Search

Predicting the price of electricity is crucial for the operation of power systems. Short-term electricity price forecasting deals with forecasts from an hour to a day ahead. Hourly-ahead forecasts offer expected prices to market participants before operation hours. This is especially useful for effective bidding strategies where the bidding amount can be reviewed or changed before the operation hours. Nevertheless, many existing models have relatively low prediction accuracy. Furthermore, single prediction models are typically less accurate for different scenarios. Thus, a hybrid model comprising least squares support vector machine (LSSVM) and genetic algorithm (GA) was developed in this work to predict electricity prices with higher accuracy. This model was tested on the Ontario electricity market. The inputs, which were the hourly Ontario electricity price (HOEP) and demand for the previous seven days, as well as 1-h pre-dispatch price (PDP), were optimized by GA to prevent losing potentially important inputs. At the same time, the LSSVM parameters were optimized by GA to obtain accurate forecasts. The hybrid LSSVM-GA model was shown to produce an average mean absolute percentage error (MAPE) of 8.13% and the structure of this model is less complex compared with other models developed in previous studies. This is due to the fact that only two algorithms were used (LSSVM and GA), with the load and HOEP for the week preceding the forecasting hour as the inputs. Based on the results, it is concluded that the proposed hybrid algorithm is a promising alternative to produce good electricity price forecasts.

Predicting electricity price has now become an important task for planning and maintenance of power system. In medium term forecast, electricity price can be predicted for several weeks ahead up to a year or few months ahead. It is useful for resources reallocation where the market players have to manage the price risk on the expected market scenario. However, researches on medium term price forecast have also exhibit low forecast accuracy. This is due to the limited historical data for training and testing purposes. Therefore, an optimization technique of Bacterial Foraging Optimization Algorithm (BFOA) for Least Square Support Vector Machine (LSSVM) was developed in this study to provide an accurate electricity price forecast with optimized LSSVM parameters and input features. So far, no literature has been found on feature and parameter selections using the LSSVM-BFOA method for medium term price prediction. The model was examined on the Ontario power market; which is reported as among the most volatile market worldwide. Monthly average of Hourly Ontario Electricity Price (HOEP) for the past 12 months and month index are selected as the input features. The developed LSSVM-BFOA shows higher forecast accuracy with lower complexity than the existing models.

Predicting electricity price has now become an important task in power system operation and planning. An hour-ahead forecast provides market participants with the pre-dispatch prices for the next hour. It is beneficial for an active bidding strategy where amount of bids can be reviewed or modified before delivery hours. However, only a few studies have been conducted in the field of hour-ahead forecasting. This is due to most power markets apply two-settlement market structure (day-ahead and real time) or standard market design rather than single-settlement system (real time). Therefore, a hybrid multi-optimization of Least Square Support Vector Machine (LSSVM) and Bacterial Foraging Optimization Algorithm (BFOA) was designed in this study to produce accurate electricity price forecasts with optimized LSSVM parameters and input features. So far, no works has been established on multistage feature and parameter optimization using LSSVM-BFOA for hour-ahead price forecast. The model was examined on the Ontario power market. A huge number of features were selected by five stages of optimization to avoid from missing any important features. The developed LSSVM-BFOA shows higher forecast accuracy with lower complexity than most of the existing models.

Electricity price forecasting has become an important part of power system operation and planning. In a pool- based electric energy market, producers submit selling bids consisting in energy blocks and their corresponding minimum selling prices to the market operator. Meanwhile, consumers submit buying bids consisting in energy blocks and their corresponding maximum buying prices to the market operator. Hence, both producers and consumers use day ahead price forecasts to derive their respective bidding strategies to the electricity market yet reduce the cost of electricity. However, forecasting electricity prices is a complex task because price series is a non-stationary and highly volatile series. Many factors cause for price spikes such as volatility in load and fuel price as well as power import to and export from outside the market through long term contract. This paper introduces an approach of machine learning algorithm for day ahead electricity price forecasting with Least Square Support Vector Machine (LS-SVM). Previous day data of Hourly Ontario Electricity Price (HOEP), generation’s price and demand from Ontario power market are used as the inputs for training data. The simulation is held using LSSVMlab in Matlab with the training and testing data of 2004. SVM that widely used for classification and regression has great generalization ability with structured risk minimization principle rather than empirical risk minimization. Moreover, same parameter settings in trained SVM give same results that absolutely reduce simulation process compared to other techniques such as neural network and time series. The mean absolute percentage error (MAPE) for the proposed model shows that SVM performs well compared to neural network.

Fuzzy Expert Systems have been used to solve complex problems efficiently in the case where information available is in descriptive form rather than quantitative number. This study has aimed to use Fuzzy expert systems to estimate the labor production rates through incorporating the influence of qualitative and quantitative factor. Production rate values of concreting of columns and their influential factors have been collected from questionnaire survey. Overall ten influential factors of qualitative and quantitative nature are selected for collecting the data during questionnaire survey based on the Likert scale of 1 to 5. Fuzzy expert system developed in this study has been compared with the two previously used Fuzzy expert systems for productivity estimation. Performance of the previous systems and system developed in this study has been compared by calculating Root Mean Square Error. The findings revealed that system developed in the study gives high linguistic and numerical accuracies as compare to the previous systems with least Root Mean Square Error. Hence, the developed Fuzzy expert system can be used reliably for estimating labor productivity by the construction Industry.

Magnetic sensor is a relatively new method to collect time-resolved partial discharge (PD) signals in XLPE cables. This paper proposes a simple yet effective method to recognize patterns of PD signals obtained from the magnetic sensor. The technique consists of wavelet transformation to de-noise the signals, statistical analysis to extract features and multi-layer perceptron back propagation (MLP BP) neural network to classify different types of PD signals. The result is elaborated in this paper.

This paper presents a concept of predicting lightning with the data from radiosonde and a more reliable dataset of lightning occurances from Tenaga Nasionl Berhad Research Sdn Berhad (TNBR). The location of interest in this research is Kuala Lumpur International Airport (KLIA). The engine used for prediction is of a Neural Network Back Propogation type (ANN-BP). The initial results show that the combination of datasets and engine are workable, however the prediction results seem to be more biased towards lightning days as compared to non-lightning days.

This paper discusses the feasibility of operating an artificial neural network - back propagation (ANN-BP) in a handheld device. Comparisons were done between operating an ANN-BP on a desktop versus a handheld device in duration of time and accuracy. It was found that by implementing an ANN-BP on a handheld device, the speed was slower as compared to running on a desktop. The accuracy of results did not differ much based on the device the ANN-BP was executed upon. As a conclusion, the viability of using a handheld device to run ANN-BP is not practical with current processor speeds as the processing time is approximately 300 times longer than that of a desktop.

Artificial intelligence is used to predict the onset of diabetes based on data measured from Pima Indians. This research is comparing the results gained from using same artificial neural networks- back propagation (ANN-BP) engine for 2 differently prepared data. The first data set consists of the entire data set which is cross validated, while the second dataset is segregated into 2 groups using Kohonen Self Organizing Maps (SOM) which are then cross validated. Splitting the files prior to implementing the cross validation improves the general accuracy of the ANN-BP whereby the positively predicted diabetes cases percentage increased from 72% to 99%. Meanwhile the prediction of the negative diabetic cases percentage increased from 80% to 97%.

In this new informative era, data and information is the most important asset to organizations. A large amount of money and manpower have been spent in data gathering, data entry, and storage every year. In Malaysia, data gathering is still largely done through manually filled forms. This data is then entered and stored into databases in government and private organizations manually. Such mode of data entry and storage requires a lot of manpower and is time consuming. At the Centre for Artificial Intelligence and Robotics (CAIRO) of Universiti Teknologi Malaysia, research is being carried out to design a system for automated data entry of handwritten-filled forms. The system consists of a high-speed scanner with an autofeeder and a computer. In the first phase, software was developed that allows the user to design the template of existing forms such that only the regions of interests are captured. The next phase involved the design of software to capture handwritten characters in the regions of interest through the scanned forms. Image processing techniques are then used to filter and improve the image of the scanned handwritten characters before they are recognized using a neural network algorithm. Once the characters are identified and verified, they are automatically stored into a database. This system can be used efficiently in many organizations that involve gathering and processing of a large number of data such as the National Registration Department, Survey Research Malaysia, Kementerian Pendidikan and Lembaga Hasil Dalam Negeri.