Solar irradiance forecasting based on direct explainable neural network

Wang, Huaizhi; Cai, Ren; Zhou, Bin; Aziz, Saddam; Qin, Bin; Voropai, N.I.; Gan, Lingxiao; Barakhtenko, Evgeny

doi:10.1016/j.enconman.2020.113487

Cited by 72 publications

(28 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It has been discussed by D'Amico et al 50 and Wang et al 51 that the number of hidden neurons should be less than the number of inputs used in the neural network model. Also, the number of hidden layers ranges from 1 to 3.…”

Section: Optimal Bra‐based Lf Strategy Incorporating Pccmentioning

confidence: 99%

Optimal BRA based electric demand prediction strategy considering instance‐based learning of the forecast factors

Waseem

Liu

Zhang

et al. 2021

Int Trans Electr Energ Syst

View full text Add to dashboard Cite

With the grid's evolution, the end-users demand becomes more vital for demand side management (DSM). Accurate load forecasting (LF) is critical for power system planning and using advanced demand response (DR) strategies.To design efficient and precise LF, information about various factors that influence end-users demand is required. In this paper, the impact of different factors on electrical demand and capacity of climatic factors existence and their variation is discussed and analysed. The Pearson correlation coefficient (PCC) is utilized to express the degree of electric demand correlation with metrological and calendar factors. Then, the optimal-Bayesian regularization algorithm (BRA) based on ANN for LF is presented. The effect of the number of neurons in hidden layers on output is observed to select the most appropriate option. Additionally, heating degree days (HDDs) and cooling degree days (CDDs) indices are investigated to consider the impact of air conditioners' (ACs) loads in different seasons. Case studies on data from Dallas, Texas, USA, are used to

show abstract

Section: Optimal Bra‐based Lf Strategy Incorporating Pccmentioning

confidence: 99%

Optimal BRA based electric demand prediction strategy considering instance‐based learning of the forecast factors

Waseem

Liu

Zhang

et al. 2021

Int Trans Electr Energ Syst

View full text Add to dashboard Cite

show abstract

“…Wang et al [16] and Sethi and Kantardzic [17] Neural networks/deep learning approaches have not seen sufficient adoption, despite growing interest in them, due to their complex black-box nature and lack of explainability and interpretability…”

Section: Source Issue or Concernmentioning

confidence: 99%

“…Wang. et al [16] Explainability is of great importance, therefore, proposed a new approach through direct explainable neural networks that can provide further insights in the input-output relationship to assist in result interpretation and model explanation.…”

Section: Source Issue or Concernmentioning

confidence: 99%

“…Despite the fact that several well-known metrics and procedures for forecasting performance assessment are presented in the literature, most studies are limited to assessing performance of the entire evaluation period, while some further detail at how performance varies with month or season [16,[29][30][31], but comparatively few studies provide details at the type-of-day [15] level or for specific values or ranges for K t [14,32]. This analysis is of interest, because as Figure 1 shows, the same method can yield very different results as a function of the variability of the location under analysis.…”

Section: Present Work and Scientific Contributionsmentioning

confidence: 99%

See 1 more Smart Citation

Short-Term Deterministic Solar Irradiance Forecasting Considering a Heuristics-Based, Operational Approach

2021

View full text Add to dashboard Cite

Solar energy is an economic and clean power source subject to natural variability, while energy storage might attenuate it, ultimately, effective and operationally feasible forecasting techniques for energy management are needed for better grid integration. This work presents a novel deterministic forecast method considering: irradiance pattern classification, Markov chains, fuzzy logic and an operational approach. The method developed was applied in a rolling manner for six years to a target location with no prior data to assess performance and its changes as new local data becomes available. Clearness index, diffuse fraction and irradiance hourly forecasts are analyzed on a yearly basis but also for 20 day types, and compared against smart persistence. Results show the proposed method outperforms smart persistence by ~10% for clearness index and diffuse fraction on the base case, but there are significant differences across the 20 day types analyzed, reaching up to +60% for clear days. Forecast lead time has the greatest impact in forecasting performance, which is important for any practical implementation. Seasonality in data gaps or rejected data can have a definite effect in performance assessment. A novel, comprehensive and detailed analysis framework was shown to present a better assessment of forecasters’ performance.

show abstract

“…Due to the wide layout of the power grid, the large-scale deployment of smart meters should consume a lot of resources. In order to save the energy consumption of distributed terminal nodes, and reduce the non-essential data transmission, it is necessary to study modern data mining technology, in integration with machine learning algorithms (Wang et al, 2020;Li Z. et al, 2021). The application of indirect data anomaly detection as well as some preprocessing and analyzing technologies is much necessary to achieve the online detection of power theft.…”

Section: Introductionmentioning

confidence: 99%

Electricity Theft Detection in Power Consumption Data Based on Adaptive Tuning Recurrent Neural Network

et al. 2021

View full text Add to dashboard Cite

Electricity theft behavior has serious influence on the normal operation of power grid and the economic benefits of power enterprises. Intelligent anti-power-theft algorithm is required for monitoring the power consumption data to recognize electricity power theft. In this paper, an adaptive time-series recurrent neural network (TSRNN) architecture was built up to detect the abnormal users (i.e., the electricity theft users) in time-series data of the power consumption. In fusion with the synthetic minority oversampling technique (SMOTE) algorithm, a batch of virtual abnormal observations were generated as the implementation for training the TSRNN model. The power consumption record was characterized with the sharp data (ARP), the peak data (PEA), and the shoulder data (SHO). In the TSRNN architectural framework, a basic network unit was formed with three input nodes linked to one hidden neuron for extracting data features from the three characteristic variables. For time-series analysis, the TSRNN structure was re-formed by circulating the basic unit. Each hidden node was designed receiving data from both the current input neurons and the time-former neuron, thus to form a combination of network linking weights for adaptive tuning. The optimization of the TSRNN model is to automatically search for the most suitable values of these linking weights driven by the collected and simulated data. The TSRNN model was trained and optimized with a high discriminant accuracy of 95.1%, and evaluated to have 89.3% accuracy. Finally, the optimized TSRNN model was used to predict the 47 real abnormal samples, resulting in having only three samples false predicted. These experimental results indicated that the proposed adaptive TSRNN architecture combined with SMOTE is feasible to identify the abnormal electricity theft behavior. It is prospective to be applied to online monitoring of distributed analysis of large-scale electricity power consumption data.

show abstract

Solar irradiance forecasting based on direct explainable neural network

Cited by 72 publications

References 35 publications

Optimal BRA based electric demand prediction strategy considering instance‐based learning of the forecast factors

Optimal BRA based electric demand prediction strategy considering instance‐based learning of the forecast factors

Short-Term Deterministic Solar Irradiance Forecasting Considering a Heuristics-Based, Operational Approach

Electricity Theft Detection in Power Consumption Data Based on Adaptive Tuning Recurrent Neural Network

Contact Info

Product

Resources

About