P24
Machine learning techniques as innovative part The controller developed in STORM makes use of machine learning algorithms that are based on finding and describing structural patterns indata. These techniques have theadvantage that they ‘learn’ the behaviour of the network by themselves without the need to be extensively tuned at the installation and by using a set of training data. Once the behaviour of the network is known, the algorithms will optimize the behaviour of the network, dependant on the chosen control strategy, or a combination of control strategies.
and cold to a so-called backbone network. This backbone network (designed for a temperature regime of 28/16°C) provides heat and cold to the cluster networks (see figure 1), to which all buildings are connected. The ambition is to make these clusters more energy self-sufficient by maximising the energy exchange between buildings and thus by minimizing the energy exchange with the backbone network. This is vital because the capacity of the backbone network and the underground storage system is limited: only the net excess of heat will be transported through the backbone network and eventually delivered to or extracted from the mines. Therefore, the mines will be utilised more as a gigantic aquifer thermal energy storage system (ATES) rather than a finite source of heat and cold. As a result, more clusters and thus more buildings can be connected to the existing backbone mine water system, and expansions of the network become possible without any major investments (see figure 5). However, for fully deploying this system, an automated and smart control system is necessary (see figure 2).
Figure 4: control strategies in district heating networks
Multiple control strategies The controller is able to deal with a number of control strategies (see figure 4) since not every grid operator has the same objectives: • In the ‘cell balancing’ strategy, the controller aims to balance the demand and the production within a local cluster of producers and consumers of heat and/or cold, making the cluster as self-sufficient as possible. This means that energy exchange within the cluster is maximized, and only when an imbalance in the cluster persists, energy is exchanged with the other networks'. These strategy will obviously be applied in the Mijnwater network. • Secondly, there is a ‘peak shaving/valley filling’ control strategy included. In district heating grids supplied by cogeneration units (CHP) or renewable energy sources, the source is often not dimensioned for peak load conditions, since the investment costs for these sources are usually high and peak load conditions only occur a number of hours per year. Therefore, in these grids, the sustainable source is assisted by cheaper fossil fuel fired installations, like gas boilers for example, serving as peak production units. In this strategy, the objective is to minimize or even eliminate the running hours of the peak installations. In this way, new networks investments in this peak installations can be avoided. • In the last control strategy, the DHC network provides balancing services to the electrical grid, called ‘market interaction’. When electric and thermal grids are coupled, for example by means of a heat pump and/or a CHP, the intrinsic flexibility in the DHC network can be used to control these heat pumps or CHPs depending on prices on the energy market, which on their term are in relation to the availability of renewable power on the electrical grids. This control strategy therefore not only maximises the profit of the networks operator, but also supports the balance in the electrical grid.
Demonstration at Rottne, Växjö, Sweden The district heating system operated by Växjö Energi in Rottne, Sweden is the second demonstration site (see figure 3). The production is based on two wood chip boilers, complemented with a traditional oil boiler for peak load coverage. The purpose of the STORM project here is to minimize the oil usage and optimize the operational behaviour of the bio-fuel boilers. Currently, a smart heat grid technology is already available in the network and will be upgraded with STORM technology. In comparison to the demo site in Heerlen, this network is a more traditional set-up and acts as a model for many district heating networks in Europe. It will ensure replicability for other DHC networks in Europe. Figure 3: Technical lay-out of Rottne, Sweden Rottne DHC network (blue lines) with consumer sub-stations (green buildings), NODA IEC installations (blue circles) and production site (brown circle).
E N E R G Y A N D E N V I R O N M E N T
Made with FlippingBook Ebook Creator