Production planning for distributed district heating networks with

At the Modelica Conference in Paris, I had the opportunity to present to the audience a work that I, together with colleagues from Modelon, Vattenfall R&D and SICS Swedish ICT, carried out over 2014 – employing for production planning in distributed district heating networks.

Now I’m glad to offer you a glimpse into this topic that lately has increased in interest. The models and applied tool have high relevance for operators of both thermal power plants and district heating networks.

Using our approach, the quality of the production plans can be substantially increased as the supply temperature and flow are optimized together with the unit loads. The production plan accounts also for the typical operational constraints in the network: pump capacity or minimum temperature at the customer station.

The improved plan quality is a direct result of the chosen technology:

  1. The network and power plant are described by physical models. Existing standard approaches include considerable model simplifications and typically use linear models.
  2. Non-linear dynamic optimization techniques make it possible:

                     - to optimize the physical models without simplifications
                     - and to set constraint on any temperature, flow or pressure in the system.

Our workflow proposes to separate the production planning problem into two optimization problems in series:

  • a so called Unit Commitment Problem (UCP), expressed with piecewise linear and discrete time models, to generate the optimal status (on/off) of the plants. This is the standard approach based on linear programming.
  • an Economic Dispatch Problem (EDP), expressed with continuous time and physics-based Modelica models, assuming that the status of the plants is known from the UCP optimization. This is the critical step in our approach that uses nonlinear optimization techniques to optimize all continuous variables like temperature, flow and pressure.

By our proposed method and optimization solutions, engineers in the field can for example:

  • learn the limits of performance of the power plant and the network
  • increase the cogeneration plant efficiency
  • improve the network economy by minimizing production and operation costs, as well as heat losses.

We applied our methods on a representation of the district heat network in Uppsala, including three production units.

Our study included four cases with incremental complexity:

  1. no network delay
  2. network delay and heat loss
  3. distributed network (several customer clusters)
  4. case comprising several days and units and a distributed network

With this study we could demonstrate that separation into two optimization studies works well and that limitations of real plants can conveniently be incorporated. The optimization model for the network showed among others that:

  • it is possible to optimize systems with variable time delays
  • you can lower power plant production peaks in production planning
  • heat accumulation, both in network and accumulator, can be exploited for higher flexibility and electricity production
  • understand effects of temperature versus mass flow changes in pipes

For the ones of you who have attended our Tutorial at the Modelica Conference 2015 you can recognize the tool used in the training.

I’ll be happy to hear more about how this work is relevant to yours, and to answer questions on how to implement such an approach and what you can expect as results.

Per-Ola Larsson holds a Ph.D. in Automatic Control and an M.Sc. in Electrical Engineering, both from Lund University and with focus on optimization, process control and signal processing. At Modelon he is a simulation consultant and project manager in several consultancy and research projects within the field of thermodynamics, containing both steady-state and dynamic models as well as several types of optimizations. 


by Per-Ola Larsson