Water Demand Forecasting Model Research Articles

A Water Demand Forecasting Model Based on Generative Adversarial Networks and Multivariate Feature Fusion

Accurate long-term water demand forecasting is beneficial to the sustainable development and management of cities. However, the randomness and nonlinear nature of water demand bring great challenges to accurate long-term water demand forecasting. For accurate long-term water demand forecasting, the models currently in use demand the input of extensive datasets, leading to increased costs for data gathering and higher barriers to entry for predictive projects. This situation underscores the pressing need for an effective forecasting method that can operate with a smaller dataset, making long-term water demand predictions more feasible and economically sensible. This study proposes a framework to delineate and analyze long-term water demand patterns. A forecasting model based on generative adversarial networks and multivariate feature fusion (the water demand forecast-mixer, WDF-mixer) is designed to generate synthetic data, and a gradient constraint is introduced to overcome the problem of overfitting. A multi-feature fusion method based on temporal and channel features is then derived, where a multi-layer perceptron is used to capture temporal dependencies and non-negative matrix decomposition is applied to obtain channel dependencies. After that, an attention layer receives all those features associated with the water demand forecasting, guiding the model to focus on important features and representing correlations across them. Finally, a fully connected network is constructed to improve the modeling efficiency and output the forecasting results. This approach was applied to real-world datasets. Our experimental results on four water demand datasets show that the proposed WDF-mixer model can achieve high forecasting accuracy and robustness. In comparison to the suboptimal models, the method introduced in this study demonstrated a notable enhancement, with a 62.61% reduction in the MSE, a 46.85% decrease in the MAE, and a 69.15% improve in the R2 score. This research could support decision makers in reducing uncertainty and increasing the quality of water resource planning and management.

An ecological stability-oriented model for the conjunctive allocation of surface water and groundwater in oases in arid inland river basins

Abstract With the continuous development of the population and social economy, the spatial and temporal distribution of water resources in arid inland river basins is severely uneven, and there is a sharp contradiction between agricultural water use and ecological water use. Irrational development and utilization of water resources has led to many problems, such as shrinking oases and drying lakes. To solve this problem, this study proposes a multiobjective, multiwater-source, ecological stability-oriented double-layer model for optimal allocation of water resources based on the large-scale system decomposition–coordination principle, the water balance principle, and a water supply and demand forecasting model. This model can resolve the contradiction between agricultural water use and ecological water use by optimizing and adjusting the crop planting structure, industrial structure, and the amount of water allocated to and groundwater level in each region and thereby achieve ecological stability and restoration of oases. The developed model was applied to the Heihe River Basin in an inland region of Northwest China. The long-term time series data of 2000–2016 were used to construct and calibrate the model. Finally, the practical ecological stability-oriented plan for conjunctive allocation of surface water and groundwater in different plan years was proposed. This model enriches the research results related to the conjunctive allocation of surface water and groundwater and provides a reference for the ecological restoration of oases in arid inland river basins.

Water use and demand forecasting model for coal-fired power generation plant in China

China is planning to expand its coal power generation to meet its energy demand and support the economic development. The current level of water use for thermal power generation is 8% of total water use, China is a water-stressed country which is facing many new challenges including climate change and population growth. China’s future coal power industry will add further pressures on already stressed water resources. This raised the key question on how the limited water resources can be managed to meet the demand of planned coal power expansion. A great level of understanding on the present status of water use and forecasting future demand in coal power plant is very important to answer this question. However, knowledge gap, data availability and accessibility are the major challenge in this regard. This paper attempts to improve the knowledge of the water demand in the coal power generation plant in China by using a simple water use model. Furthermore, a method is proposed to forecast future water demand in coal power plant. The proposed method is applied for forecasting water demand in Shaanxi coal power bases in Northern China under four scenarios. The results showed that the future water demand for Shaanxi coal power base will increase by 102–161% compared to current use under different scenarios in order to increase the production capacity by 206%. Adopting the optimum level of current status of water use, it is possible to limit the increase in water demand by 102% or 47.119 million-m3. It is expected that the finding of the study would help decision-making processes in water resources management in Chinese coal power generation.

A Probabilistic Short-Term Water Demand Forecasting Model Based on the Markov Chain

This paper proposes a short-term water demand forecasting method based on the use of the Markov chain. This method provides estimates of future demands by calculating probabilities that the future demand value will fall within pre-assigned intervals covering the expected total variability. More specifically, two models based on homogeneous and non-homogeneous Markov chains were developed and presented. These models, together with two benchmark models (based on artificial neural network and naïve methods), were applied to three real-life case studies for the purpose of forecasting the respective water demands from 1 to 24 h ahead. The results obtained show that the model based on a homogeneous Markov chain provides more accurate short-term forecasts than the one based on a non-homogeneous Markov chain, which is in line with the artificial neural network model. Both Markov chain models enable probabilistic information regarding the stochastic demand forecast to be easily obtained.

A Short-Term Water Demand Forecasting Model Using a Moving Window on Previously Observed Data

In this article, a model for forecasting water demands over a 24-h time window using solely a pair of coefficients whose value is updated at every forecasting step is presented. The first coefficient expresses the ratio between the average water demand over the 24 h that follow the time the forecast is made and the average water demand over the 24 h that precede it. The second coefficient expresses the relationship between the average water demand in a generic hour falling within the 24-h forecasting period and the average water demand over that period. These coefficients are estimated using the information available in the weeks prior to the time of forecasting and, therefore, the model does not require any actual calibration process. The length of the time series necessary to implement the model is thus just a few weeks (3–4 weeks) and the model can be parameterized and used without there being any need to collect hourly water demand data for long periods. The application of the model to a real-life case and a comparison with results provided by another model already proposed in the scientific literature show it to be effective, robust, and easy to use.

A Framework for Sustainable Urban Water Management through Demand and Supply Forecasting: The Case of Istanbul

The metropolitan city of Istanbul is becoming overcrowded and the demand for clean water is steeply rising in the city. The use of analytical approaches has become more and more critical for forecasting the water supply and demand balance in the long run. In this research, Istanbul’s water supply and demand data is collected for the period during 2006 and 2014. Then, using an autoregressive integrated moving average (ARIMA) model, the time series water supply and demand forecasting model is constructed for the period between 2015 and 2018. Three important sustainability metrics such as water loss to supply ratio, water loss to demand ratio, and water loss to residential demand ratio are also presented. The findings show that residential water demand is responsible for nearly 80% of total water use and the consumption categories including commercial, industrial, agriculture, outdoor, and others have a lower share in total water demand. The results also show that there is a considerable water loss in the water distribution system which requires significant investments on the water supply networks. Furthermore, the forecasting results indicated that pipeline projects will be critical in the near future due to expected increases in the total water demand of Istanbul. The authors suggest that sustainable management of water can be achieved by reducing the residential water use through the use of water efficient technologies in households and reduction in water supply loss through investments on distribution infrastructure.

Research on Industrial Water Demand Forecasting Model and its Application in China

Industrial water demand forecasting is needed for the effective planning and management of water supply systems. The paper first made impacting factor analysis of industrial water demand. Data analysis showed that there was a converse “U” type relationship between industrial water demand and industrial value added. There was a negative correlation relationship between industrial water demand and the recycling rate of it. With multiple regression method, industrial water demand forecasting model was established. In the supposed scenarios, we applied the model to predict Chinese industrial water demand in 2014 and 2015.The results were 141.04 billion m3 in 2014 and 137.79billion m3 in 2015.

An integrated water resources management strategy for Al-Ain City, United Arab Emirates

Abstract. Al-Ain is the second largest city in the Emirate of Abu Dhabi and the third in the UAE. Currently, desalination plants are the only source of drinking water in the city with an average daily supply of 170 MIG. Recently, Abu Dhabi Urban Planning Council (UPC) released Al-Ain 2030 Plan. Projects suggested in this plan, over and above the expected natural population growth, will certainly put additional stress on the water resources in the city. Therefore, Al-Ain city seems to be in urgent need for an integrated water resources management strategy towards achieving sustainable development. This strategy will contain three main components; namely, a Water Demand Forecasting Model (WDFM), a Water Budget Model (WBM), and a Water Resources Optimization Model (WROM). The main aim of this paper is to present the WBM that estimates all inflows and outflows to assess water resources sustainability in the city.

Water Demand Forecasting Model for the Metropolitan Area of São Paulo, Brazil

This work is concerned with forecasting water demand in the metropolitan area of Sao Paulo (MASP) through water consumption, meteorological and socio-environmental variables using an Artificial Neural Network (ANN) system. Possible socio-environmental and meteorological conditions affecting water consumption at Cantareira water treatment station (WTS) in the MASP, Brazil were analyzed for the year 2005. Eight model configurations were developed and used for the Cantareira WTS. The best performance was obtained for 12-h average of the input variables. The ANN model performed best with three times steps in advance. The hourly forecasting was obtained with acceptable error levels. Model results indicate an overall tendency for small errors. The proposed method is useful tool for water demand forecasting and water systems management. The paper is an important contribution since it takes into account weather variables and introduces some diagnostic studies on water consumption in one of the largest urban environments of the planet with its unique peculiarities such as anthropic affects on weather and climate that feeds back into the water consumption. The averaging is a low pass filter indeed and we used it to improve Signal to Noise Ratio (SNR).

Assessment of the Predictive Uncertainty within the Framework of Water Demand Forecasting by Using the Model Conditional Processor

Abstract An application of the Model Conditional Processor (MCP) is here presented to assess the predictive uncertainty in water demand forecasting related to water distribution systems. The MCP allows for the estimate of the forecast and its uncertainty through the elaboration of the forecasts provided by more than one deterministic forecasting model. The approach is applied to the hourly water consumptions of a town in the northern Italy and the results highlight its effectiveness, provided that the data set used for the parameterization is effectively representative of the accuracy of the real time water demand-forecasting model.

Quantification of Water Savings due to Drought Restrictions in Water Demand Forecasting Models

This paper presents a technique to quantify water savings due to implementation of water restrictions by adopting water restriction indexes as a continuous numerical predictor variable in regression analysis. The adopted modeling technique compares four methods: yearly base difference method, weighted average method, before and after method, and expected use method. These methods are applied to single and multiple dwelling residential sectors in the Blue Mountains region, Australia. In the study, three forms of multiple regression techniques are adopted: raw data, semi-log, and log-log. The model performances are evaluated by a number of statistics such as relative error, Nash-Sutcliffe coefficient, and percentage bias. Moreover, the potential of using the water restriction savings and water conservation savings as continuous predictor variables in the water demand forecasting model is investigated. The performances of different modeling techniques are evaluated using split-sample and leave-one-out cross-validation methods. The yearly base difference method is found to quantify the water savings more accurately in that the savings due to Level 1, Level 2, and Level 3 water restrictions are found to be approximately 9, 18, and 20%, respectively, for the single dwelling residential sector and approximately 4, 8, and 9%, respectively, for the multiple dwelling residential sector. The semi-log model coupled with yearly base difference method is found to perform the best in predicting water demand for both the single and multiple dwelling residential sectors with a relative error of about 3%.

Hybrid Water Demand Forecasting Model Associating Artificial Neural Network with Fourier Series

This paper addresses the problem of water demand forecasting for real-time operation of Water Supply Systems. The present study was conducted to identify the best fit model using hourly consumption data from the Water Supply System of Araraquara, Sao Paulo, Brazil. Like most recent research, it uses Artificial Neural Networks (ANNs) in view of their enhanced capability to match or even improve on the regression model forecasts. The ANNs used were the Multilayer Perceptron with the Backpropagation algorithm (MLP-BP), the Dynamic Neural Network (DAN2) and two hybrid ANNs. The hybrid models use the error produced by the Fourier series forecasting as input to the MLP-BP and DAN2, called respectively ANN-H and DAN2-H. Other inputs, such as the demand and weather variables (temperature and humidity) were selected based on autocorrelation and correlation analysis, respectively. The results are promising with the hybrid models in general performing better than the usual ANN models. DAN2-H, identified as the best model, produced a Mean Absolute Error of 2.25 L/s and 2.06 L/s for training and test set respectively, which represent about 8% of the average consumption.