Emil Pelikán

A rigorous inter-comparison of ground-level ozone predictions

Novel statistical approaches to prediction have recently been shown to perform well in several scientific fields but have not, until now, been comprehensively evaluated for predicting air pollution. In this paper we report on a model inter-comparison exercise in which 15 different statistical techniques for ozone forecasting were applied to ten data sets representing different meteorological and emission conditions throughout Europe. We also attempt to compare the performance of the statistical techniques with a deterministic chemical trajectory model. Likewise, our exercise includes comparisons of sites, performance indices, forecasting horizons, etc. The comparative evaluation of forecasting performance (benchmarking) produced 1340 yearly time series of daily predictions and the results are described in terms of predefined performance indices. Through analysing associations between the performance indices, we found that the success index is of outstanding significance. For models that are excellent in predicting threshold exceedances and have a high success index, we also observe high performance in the overall goodness of fit. The 8-h average ozone concentration forecast accuracy was found to be superior to the 1-h mean ozone concentration forecast, which makes the former very significant for operational forecasting. The best forecasts were achieved for sites located in rural and suburban areas in Central Europe unaffected by extreme emissions (e.g. from industries). Our results demonstrate that a particular technique is often excellent in some respects but poor in others. For most situations, we recommend neural network and generalised additive models as the best compromise, as these can handle nonlinear associations and can be easily adapted to site specific conditions. In contrast, nonlinear modelling of the dynamical development of univariate ozone time-series was not profitable.

Statistical models to assess the health effects and to forecast ground-level ozone

By means of statistical approaches we attempt to bridge both aspects of the ground-level ozone problem: assessment of health effects and forecasting and warning. Disagreement has been highlighted in the literature recently regarding the adverse health effects of tropospheric ozone pollution. Based on a panel study of children in Leipzig we identified a non-linear (quadratic) concentration-response relationship between ozone and respiratory symptoms. Our results indicate that using ozone as a linear covariate might be a misspecification of the model, which might explain non-uniform results of several field studies in health effects of ozone. We conclude that there is urgent demand for forecasting episodes of high ozone that may help susceptible persons to avoid high exposure. Novel approaches to statistical modelling and data mining are helpful tools in operational smog forecasting. We present a rigorous assessment of the performance of 15 different statistical techniques in an inter-comparison study based on data sets from 10 European regions. To evaluate the results of the inter-comparison exercise we suggest an integrated assessment procedure, which takes the unbalanced study design into consideration. This procedure is based on estimating a statistical model for the performance indices depending on predefined factors, such as site, forecasting technique, forecasting horizon, etc. We find that the best predictions can be achieved for sites located in rural and suburban areas in Central Europe. For application in operational air pollution forecasting we may recommend neural network and generalised additive models, which can handle non-linear associations between atmospheric variables. As an example we demonstrate the application of a Generalised Additive Model (GAM). GAMs are based on smoothing splines for the covariates, i.e., meteorological parameters and concentrations of other pollutants. Finally, it transpired that respiratory symptoms are associated with the daily maximum of the 8-h average ozone concentration, which in turn is best predicted by means of non-linear statistical models. The new air quality directive of the European Commission (Directive 2002/3/EC) accounts for the special relevance of the 8h mean ozone concentration.

Wind power forecasting by an empirical model using NWP outputs

This paper presents a simple and robust wind power forecasting approach using inputs from a state-of-the-art numerical weather prediction models (NWP) with mesoscale resolution. The model can be used for short-term and longer term forecasting horizon up to 72 hours ahead. The forecasting ability of the presented approach is demonstrated using real power production data from the Czech Republic.

K-MEANS CLUSTERING FOR PROBLEMS WITH PERIODIC ATTRIBUTES

The K-means algorithm is very popular in the machine learning community due to its inherent simplicity. However, in its basic form, it is not suitable for use in problems which contain periodic attributes, such as oscillator phase, hour of day or directional heading. A commonly used technique of trigonometrically encoding periodic input attributes to artificially generate the required topology introduces a systematic error. In this paper, a metric which induces a conceptually correct topology for periodic attributes is embedded into the K-means algorithm. This requires solving a non-convex minimization problem in the maximization step. Results of numerical experiments comparing the proposed algorithm to K-means with trigonometric encoding on synthetically generated data are reported. The advantage of using the proposed K-means algorithm is also shown on a real example using gas load data to build simple predictive models.

Fuzzy power management for environmental monitoring systems in tropical regions

Remote environmental monitoring systems require effective energy management to allow reliable long-term operation without frequent maintenance to replace or recharge batteries. To design and analyze relevant energy management strategies, we have developed Simulink-based models of a recently constructed monitoring device to evaluate its potential performance. The model uses long-term solar energy data from two locations, Chamela, Mexico, and Fairview, Canada, to estimate the energy harvesting capabilities of the device. Using the simulator, we have developed and evaluated a fuzzy energy management strategy that determines how the device should operate to match the solar energy profile in each location. Solar energy in Chamela, Mexico is abundant and consistent so an energy harvesting remote monitoring device could have a high activity level without risking device failure. Fairview, Canada, has limited solar resources in the winter but plenty in the summer; a device dependent upon this energy source must adapt its activity level to match energy availability or risk running out of energy. While the simulated device in Mexico outperforms the one in Canada, both succeed in matching the available environmental resources and largely avoid energy related device failure. In the future, their performance can be improved by optimizing the designed strategies and further improving the details of the simulation.

Identification of critical aging segments and hotspots of power transmission lines

Localization of hotspots and critical aging segments of transmission lines is important for operation and asset management of power transmission systems. Conductors can lose their tensile strength due to the adverse effects of conductor aging caused by annealing. Although the loss of conductor strength is gradual, it accumulates over time and increases the probability of outages and blackouts. Therefore, it is important to keep track of conductor temperatures over time and in space, in order to identify segments of power transmission network that may require more close attention, repairs, or reinforcements. This paper describes and illustrates a new methodology for localization of hotspots and identification of critical aging segments of power transmission lines. The methodology uses load information and weather conditions derived from historical weather reanalysis to derive a time series of spatially resolved map of conductor temperatures. The temperature map is then used to estimate loss of conductor tensile strength for each span of the transmission line. The process is illustrated for a sample transmission line, using assumed load current and historical weather information spanning a period of five years. The simulation results show that the proposed approach provides information vital for transmission network operating procedures and transmission asset management.

Statistical modeling of energy production by photovoltaic farms

This paper introduces several alternative statistical approaches to modeling and prediction of electric energy generated by photovoltaic farms. The statistical models use outputs of a numerical weather prediction model as their inputs. Presented statistical models allow for easy-to-compute predictions, both in temporal sense and for out-of-sample individual farms. Model performance is illustrated on a sample of real photovoltaic farms located in the Czech Republic.

Statistical models for disaggregation and reaggregation of natural gas consumption data

In this paper, we present a unified framework for natural gas consumption modeling and forecasting. This consists of models of GAM class and their nonlinear extension, tailored for easy estimation, aggregation and treatment of the delayed relationship between temperature and consumption. Since the consumption data for households and small commercial customers are routinely available in many countries only as long-term sum meter readings, their disaggregation and possibly reaggregation to different time intervals is necessary for a variety of purposes. We show some examples of specific models based on the presented framework and then we demonstrate their use in practice, especially for the disaggregation and reaggregation tasks.

Support Vector Regression of multiple predictive models of downward short-wave radiation

Accurate forecasts of weather conditions are of the utmost importance for the management and operation of renewable energy sources with intermittent (stochastic) production. With the growing amount of intermittent energy sources, the need for precise weather predictions increases. Production of energy from renewable power sources, such as wind and solar, can be predicted using numerical weather prediction models. These models can provide high-resolution, localized forecast of wind speed and solar irradiation. However, different instances of numerical weather prediction models may provide different forecasts, depending on their properties and parameterizations. To alleviate this problem, it is possible to employ multiple models and to combine their outputs to obtain more accurate localized forecasts. This work uses the machine-learning tool of Support Vector Regression to amalgamate downward short-wave radiation forecasts of several numerical weather prediction models. Results of SVR-based multi-model forecasts of irradiation at a large set of locations show a significant improvement of prediction accuracy.

Air pollution episodes: Modelling tools for improved smog management (APPETISE)

Most ambient air quality models are deterministic models or rely upon simple regression based statistics. Their success, however, is limited either by their failure to capture the non-linear behaviour of air pollutants, or the incomplete understanding of the physical and chemical processes involved. The APPETISE project aims to develop and test the suitability of novel non-linear statistical methods to improve the ability to accurately forecast variations in air quality. It also aims to develop methods for handling missing data, which will have generic applications for other real data situations. The work is being carried out over a period of 2 years by a consortium from 9 institutions from 5 different European countries and is funded under the European Union Fifth Framework Programme. The project concentrates on 4 key pollutants; nitrogen oxides, particulates, ground level ozone and sulphur dioxide. Since it is likely that different methods and models will work best under different situations an ensemble approach will be utilised to improve the confidence held in any given prediction. The project will work towards the construction of a prototype air quality prediction and warning system the performance of which will be tested against existing systems.