Background
Accurate and efficient forecasts of energy consumption and production are a fundamental precondition for dynamic and fine-grained scheduling. Based on forecasts, schedules for RES supply and demand are initially computed and afterwards incrementally maintained if forecast values change over time. Specific characteristics of energy time series like multi-seasonality (daily, weekly, annual) or dependency on external information like weather or calendar events motivate the employment of forecast models tailor-made for the energy domain. In addition, different forecast horizons (short-term, mid-term, long-term) as well as the forecasting of flex-offers have to be provided. Finally, forecasting faces the challenge of a large scale hierarchical system with high update rates of new measurements and evolving time series, which require continuous model evaluation and adaptation.
Overview Forecasting Approach
Our system architecture consists of two main components: (1) the transparent forecast model creation and usage and (2) the transparent forecast model update and maintenance.
Model Creation
The model creation component automatically creates forecast models, either beforehand or when the respective forecasts are demanded, where we apply the Engle, Granger, Ramanathan, and Vahid-Arraghi (EGRV) Model and the Triple Seasonality Holt Winters (HWT) Model. The EGRV-Model is a multi-equation energy demand forecast model that uses an individual model for each intra-day period (e.g., one model for each hour). In addition, weather information, calendar events (e.g., holidays) and context knowledge of energy types (e.g., constraints on the produced energy) are included. If the EGRV model does not provide accurate results, we fall back to the alternative (more robust) HWT-Model, which is a energy specific adaptation of the general purpose Holt-Winters exponential smoothing forecast model. Model creation involves computationally expensive parameter estimation, where we reuse existing well-established local (e.g., Downhill-Simplex) and global (e.g., Simulated Annealing) parameter estimators.
Model Usage
The scheduling component can explicitly request forecast values or may register forecast queries as continuous queries in order to obtain notifications whenever the forecast values change significantly.
Model Update and Maintenance
A continuous stream of new measurements require a continuous maintenance of forecast models. For each new time series value, we update our forecast models that consists of a simple update of smoothing constants or the shift of lagged input values. This implies low additional costs. Due to changing time series characteristics, the accuracy of the forecast models might be reduced over time, which poses the necessity of adapting the model parameters. To evaluate the need for a model adaptation, we offer different model evaluation strategies (e.g., time- or threshold-based). Furthermore, the model adaption exploits the context knowledge of previous model estimations in order to speed up this time-consuming process of parameter re-estimation.
Research Results
Forecasting always needs to cope with the trade off between forecast accuracy and runtime of parameter estimation. We offer different optimizations that address this challenge in terms of model creation on physical (parallelized model creation) and logical level (hierarchical forecasting), model usage (publish-subscribe forecast queries) and model maintenance (context-aware model adaption).
Parallelized Model Creation
Energy-domain-specific multi-equation forecast models (e.g., EGRV-Model) comprise a large number of parameters, for what reason the estimation of such models is time consuming. As multi-equation models consist of several independent individual models, we can reduce the time needed for estimating such models by partitioning and parallelization. Therefore, we horizontally partition the time series according to the multi-equation access pattern and parallelize the model estimation process according to the resulting independent data partitions.
Hierarchical Forecasting
Based on the hierarchical organization of the energy market, the macroscopic system architecture is inherently distributed, where at each system node, one or several forecast models might be created and used according to the scope of the particular role. Beside the use of individual forecast models, forecast models can be used to aggregate or disaggregate forecast values without the need for individual models at each system node. Therefore, we provide an advisor component that computes for a given hierarchical structure a configuration of forecast models according to specified accuracy and runtime constraints.
Publish-Subscribe Forecast Queries
The scheduling component does not always need or even not want to have the most up-to-date forecast values as every new forecast value triggers the computationally expensive maintenance of schedules. Only if forecast values change significantly, notifications are required. Therefore, in addition to requesting forecast values, we offer the interaction scheme of so-called publish-subscribe forecast queries. Hereby, our goal is to minimize the overall costs of the subscriber.
Context-Aware Model Adaptation
The development of energy time series strongly depends on background processes and influences that together form the context of a time series. Observing these context information offers the possibility of storing previous models in conjunction to their corresponding context information within a repository to reuse them whenever a similar context reoccurs. This kind of case-based reasoning approach achieves a higher forecast accuracy in less time, especially for complex models.
Accurate and efficient forecasts of energy consumption and production are a fundamental precondition for dynamic and fine-grained scheduling. Based on forecasts, schedules for RES supply and demand are initially computed and afterwards incrementally maintained if forecast values change over time. Specific characteristics of energy time series like multi-seasonality (daily, weekly, annual) or dependency on external information like weather or calendar events motivate the employment of forecast models tailor-made for the energy domain. In addition, different forecast horizons (short-term, mid-term, long-term) as well as the forecasting of flex-offers have to be provided. Finally, forecasting faces the challenge of a large scale hierarchical system with high update rates of new measurements and evolving time series, which require continuous model evaluation and adaptation.
Overview Forecasting Approach
Our system architecture consists of two main components: (1) the transparent forecast model creation and usage and (2) the transparent forecast model update and maintenance.
Model Creation
The model creation component automatically creates forecast models, either beforehand or when the respective forecasts are demanded, where we apply the Engle, Granger, Ramanathan, and Vahid-Arraghi (EGRV) Model and the Triple Seasonality Holt Winters (HWT) Model. The EGRV-Model is a multi-equation energy demand forecast model that uses an individual model for each intra-day period (e.g., one model for each hour). In addition, weather information, calendar events (e.g., holidays) and context knowledge of energy types (e.g., constraints on the produced energy) are included. If the EGRV model does not provide accurate results, we fall back to the alternative (more robust) HWT-Model, which is a energy specific adaptation of the general purpose Holt-Winters exponential smoothing forecast model. Model creation involves computationally expensive parameter estimation, where we reuse existing well-established local (e.g., Downhill-Simplex) and global (e.g., Simulated Annealing) parameter estimators.
Model Usage
The scheduling component can explicitly request forecast values or may register forecast queries as continuous queries in order to obtain notifications whenever the forecast values change significantly.
Model Update and Maintenance
A continuous stream of new measurements require a continuous maintenance of forecast models. For each new time series value, we update our forecast models that consists of a simple update of smoothing constants or the shift of lagged input values. This implies low additional costs. Due to changing time series characteristics, the accuracy of the forecast models might be reduced over time, which poses the necessity of adapting the model parameters. To evaluate the need for a model adaptation, we offer different model evaluation strategies (e.g., time- or threshold-based). Furthermore, the model adaption exploits the context knowledge of previous model estimations in order to speed up this time-consuming process of parameter re-estimation.
Research Results
Forecasting always needs to cope with the trade off between forecast accuracy and runtime of parameter estimation. We offer different optimizations that address this challenge in terms of model creation on physical (parallelized model creation) and logical level (hierarchical forecasting), model usage (publish-subscribe forecast queries) and model maintenance (context-aware model adaption).
Parallelized Model Creation
Energy-domain-specific multi-equation forecast models (e.g., EGRV-Model) comprise a large number of parameters, for what reason the estimation of such models is time consuming. As multi-equation models consist of several independent individual models, we can reduce the time needed for estimating such models by partitioning and parallelization. Therefore, we horizontally partition the time series according to the multi-equation access pattern and parallelize the model estimation process according to the resulting independent data partitions.
Hierarchical Forecasting
Based on the hierarchical organization of the energy market, the macroscopic system architecture is inherently distributed, where at each system node, one or several forecast models might be created and used according to the scope of the particular role. Beside the use of individual forecast models, forecast models can be used to aggregate or disaggregate forecast values without the need for individual models at each system node. Therefore, we provide an advisor component that computes for a given hierarchical structure a configuration of forecast models according to specified accuracy and runtime constraints.
Publish-Subscribe Forecast Queries
The scheduling component does not always need or even not want to have the most up-to-date forecast values as every new forecast value triggers the computationally expensive maintenance of schedules. Only if forecast values change significantly, notifications are required. Therefore, in addition to requesting forecast values, we offer the interaction scheme of so-called publish-subscribe forecast queries. Hereby, our goal is to minimize the overall costs of the subscriber.
Context-Aware Model Adaptation
The development of energy time series strongly depends on background processes and influences that together form the context of a time series. Observing these context information offers the possibility of storing previous models in conjunction to their corresponding context information within a repository to reuse them whenever a similar context reoccurs. This kind of case-based reasoning approach achieves a higher forecast accuracy in less time, especially for complex models.