logo Datameteo Datameteo previsioni meteo

Hydro and Mini Hydro Power Plant Forecast Predictions

Forecasting of hydropower plants production with a “physically based” approach using high resolution weather models As known in many countries part of the imbalance costs (due to difference between energy forecasted and actually delivered to the grid) are now charged also to electricity producers from renewable sources

As a consequence, appears nowadays strong the demand of support from specific forecasting systems that are able to determine, with a reasonable margin of error, the potential production of the plants with some days advance.

The demand related to the recent Resolution is furthermore part of a wider context of potential production plants forecasting, also with longer temporal scales, related to the improvement of plant management.

Simultaneously, the modeling software in water resources and environmental issues are continuously more implemented in decision and management support platforms. In many operative fields, in fact, there is the necessity to have available this kind of platforms that, based on a strong and detailed representation of relatively complex physical phenomenology, allow to provide the synthetic information needed to elaborate operation and strategic decisions.

For example, our MeteoBrowser platform, allows you to easily check where there are precipitations of a certain entity, and thanks to temperatures, check also the type. This allows you to assess in the scope of catchment, which may be the implications for environmental monitoring and management of the basin.

Hydro Forecast long term prediction

In the image above the details of Northwest Italy with overlapping cartographic layers of air temperature and precipitation to have with a single glance the perception of how much precipitation will be expected and even what kind, if rainy or snowy.

This approach results widely effective also in the environment of systems finalized to the forecasting of potential production of hydropower plants.

Subjects involved in power production and plant management, likewise the operators of energy trading, are nowadays focusing their attention to forecasting systems based on this kind of approach, in the perspective of both a short-period (1-3 days) and a monthly basis forecast. The goals are multiples: plan of the energy delivery service, energy trading, forecasting and management of finance charge, estimation of resources on a seasonal scale.

The forecasting systems referred above represent, for the hydropower sector, a strong and adequate answer, available in short time, not only in relation with the short-period forecasting of energy production, but also for a better characterization and management on a monthly or seasonal scale.


How is it possible to forecast the potential power production of a plant?

The approach adopted is based on the experience gained by Datameteo in partnership with relevant actors in the implementation of many hydrological real-time forecasting and management systems in our country, since the end of ‘90s, both with purpose of civil protection and improved water resources management. 


The modelling component is based on different calculation code of a hydro-meteorological model that uses data of high resolution meteorological models of Datameteo.

For each plant a “site specific” approach is adopted, finding the more adequate model schematization with reference to a hydrological component and, if necessary, an hydraulic component too.


The model so built is calibrated on the basis of historical data and then made operative in automatic mode, with a daily input of field observation data from the regional networks as well weather forecasting data.



What do we mean for High Resolution Weather Forecast?

High resolution means quality weather forecast everywhere based on WRF numerical model irrespective of whether a meteorological recording station exists or not. Through the high spatial resolution, which calculates forecast for every 1-3-18 km, Datameteo produces local forecasts which include the effects of local climate, topography and soil cover. We conduct regular verification of our forecast, comparing them to actual observation data. Our weather forecast are live corrected by sat feed or statistical approaches to give the best radiation, temperature, wind predictions. Not only these meteorological variables are involved but many other can be extracted to have accurate forecast such as: precipitation, cloudiness, air mass convection.


WRF Model for accurate snow and precipitation predictions
All the atmosferical parameters ( rain amount, rain rain, snow, melted snow, model run off , soil moistuire etc..)  fraction, are available via API simply passing latitude and longitude of the location.

Which are the characteristics of the service offered?

The described approach is applicable and effective with both high dimension plants and not relevant plants (<10MW) located on secondary catchments.
The simulations are run with daily or upper frequency, with reference to two different lapses of time, short and long period.  

The short period simulations covered a lapse of time typically of 48-72 hours for the scheduling of the energy delivering to the grid on the “day before” market and on the “intra-day” markets.

The long period simulations covered lapses of time from some weeks to one year, with reference to different rainfall scenarios (typically scarce, medium, plentiful year) on the basis of the real “present” conditions of the catchment. So these statistical analysis aren’t based on historical data but on the real potential production of the plant based on the real hydrological conditions of the catchment.

The simulation results can be sent via e-mail, FTP or eventually published on a specific website.

Hydrological Scenario

Results of long period simulations – power forecasting for the following 12 months with 3 possible evolutions of the hydrologic scenario based on the present conditions of the catchment.


Which parameters will be requested for the start-up of the power forecasting?

For the start-up of the service is necessary to know the position of the plant and its main characteristics like the maximum power and the operational conditions.

Moreover, it’s necessary to have available a historical series of the plant power production of 1-2 years in order to calibrate the model.
At last, it’s necessary to activate a hyperlink for the daily acquisition of the production data (e.g. “Sorter” system).

If relevant, it’s moreover possible to take in consideration particular conditions like the possible obstruction of the grates for excessive material transport or the presence of specific prescribed operational rules, like for example a reduction of water diversion in some periods of the year.

Which results can be achieved with the forecasting service?

The performance of the forecasting system depends on the plant characteristics and on the quality of the available input data. Generally, the modeling system allows to respect the tolerance limits imposed by the AEEG Resolution N° 281/2012/R/efr for the majority of the days and the hours of the years, unless unavailability or other specific conditions on the plant, also with reference with the 2 days before forecasting.

The accuracy results moreover increased reducing the forecast period for the intra-day markets.

A relevant increase of the performance is moreover obtainable activating specific auto-correcting functionalities of the system due to the real-time or semi-postponed assimilation of the plant power production data.

As an example, the following tables show the performance obtained in the first semester of the 2012 for a big-size plant and a group of not relevant plants.

The results related to the not relevant plants are calculated singularly for each plant. The possible aggregation of many plants for zone allows
 a significant reduction of the forecasting error.

Plant type


Max Power

70 MW

Catchment Area

> 10.000 Km2


Forecast day-1

Forecast day-2

Average Error % *

8 %

10 %

Hours with Error < 10%

76 %

67 %

Hours with Error < 20%

96 %

88 %



Plants Type

Not relevants

Max Power

From 2 to 10 MW

Catchment Area

From 5 to 50 Km2


Forecast day-1

Forecast day-2

Average Error %

8 %

10-15 %

Hours with Error < 10%

62-66 %

50-58 %

Hours with Error < 20%

85-91    %

77- 83%

(*) calculated in absence of auto-correction related to the real power production data that allow a further improvement of the accuracy.

Accurate mini-idro power forecast

Comparison between real (green) and simulated (blue) power for a relevant plant. For big size plants the system performance is particularly high.


Which are the strong points of the solution proposed?

At the present the forecasting systems for hydropower plants are typically based on a statistical approach, founded on correlations between rainfall and plant power production or, in some cases, on a wide-scale modeling, that is able to schematize some components of the physical process though maintaining a “global” type approach. In both cases, the input data to the forecasting system are mainly composed from the meteorological forecast component instead of the real field observations of the processes in progress in the catchment.

With this approach it’s often necessary to operate at an aggregated scale for the plants of reduced size and it’s impossible to keep in consideration site-specific situations that characterize the contributing catchment and, therefore, the plant power production. 


The modeling approach adopted by Datameteo offers, in respect to these, several advantages that turn in a higher performance and reliability of the system, like:

•    Site specific approach
: for each plant a specific model is implemented and calibrated choosing the schematization that assure the better representation of the physical process;

•    The consequently possibility to consider specific characteristics of the plant’s contributing catchment, like the presence of lakes or other control devices, as well prescribed operational rules (minimum flow, seasonal operational variations), problems related to sediment transport, etc.;

•    Priority use of field observed data (rain gauges, thermometers and hydrometers) and, consequently, a less dependence from the weather forecasting component, affected by a higher uncertainty;

•    Use of forecasting input interpreted and corrected in relation to the context of reference;

•    Real-time or semi-postponed acquisition of plant’s power production data, that allow the model to “auto-correct” itself increasing the forecast reliability;

•    Use of deterministic and physically based modeling software, able to simulate “in continuous” the hydrological process, applied worldwide as a standard;

•    The expertise and knowledge of the staff with a huge experience in hydrological and hydraulic processes simulation, included forecasting systems in Italy for civil protection purpose.

Phisically based idro model


Which are the new frontiers?

If on one hand the development of forecasting systems leads to use a consolidated approach to create increasingly complex  and detailed models, first of all the model on the entire Po basin, on the other hand the research and technological developments lead to the rise of new demands and the availability of new types of data.

Parallel to the development of systems for forecasting and flood management, the main interest in the last decade even after taking several catastrophic events in our country, in recent years the attention of the scientific community and the general public is turning more and more towards issues of hydrological lean and dry phases.

Although the modeling of hydrological processes in normal or lean conditions shows level of complexity in some ways greater than flood events, having to consider even neglected phenomena in the representation of the latter (such as interactions with the ground or the presence of derivations), decision support systems in the field of water resource management offer to the decision maker  broader and more effective possibilities of intervention, as the regulation of reservoirs or derivations, proving again increasingly indispensable instrument in support of the taken actions, not only in emergency conditions.

In this context activities of applied research and technological innovation aimed at using satellite data have priority relevance , both radar-SAR typology and optical , with different resolution, supporting the information obtained with traditional methods

Satellite feed for iidro power

Satellite and digital model of a river: the future of environmental monitoring and forecasting on river basins using remote technology.

The main goal of such research is to evaluate the quantitative improvement of the predictive capacity of extreme events on Italian territory through the use of a set of data taken from observations of the sensors available today for Earth observation from space and those that will be mounted on satellites presumably in the next five years.

Satellite observation of the state of soil saturation, with its interpretative models, is certainly the main component of innovation, which is complemented by the ability to observe in all the weather conditions the status and morphology of watercourses. In the second instance, it is of great interest, in the context of Civil protection activities, observation of the built and post-event inundated areas.