The forecasting methods will respond to different needs:  the day-ahead operation planner needs methods that provide a load/renewables/DG forecast together with a distribution of actual outcomes that is conditional on the forecast; the rolling-horizon operations planner needs a more dynamic, time-staged production of load/renewable/DG forecast and their conditional outcome distribution; the market participant needs price distribution forecasts, and so on.  Each requires a different forecasting technique.

From a DISCO (Distribution Company) point of view, day-ahead load forecasting in the smart grid may be much more difficult than current strategies because the uncertain generation of renewable/DG and DR sources can deteriorate the net load of the downside grid. Am I right?

I feel what ever the forecasting techniques reported and being used are fairly good enough and I doubt any wider scope of improvement in them. Because a forecast can never be a reality to 100%.  Rather than thinking about the improvement  of forecasting techniques, it is high time to concentrate on the methods of bridging the gap between the load and generation. Please share any ideas if I am wrong.

Smart Grid will have an effect of smoothening the peak demands. So, the hourly load demand forecast approach has to be combined with smoothening algorithms to flatten the demand curve. Smart Grid is not expected to bring uncertainty but bring order to the demand phenomenon by dampening the rise in demand.

It depends on which part of the SG (neighbourhood, regional, distribution or transmission part) are the investigates object: for neighborthood or regional where period of itrusion of DER or DG may be etablished, answer is: yes, whereas for distribution or trasmission part answer is: no.

This is an interesting question. I believe that we need to develop new load forecasting methods that are more suitable for a smart grid environment. In future networks, we will see much more variability in demand patterns, both temporally and spatially. Load forecasting methods will need to be probabilistic, e.g. producing ranges of values rather than point forecasts, in order to better model the uncertainties due to DER and DR. In addition, there will be much more interaction between demand, DER, and electricity prices (e.g. variable prices will affect demand, and variable demands will affect prices). Therefore, load forecasting, DER forecasting and electricity price forecasting will need to be integrated, i.e. considering all of these interactions and feedback loops. Data from smart meters can help us understand these changes in load patterns in more detail and help us to produce more useful forecasts.

Dear Morteza Shabanzadeh,

a time series analysis (based on decomposition and segmentation of the load time series) could be a good method. Let me suggest you this paper:

http://www.sciencedirect.com/science/article/pii/S1110016811000330

I hope it might be helpful to your research.

Kind regards.

Emanuele

This is really an interesting question and it is relevant!

My view is that, the basic demand forecast technique/method needs not to change. What one can do is to add a capability for scenarios of smart grid and DER/DR. The need (load) will exist no matter what, but the supply method or source and the tricks around it can change, and that is why one needs to have scenarios to model the smart grids and DER/DR. Thing is, a family (household) will always need energy to cook, heat up the water, etc., but the way that energy is sourced and supplied is subject to change and that is where we need to draw scenarios.

Also, the impact will show differently at different levels of the electrical network hierarchy: the house meter box may show pure demand as used by the household; a node on the MV line may show the impact of distributed generators connected and the switching of supply caused by the smart grids technologies used, and so on.

Conclude: keep the forecast technique and simply draw scenarios to cater for the forecast results

Excellent question. Three responses: First, load forecasting is being reinvented for a world of dynamic demand, supply, price, and regulation. Given what's going on in supply, load delivery will become a vital service.

Second, to enable load delivery, dynamic system modeling (well beyond typical probabilistic and scenario-based approaches) will be required, and soon. Take a close look at our industry today  --  doesn't it remind you of Wall Street before the serious quants took charge?

Third, to permit load delivery, rates will be gradually reengineered to reward the kinds of load that benefit the system. Residential demand charges are an example. Load forecasts will be based on anticipated load delivery.

One reason load delivery will be such a valuable services is that our industry's supply-side modeling is even more troubled than its demand-side modeling. At the recent  PowerGen conference, CAISO CEO Steve Berberich said that on the supply side, "capacity will no longer be the coin of the realm. Capability will be the coin of the realm." He's right, but we don't know what it means. The capabilities of resources and dispatching protocols are not evident, and we haven't built our reliability models based on them.

If supply is about to experience a new storm of volatility and confusion, reliable load delivery will be worth more than ever. We're going to make load forecasts, and then we'll be asked to make them come true.