Hasso-Plattner-Institut
  
Hasso-Plattner-Institut
Prof. Dr. h.c. Hasso Plattner
  
 

Energy Management Project Details

More than five years of research in in-memory technology at the chair of Prof. Dr. Hasso Plattner at the Hasso Plattner Institute created the foundation for the real-time in-memory project. Thanks to our cooperation with the SAP Innovation Center Potsdam, we were able to interview business partners to obtain a wide understanding of enterprise technology and requirements of real-world applications. We received considerable support in combining research methods and business knowledge to solve our project goals by our supervisors Matthieu Schapranow and Christian Schwarz. The main goal of the project was to be able to save and analyse data of 100 Million Smart Meters using a column-based in-memory database. From our point of view this number covers the expected number of smart meters in Germany [2]. All analyses are done on the raw data without additional aggregates.

Database and Hardware

We incorporated an in-memory database, which creates the basis of our research prototype. In contrast to traditional databases, our database holds the entire set of operational data in-memory. By eliminating the need for hard disk drives for storing data on disk, accessing data in-memory becomes about 50,000 times faster than from disk [1]. This technology allows the processing of 100 Million Smart Meter Readings we expect to be inserting into the database every 15 minutes without unacceptable hardware effort. We used a commodity blade server. This server had four 8-Core CPUs and 256GB of DDR3 1066MHz main memory installed.

Smart Meters

Each energy customer in a Smart Grid will have at least one Smart Meter. This will be read periodically to obtain the energy consumption of the last reading interval. At the moment we assume that these intervals are 15 minutes. This is a commonly used standard although there is no absolute standardization of the smart metering technology [3]. A data stream of 100 Million reading values per 15 minutes comes up this way.

Rates

Electricity rates in the future Smart Grid will consist of timeslots a.k.a. time of use buckets [4]. These timeframes describe when a specific price per kWh is valid. They can be set in any granularity so a rate can have for each 15-minute period a different timeslot. We implemented a fully flexible rate, which has 35040 timeslots and can have different prices in every quarter of an hour per year. This allows to react directly to supply and demand and apply appropriate price adjustments in real-time.

Billing

For every Smart Meter detailed consumption data is recorded in the database for each 15-minute period. Each of these readings is saved with the appropriate timestamp and each smart meter has a rate for a specified timeframe. With the help of the in-memory database it is possible to calculate which reading goes with which rate within seconds. Then the proper timeslot for this timestamp is being searched and the consumption in kWh is multiplied with the timeslot’s price. This is done for each reading. At the end all the costs for the single readings are added together. All calculations are done on database level. The appropriate query provides the billing value and is processed by the database within three seconds on the described hardware. Due to parallelization up to 1,000 of these queries can be done within ten seconds. Furthermore, we provide an extended model that allows defining various price adjustments for each rate. These can be diverse bonuses or similar modifications. They do not have any noticeable influence on the execution time of the query.

Consumption Prediction

Predictions can be made on the basis of the historical consumption data of a Smart Meter. This process does not need any pre-aggregations and is executed on the raw data. In the context of our project we evaluated existing prediction algorithms and implemented one that reached a deviation on average of 6.41%. This was not proven generally – only tested on the real metering data we were provided with. The prediction allows us to calculate possible costs with all rates and compare which rate is the most reasonable one. Since this is done in less than two seconds a customer can get an overview of the predicted costs with every rate and then decide whether he wants to change his own rate in real-time [5].

Switching the Rate

In our system, switching the rate means two simple database operations. This is best explained by an example: Customer “A” wants to switch at the time “t”. Since the beginning of the year he has had “Rate 1” and now plans on changing to “Rate 2”. Rate 1 is valid till the end of the year. On the iPad he can issue a respective request by a simple push of a finger. Subsequently, in the database the validity date of the currently active connection between Smart Meter “A” and “Rate 1” is set to “t”. A new record is created for the Smart Meter of customer “A” with “Rate 2”, featuring “t” as validity start point. For all following billings and cost accountings in the database the values read from the Smart Meter of customer “A” are set against the timeslots from “Rate 2”. The switching process is illustrated more clearly below. The tables do not represent the actual database schema, but rather are meant to point out respective values of the database entries and make the steps more comprehensible. (I) Original state at time “t“:

  • Rate 1 and Rate 2 are generally valid to the end of 2011
  • Customer “A“ uses Rate 1 since the beginning of 2011
Customer Rate
  Name Start End
A Rate 1 01/01/2011 12/31/2011

(II) Action: Customer uses iPad to switch to “Rate 2“ (III) Final state after switching at time t:

  • Customer A having had Rate 1 until time t is saved
  • Customer A using Rate 2 from time t is saved
Customer Rate
  Name Start End
A Rate 1 01/01/2011 t
A Rate 2 t 12/31/2011

References

[1] Plattner, H., & Zeier, A.: In-Memory Data Management: An Inflection Point for Enterprise Applications (1st ed.). Springer 2011.
[2] Statistisches Bundesamt: Haushalte
[3] E-Energy Modellstadt Mannheim: Praxistests im Projekt
[4] Schapranow, M.P., Kühne, R., Zeier, A.: Real-Time Billing in Smart Grid Infrastructures. In: Power and Energy Student Summit 2010 - Integration of Renewable Energies into the Grid (2010)
[5] Galitz, W.O.: The Essential Guide to User Interface Design: An Introduction to GUI Design Principles and Techniques. Wiley & Sons, 3rd ed. edn. (Apr 2007)