servers, whose replacement usually affects the DC long-term performance and consequently change the DC KPIs. But the software installed by CATALYST is a relevant change for a DC because, even though the “static parameters” are the same, because the equipment is the same, CATALYST allows the DC to use its own resources in a different way. That is the reason why the project computes some KPIs for each experimental test case, even though the tests do not imply any change to the equipment.
All the tests performed so far are temporary experiments that cannot change the metrics that characterize the data center but we compute KPIs, applied to the timeslot of the test (from the time before disconnection to the time when the DC comes back to the state similar to the initial one) because their values give us some hints about how the KPIs can change depending not only on the flexibility service we are testing but also on the frequency this flexibility service is provided.
As an example, if the CATALYST framework, once installed in the DC, decides that the DC must work in islanded mode for two hours every day, we are sure that metrics like CO2 Savings will change significantly. And we have applied no changes to the hardware, but we have applied a relevant change to the software that manages the DC. We must obviously keep in mind that, if we compute a KPI in a short timeslot (like the two hours of an experiment), we cannot say that the KPI of DC is the new value that we have computed. We can simply say that that value is an indication of how that KPI can change depending on the frequency the flexibility action is applied.
The results obtained so far in Pont Saint Martin are only part of the expected results of the project: other three pilots are validating the CATALYST technological solution by testing functionalities that cannot be applied to a colocation data center like the one in Pont Saint Martin, such us the IT Workload Validation. So, it’s important to stay tuned…