Accurately forecasted demand responses are necessary for the sustainable electricity systems

16.11.2015

Distributed dynamic demand side responses can be made fast, reliable, predictable and inexpensive. Such demand side responses are excellent for balancing the increasingly intermittent generation. Typically distributed demand responses are implemented in a way that many good properties such as response speed and predictability are lost. Unpredictable responses implemented in large scale mean increasing balancing errors for the aggregators and retailers and the whole electricity system. Managing large balancing errors typically causes increasing costs and emissions, because fossil fuel power plants, centralised electricity storages, and massive network investments are used for the purpose. It is much better to automate and improve the demand response related processes, continue opening the markets for demand side flexibility without removing its incentives for response forecasting, and develop and implement models and methods that forecast and optimise the responses accurately. I uploaded a two page paper to explain the importance of accurate response forecasting and the contributions of the project RESPONSE.

 

Background

 

The electricity generation is increasingly based on renewable energy sources that are often intermittent and distributed. The electricity systems urgently need more balancing power and inertia in order to maintain the security of the power supply at a level that we are used to. Balancing with centralised power plants only is not any more a sustainable solution. It is both expensive and polluting. Electricity storage technologies keep on developing, but they are still very far from providing an economically feasible and environmentally sustainable main solution for balancing the electricity systems. At the same time we have much cheap and environmentally friendly suitable control capacity for fast responses unused. Most of this unused demand side response potential is at small and medium size electricity consumers. Demand side responses are faster, more reliable, better located and less expensive that most generation side balancing resources. The main limitation of demand side responses is the duration of the responses. So they cannot replace the need of power plants as slow reserves. The reasons for not adequately investing in demand side responses include the following. 1) Electricity market structures and business models and the power system balance control mechanisms and automation systems are still too centralised, hierarchical and complex for the access of demand side resources. 2) Reliability, cyber security and the interfaces of building and home automation are a challenge. 3) Verification of fast distributed responses may be expensive. 4) The demand side responses have often been too unpredictable and unreliable. These barriers are gradually being removed and the engagement of demand response is increasing.

 

Developing models for response forecasting and optimisation in project RESPONSE 

 

The Finnish Academy research project RESPONSE develops methods needed for solving the two last mentioned points. It develops accurate and maintainable models for forecasting and optimisation of the control responses of small electricity consumers such as houses that use electricity for heating, cooling and other processes that include some kind of energy storage.

The project RESPONSE continues earlier research collaboration of the partners. For example, several forecasting methods were compared and their mutual strengths and weaknesses assessed before the start of the project [1]. There the assessed methods included smart metering data based profile models, a neural network (NN) model, and a Kalman-filter based predictor with input nonlinearities and a physically based main structure. Then an optimization scheme based on multi-objective genetic algorithm (GA) for designing data-driven models for short-term forecasting of electric loads was developed and reported [2]. After that a hybrid model for short-term forecasting of aggregated thermal loads and their control responses was developed and studied using data of about 700 customers from dynamic load control field tests [3]. The hybrid model developed combines partly physically based models with a support vector machine (SVM) and forecasts the responses more accurately than any of the component methods alone.

The research topics for the future include alternative hybrid forecasting approaches, both aggregated models and individual house models, preparing more field tests data for the development and analysis of the methods, development and analysis of criteria for the comparison of methods, the value of forecasting accuracy for the electricity market and network, suitability of specific forecasting methods for different purposes such as managing local network constraints or optimisation of the market balance, statistical relevance of the results, confidence intervals of the forecasted values and performance criteria, methods and criteria for the optimisation of the responses, etc.

 

Modelled demand side responses improve managing of the electricity balance

 

Some demand response actors and campaigns try to promote unpredictable or slow responses. That means wasting the big relative strengths of the demand side resources. In large scale it also means increasing balancing errors for the aggregators and retailers and eventually to the whole electricity system. Managing the resulting balancing errors typically causes losses of money and increased emissions, because fossil fuel power plants, centralised electricity storages, and massive network investments are used for the purpose.  Thus removing the DR actors from the incentives to reduce the balancing errors of their own balance and the whole electricity system is not a sustainable solution. It is much better to automate the DR related processes and develop and implement models and methods that forecast and optimise the responses accurately.

 

References

[1] P. Koponen, A. Mutanen, H. Niska, "Assessment of Some Methods for Short-Term Load Forecasting", IEEE PES ISGT 2014, Istanbul Turkey, 12-15 October 2014. 6 p.

[2] H. Niska, P. Koponen, A. Mutanen, “Evolving smart meter data driven model for short-term forecasting of electric loads" IEEE Tenth International Conference on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP) 2015, Singapore, 7-9 April 2015, 6 p.

[3] P. Koponen, H. Niska, “Hybrid Model for Short-Term Forecasting of Loads and Load Control Responses", paper submitted to PSCC 2016, 7 p.

Text: Pekka Koponen

Viimeksi muokattu 9.3.2016
Seuraa meitä:
FacebookSlideshareTwitterYoutube
VAIHDE 029 533 5000
KIRJAAMO 029 533 5049
FAKSI 029 533 5299
   
SÄHKÖPOSTI etunimi.sukunimi@aka.fi
AUKIOLO Arkisin 8.00-16.15
   
HENKILÖHAKU »
YHTEYSTIEDOT, LASKUTUS  JA
REKISTERISELOSTEET»
KYSYMYKSET JA PALAUTE »