1 Introduction

The increase in prices of energy sources based on oil, gas and coal in the world in addition to the problem of global warming has led countries around the world to increase reliance on renewable energy sources, especially hydropower. The Turkish government ambitiously wants to provide 30 % of the country’s electricity demand from renewable energy sources by 2023, (Melikoglu 2013).

There isn’t specified definition of small hydropower plant (SHP) in the world, for example SHP in Sweden is limits to 1.5 MW, while in china is up to 25 MW. The European Small Hydropower Association defines the small hydropower up to 10 MW (Lins and Laguna 2005). The spatial analysis for the selection an optimum site to set up a SHP is one of the complex issues because of the large number of influencing variables in the selection of that site.

In the design of SHP the most important problem facing the designer is ungauged basin with no flow duration curve at the SHP site. Topographic criteria is another factor facing the designer to assess the site location. Available head has a critical influence on the performance of the run of river hydropower projects. Searching for maximum head with a short water way as possible within a watershed is considered to maximize the power generating and reduce the cost of a plant. These two factors increase the difficulty to select the location of the project site. For any hydropower study on poorly gauged or ungauged watershed, a methodology has to be selected for determination of runoff. Integrated hydrological modeling is considered as one of the most modern technologies in the preparation of hydrological studies, especially with the tremendous development in the field of space survey using satellite technology.

Geographic information system (GIS) and hydrologic modeling are successfully utilized in the last two decades to develop a methodology to predict hydropower energy for any proposed hydropower project. The first applications of GIS for hydropower estimation were made in the United Kingdom (Punys et al. 2011). GIS application was used to a site selection of a small run-of-river hydropower project in Thailand by considering engineering, economic, environmental criteria and social impact (Rojanamon et al. 2009). In South Korea, a GIS-based methodology to identify suitable SHP sites by geospatial criteria was proposed (Yi et al. 2010). In this study, the location analysis methodology is focusing for searching for alternative locations rather than selecting the suitable site. Remote sensing and GIS with hydrologic modeling were applied to Solakli watershed, Eastern Black sea region for hydropower purpose. Basin boundary, area, (minimum, maximum and mean) elevation and slope information of the basin have been derived from the digital elevation model DEM (Coskun et al. 2010). SWAT2000 was applied to assess spatial availability of water resources and hydropower potentiality on Kopili River in North East, India by using the available metrological and stream flow data (Kusrea et al. 2010).

The results identified about 133 MW were available in 107 sites that reduction the carbon emission up to 125 t of CO2. SWAT model and spatial technologies were used to assess the runoff in the Mat River basin of Southern Mizoram, India. The results identified 33 hydropower potential sites within 147 km2. A total of 3039, 1127 and 805 kW can be harnessed at 50, 75 and 90 % dependability respectively (Pandey et al. 2015).

Applications of integrated hydrological modeling on hydropower assessment especially for ungauged basins become a powerful technique for flow simulation for design purposes. One of the powerful rainfall-runoff hydrological model on the world is HEC-HMS (US Army Corps of Engineers 2000). It is designed to integrate all the physical events of the rainfall-runoff process by simulation of runoff corresponding to different geographic area from large river watersheds to small river watersheds. The application of HEC-HMS with continuous soil moisture accounting model was applied to twelve basins in the north of Spain by using numerical tools, the model was developed to fill the incomplete daily rainfall series data during thirty three water years (García et al. 2008). Two hydrological models WBNM and HEC-HMS with GIS procedure were used to simulate runoff hydrographs of a small urban watershed located in Azzaba city in the North of Algera. The objective of the study is to test the effect of watershed size and time steps on the shape of runoff hydrograph (Laouachria and Mansouri 2015). HEC-GeoHMS model was used for flood forecasting in Masai and Wan’an watersheds in China (Oleyiblo and LI, 2010). Arc GIS 9.2 and HEC-HMS were used to calibrate and validate the flow data for the Attanagalu Oya watershed in Sri Lankan to generate long term flow data (Halwatura and Najim 2013). Impact of climate change on hydropower potential in Kulekhani hydropower project of Nepal was studied by Shresta et al. (2014).

The application of gene expression programming (GEP) is widely used in hydrologic engineering especially in the last decade. GEP was used for prediction of the monthly water level data of Van Lake, Turkey by Aytek, et al. (2014). GEP, linear genetic programming (LGP) and logistic regression were employed in order to forecast peak flood discharges for 543 ungauged sites across Turkey (Seckin and Guven 2012). LGP was used for predicting the peak monthly discharges from statistical downscaling and to evaluate the impacts of the global warming and climate change on determining of the flood discharge by considering different scenarios (Tofiq and Guven 2014), and suspended sediment yield estimation (Kisi and Guven 2010; Guven and Kisi 2011). GEP was used for modeling reference evapotranspiration using routing weather data from the tropical seasonally dry regions of West Africa in Burkina Faso by Traore and Guven (2013). GEP was also successfully applied in a wide range of subjects in water resources engineering (Guven 2009; Guven and Kisi 2011; Azamathulla et al. 2011; Azamathulla 2012; Kisi et al. 2012; Azamathulla and Jarrett 2013; Mehr et al. 2013,2014; Karbasi and Azamathulla 2015; Zorn and Shameldin 2015).

The objective of this study is to develop a computer program for run-of-river hydropower plant site assessment under different (hydrological, metrological and economical) criteria. The model is coupled with GEP to generate a flow duration curve at any location within a watershed for design purposes. The proposed program is capable to evaluate the optimum production of electricity for any watershed by selecting the optimum sites to install run of river hydropower plants and to assess the total potential of electricity that can be generated. In the literature, the earlier studies dealt with single watershed projects with unique criteria of optimum site location, however, this study divides the main watershed into sub-sections and searches for optimum site locations for the pre-defined multi-criteria. And finally, more than one hydropower project location are determined. This makes the novelty of the proposed methodology.

2 Study Area

The study area is Hurman River watershed, one of the major Ceyhan river tributaries in Turkey. The watershed is located in geographical coordinates as 360 32′ 00″–370 07′ 00″ E–380 22′ 00″–380 44′ 00″ N. The Hurman River watershed is mountainous with narrow deep valleys, that is suitable for hydroelectric generation, especially by run-of-river HPP type. Figure 1 shows the location of the study area. The main soil texture for Hurman River watershed is clay < 80 %, so the hydrological soil type for Hurman River is “C” according to soil conservation service (SCS 1986). The dominant land use in the study area are meadows, pastures and arable land.

Fig. 1
figure 1

Location of the study area

Full size image

3 Methodology

3.1 Watershed Delineation

The watershed modeling system (WMS) version 8.3 is selected as integrated hydrological model in this study. Land use and soil type for the Hurman River watershed were input to the model as GIS files. Digital elevation model (ASTER GDEM worldwide elevation data 1 arc-second resolution) was analyzed by TOPAZ module in WMS to calculate flow direction and flow accumulation to extract the stream network. A drainage module in WMS was used to delineate the watershed boundary (see Fig. 2). Hurman River watershed area is 940.49 km2 upstream Tanir flow station, and watershed slope is 0.18 m/m.

Fig. 2
figure 2

Hurman River watershed boundary

Full size image

3.2 Runoff Simulation

The main objective of using the WMS is to create a conceptual model to simulate the runoff by one of the rainfall-runoff models in WMS. Hydrological model HEC-HMS was selected to complete the rainfall-runoff simulation in this study. The model includes separate models to represent each component of the runoff process which includes computation of losses, runoff volume, direct runoff and base flow. The initial loss and constant rate method was selected to simulate the infiltration loss, Clark unit hydrograph was used for direct runoff transformation and the constant monthly varying value was used to estimate the base flow. Four parameters were identified to calibrated: Canopy interception storage, constant loss rate, Clark storage coefficient and Clark time of concentration. Calibration was completed automatically using optimization manager in HEC-HMS model to minimize the bias between observed and simulated stream flow data. Sum squared residuals was selected as objective function and the Univariate Gradient search algorithm was selected to minimize the objective function. The relative changes in the model output with respect to change in the model input parameters were examined by sensitivity analysis.

3.3 Flow Duration Curve Generation

The important hydrological parameter for design of run-of-river HPP is flow duration curve (FDC). The FDC method is used in individual project and it is not applicable for more than one projects. The FDC is developed using available stream flow record including at least 30 years of monthly discharge data. It is necessary to find a FDC for each proposed HPP. Generating FDC for any location within a watershed is the key for design of HPP project. Many statistical approaches for FDC analysis were developed by researchers in the world (Booker and Snelder 2012; Liucci et al. 2014; Longobardi and Villani 2013; Mendicino and Senatore 2013; Niadas 2005) to solve the predication of FDC problem by using statistical methods. In this study, a new methodology to solve this problem is developed. The watershed drainage area was divided to sub-basins by adding outlets at each 1 km distance along the stream network. Daily rainfall data were used to simulate the runoff and generate streams flows at each sub-basins outlet for FDC analysis. GEP is used to analyze the simulated FDCs to develop a mathematical expression to generate FDC at any location along the basin’s stream network.

3.3.1 Gene Expression Programming (GEP)

GEP is an extension to the genetic programming (GP) that was invented by Ferreira (2001). GP was used in the field of hydrologic engineering more widely in the last decade. The basic difference between GEP and GP is the representation of computer program. GEP is a symbolic regression algorithm to form a mathematical function that fits a set of data. The GEP is using a genetic, evolution algorithm and symbolic regression to fit the data to optimum mathematical function. GEP program is encoded as linear strings of fixed length (chromosomes or genomes) and then expressed as nonlinear entities of different shapes and sizes (Ferreira 2001, 2006).

The expressions for the problems solution are generated automatically in GEP by coding the expression as a tree structure with nodes (function) and leaves (terminal). A fitness function is used to evaluate the generated candidates and a modification evaluation cycle is applied for the higher performance candidates and then re-evaluated until achieved an optimum solution. The GEP code is very simple, the relation between the symbols of the nodes and the chromosome is represented in the trees in one to one relation. The genetic expression relation genes are composed of two parts (head and tail). The head includes variable and constants that are used to encode a mathematical expression and the tail includes terminal symbols. In this study, GEP was employed to perform a symbolic regression to predict a flow duration curve at any site within the Hurman watershed based on the results of the rainfall-runoff model.

3.4 Hydropower Computation Program

A computer program was developed by Visual Basic Programming in Microsoft Excel interface for hydropower potential computation. Computation includes hydrological and spatial analysis of the watershed under different criteria (hydrological, spatial and economic). The program was designed on the basis of the ability to select the best solution from several alternative criteria. The proposed program requires the following inputs:

  1. 1-

    A file that contains the coordinates of the stream network (UTM-coordinates), drainage area upstream each coordinates and the elevation for each point along the streams network.

  2. 2-

    A mathematical expression for flow duration curve generation.

  3. 3-

    Design criteria which includes minimum distance between the two projects, maximum distance from weir to power house, total efficiency and power or head.

The proposed hydropower program analyzes the stream network data and selects the best locations to install the run-of-river HPP projects under different criteria. Two major locations in run-of-the river HPP projects are the location of weir and the location of powerhouse. The hydropower program searching algorithm starts from the basin outlet towards the upstream direction. If the design focuses on the installed power value of the project, the program starts with an initial point as power house and then searches for the weir location upstream the power house along the stream. The program estimates upstream drainage area and discharge at this proposed weir site, and then calculates the head and power for the project. If the computed power value matches the input power criteria the location of the weir is selected and the program starts to find another location to install a new plant after matching the space distance criteria between the two adjacent projects.

If the design focuses on head criteria, the same searching procedure is applied except the program focusing on the head rather than the power. The program uses the elevation of power house location as reference point and then checks the difference in elevation between the proposed power house location elevation and upstream points along the stream to match the required difference head. If the computed head value matches the head criteria value, the location of the weir is selected and the program starting to find another location to install a new project after matching the space distance criteria between the two projects.

Figure 3 exhibits the searching algorithm for the hydropower program site selection. The hydropower program is also able to analyze the whole watershed according to economic criteria (total length of penstocks for all proposed projects, total distances between the power house and the weir for all proposed projects and power generated from all the projects) according to the head or the installed power criteria to get the optimum HPP locations.

Fig. 3
figure 3

Searching algorithm for selection of run of river project

Full size image

The power is estimated by product of head, flow, unit weight of water and total project efficiency.

$$ P={\gamma}_wQHe $$
(1)

where, P is power (kW), γw is unit weight of water (9.81 kN/m3), Q is design flow (m3/s), H is static head from intake to powerhouse (m) and e is total project efficiency. The program is used the FDC mathematical expression that developed by using the gene expression programming to calculate the design flow according to the required design flow duration.

The flow chart of the research methodology is given in Fig. 4. The program outputs included, table that contains the coordinates of weir and powerhouse in UTM, design discharge, design head, upstream drainage area for the weir and powerhouse, power in kW, distance from powerhouse to weir location and penstock length and a map containing the locations of the weir and powerhouse for all proposed HPPs.

Fig. 4
figure 4

Flow chart of proposed methodology

Full size image

4 Results and Discussion

The rainfall-runoff model parameters values are estimated by calibration process based on the goodness of fit between the simulated and observed flows data. Daily rainfall data from October, 1, 1993 to September, 1, 1996 were used to simulate the runoff by HEC-HMS and calibrated with the daily flow time series for the same period. The results showed that the canopy interception storage and Clark time of concentration parameters are not sensitive parameters and the optimized values for them are 4 mm and 14.47 h, respectively. But the runoff simulation is sensetive to constant loss rate and Clark storage coefficient parameters. The objective function sensitivity is −2.54 and −0.07 for constant loss rate and Clark storage coefficient, respectively, and the optimized values for them are 0.528 mm/h. and 24.081 h., respectively. The validation of the model was made using the simulated and observed discharge data from January, 1, 1990 to September, 30, 1992. The visual checking of the simulated and observed streamflows of the calibration and the validation periods are exhibited in Fig. 5. A good agreement between the simulated and the observed values is apparent, because there is a similar trend between simulated and observed stream flows.

Fig. 5
figure 5

Comparison of the simulated and the observed daily stream flows

Full size image

The results proved that the accuracy of HEC-HMS model with the high values of NSE as 0.95 and 0.92, and R2 as 0.94 and 0.92 for the calibration and the validation periods, respectively. The mean error (ME) is 0.19 and 0.28 m3/s, and the root mean square error (RMSE) is 0.97 and 0.88 m3/s for the calibration and the validation periods respectively. The main objective of the rainfall-runoff model in this study is to generate the daily flow series at each 1 km distance along the streams network and estimate the flow duration curves at these locations. Hurman River watershed is divided into 130 sub-basins, and 130 FDCs were generated with the simulated daily discharges. Table 1 shows the sample of the resulted FDCs for some subbasins in the Hurman River watershed.

Table 1 Samples of the FDCs at different sites along the streams network
Full size table

A mathematical expression to predict the FDC at any site within the Hurman watershed was developed by using GEP. The 130 FDCs calculated from the results of rainfall–runoff model were used to develop the GEP mathematical expression. The FDC is generated at any site within the watershed by three independent variables (drainage area, elevation of the outlet and flow duration). The optimum expression model was achieved by applying the parameters summarized in Table 2. GEP modeling revealed Eq. (2) with R2 = 0.97, MSE = 0.195 and MAE = 0.244.

$$ Q=\left(\frac{0.103\times A}{D}\right)+\left(\frac{5.154\times \left(A-D\right)}{Z}\right)+0.259 $$
(2)

where, A is drainage area (km2), D is flow duration (5, 10, ……… 95) and Z is outlet elevation (m).

Table 2 GEP model settings
Full size table

Two criteria are used to assess the potential hydropower for Hurman River by the hydropower program: i) power criteria (500, 1000, 1500 and 2000 kW) for installed capacity, and ii) the head criteria (20, 30, 40 and 50 m). Both criteria are used to estimate the optimum sites to install the run-of-the river HPP projects. Three levels of dependability design discharge (30, 40 and 50 %) were used with both criteria. The overall plant efficiency is taken as 80 %.

The minimum distance between the current project and the next project is taken as 500 m and the maximum distance between weir location and powerhouse site is 3500 km (Kusrea et al. 2010). The total generated power with total penstock length and the distance between the weir and powerhouse for the two criteria (power and head) are summarized in Table 3. The analysis of the results showed that the 1000 kW as design power criteria have given an optimum (total electricity production, total length of penstocks and total distance between the weir and powerhouse and number of projects) for the three levels of design flow duration in comparison with the other alternative.

Table 3 Summary of the hydropower program outputs for different power and head criteria
Full size table

The results in Table 3 indicate that head requirement as input criteria is not an efficient choice to select the optimum sites to install run-of-the-river HPP projects within the study area, because the number of project increases, total length of penstocks and total distance from weir to powerhouse also increases dramatically with low power that is generated in comparison with the number of projects. Figure 6 shows the projects distribution for 20 m-head-design criteria for 30 % design discharge duration. Table 4 summarizes the program outputs for 1000 kW-design power requirement for each proposed site for 30 % duration design discharge and 80 % total project efficiency as the program outputs example

Fig. 6
figure 6

HPP project distribution for 20 m-head criteria

Full size image
Table 4 Sample of hydropower program output for 1000 kW power criteria
Full size table

5 Conclusion

A computer program incorporating integrated hydrological modeling and GEP technology is suggested in this study. The proposed methodology developed in this study for HPP site assessment could overcome all the disadvantages associated with traditional methods.

The high values of the statistical indicators, especially the Nash–Sutcliffe efficiency and R2 (higher than 90 %) proved the efficiency of selected the HEC-HMS model for rainfall-runoff modeling.

GEP was used to develop an optimum mathematical expression predicting the FDC at any site along the stream network within the Hurman watershed, which gave high correlation (R2 = 0.97). The mathematical expression developed by GEP make the generation FDC easier than traditional statistical techniques used for this purpose.

The results of hydropower potential computations showed that the power criteria for the design of run-of-the-river HPP project site selection in Hurman watershed outperformed against the head criteria, because the number of projects for power criteria is less than the number of projects for head criteria for the same total hydroelectricity to be generated. The 1000 kW as input power criteria revealed optimum hydroelectricity generation for different design discharge durations.

The proposed hydropower computation program was designed to assess the potential hydropower for any watershed just by the limited inputs corresponding to the watershed under study. The proposed methodology is believed to be a strong alternative to the existing ones.