Predictive supply temperature optimization of district heating networks
Laakkonen, Leo (2016)
Laakkonen, Leo
2016
Automaatiotekniikan koulutusohjelma
Teknisten tieteiden tiedekunta - Faculty of Engineering Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2016-11-09
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tty-201610244637
https://urn.fi/URN:NBN:fi:tty-201610244637
Tiivistelmä
In combined heat and power (CHP) the waste heat of power production is used for heating the water in district heating (DH) plants. Fluctuating power production in CHP plants may cause unwanted disturbances in district heating networks (DHN), which leads to the situation that the best efficiency in CHP production is not achieved. Although DH -systems are usually automated, the supply temperature is still primarily chosen manually by the operator or it is based on current outdoor temperature. This is because of the uncertain heat demand in near future and uncertain behaviour of delay from heat supplier to consumers, which make the temperature scheduling challenging.
In this work, future heat demand and return water temperature are predicted based on outdoor temperature forecast and process data history using neural network predictors. Consumers in network are presumed to be similar, but their distances from production sites vary thus creating a distribution function of range. Delay is modelled as a distribution function based on the distances between heat consumers and the suppliers, which weights the supply temperatures from last few hours calculating the average supply temperature received by the consumers. The brute force optimizer utilizes these models to optimize the supply temperature by minimizing heat loss and pumping costs. Delays are dependent on mass flows, but they are not set as variables during optimizations due to formulation and performance challenges. Instead, the delays are determined for each optimization cycle based on mass flows of earlier cycle and they are iterated as long as delays are converged. The resulting supply temperature curve is a discrete curve that cuts the heat load peaks by charging and discharging the energy content of the DHN. Optimization keeps the supply water temperature and flow rates in control and stabilizes the network smoothly and efficiently after disturbances.
In this work, the optimization is demonstrated in case study of Kuopio DHN, operated by Kuopion Energia Oy. Models are fitted and calibrated into Kuopio DHN and the optimization is compared to the measured supply temperatures and instructional temperatures by Energiateollisuus ry. Standard deviation of heat load predictor was 6.3 MW, return temperature predictor 0.78 ℃ and for delay distribution model 0.30 ℃. Main actions of optimization were delay prediction, reduction of supply temperature and minimizing pumping during high pumping costs. Optimization gave savings of 1.2 – 1.7 % on heat delivery.
In this work, future heat demand and return water temperature are predicted based on outdoor temperature forecast and process data history using neural network predictors. Consumers in network are presumed to be similar, but their distances from production sites vary thus creating a distribution function of range. Delay is modelled as a distribution function based on the distances between heat consumers and the suppliers, which weights the supply temperatures from last few hours calculating the average supply temperature received by the consumers. The brute force optimizer utilizes these models to optimize the supply temperature by minimizing heat loss and pumping costs. Delays are dependent on mass flows, but they are not set as variables during optimizations due to formulation and performance challenges. Instead, the delays are determined for each optimization cycle based on mass flows of earlier cycle and they are iterated as long as delays are converged. The resulting supply temperature curve is a discrete curve that cuts the heat load peaks by charging and discharging the energy content of the DHN. Optimization keeps the supply water temperature and flow rates in control and stabilizes the network smoothly and efficiently after disturbances.
In this work, the optimization is demonstrated in case study of Kuopio DHN, operated by Kuopion Energia Oy. Models are fitted and calibrated into Kuopio DHN and the optimization is compared to the measured supply temperatures and instructional temperatures by Energiateollisuus ry. Standard deviation of heat load predictor was 6.3 MW, return temperature predictor 0.78 ℃ and for delay distribution model 0.30 ℃. Main actions of optimization were delay prediction, reduction of supply temperature and minimizing pumping during high pumping costs. Optimization gave savings of 1.2 – 1.7 % on heat delivery.