The role of energy audits in irrigated areas. The case of ‘Fuente Palmera’ irrigation district (Spain)

In recent years, energy consumption for irrigation has grown rapidly. Actually, nowadays energy represents a significant percentage on the total water costs in irrigation districts using energy to pressurize water. With the aim of improving energy efficiency in the Fuente Palmera irrigation district, was applied the protocol for conducting energy audits in irrigation districts developed by Spanish Institute for Diversification and Energy Savings (IDAE). The irrigated area organized in two independent sectors according to a homogeneous elevation criterion is analyzed and simulated. The potential energy savings derived from this measure was evaluated. For this purpose, a model based on the hydraulic simulator EPANET has been carried out. Its energy demand was estimated in 1,360 kWh ha –1 and its overall energy efficiency in 56%. The district was globally classified in group C (normal). Results show potential energy savings of up to 12% were obtained when the network was divided in sectors and farmers organized in two irrigation shifts. Further energy savings could be achieved by improving the hydraulic structures, such as the pumping station or the network layout and dimensions.


Introduction
Energy consumption has continuously grown in the last decades worldwide. Because of increased energy costs, energy scarcity and energy related pollution, in the last years all economic sectors have intensified their efforts to improve energy use efficiency.
Irrigated agriculture is one of the sectors that have experienced a notable increase in energy use during recent years. In Spain, this increase is mainly due to irrigation modernization programs, where open channel distribution systems are being replaced by on-demand pressurized networks. These measures have succeeded in improving irrigation efficiency. However, they have led to a significant increase in energy consumption. Actually, Corominas (2009) reported that while water use per hectare has been reduced by 23% since 1950, energy demand has been increased by 670%.
In the last years, research efforts were devoted to the improvement of irrigation efficiency by means of benchmarking techniques and performance indicators (Malano and Burton, 2001;Rodríguez-Díaz et al., 2008). Recent studies have focused on the need to improve water and energy efficiency at the same time (Pulido-Calvo et al., 2003;ITRC, 2005;Moreno et al., 2007Moreno et al., , 2009Vieira and Ramos, 2009;Daccache et al., 2010). Other research works have highlighted that energy savings of up to 20% could be achieved by introducing minor changes in the irrigation district's management practices (Rodríguez-Díaz et al., 2009;Jiménez-Bello et al., 2010;Moreno et al., 2010).
An aggravating factor is that, due to the liberalization of the Spanish Electricity Market, on 1 st January 2008 the special tariffs for irrigation disappeared and now irrigation districts have to use the same tariffs as the rest of the industry. During the months of June and July, when the peak of the irrigation demand is concentrated, most of the daily hours are included in periods of expensive tariffs.
In the irrigated agriculture sector, two main measures are proposed in these plans. The first measure is the migration from sprinkler to drip irrigation systems, and the second is the improvement of efficiency through energy audits in irrigation districts. In this way, the Spanish Institute for Diversification and Energy Saving (IDAE, 2008) proposed a protocol for conducting energy audits, which evaluates efficiency by means of water and energy performance indicators. This methodology allows detection of inefficiencies and provides information on the required improvement actions (Abadía et al., 2008).
The initial set of indicators proposed by IDAE has been extended in recent studies (Abadía et al., 2009) and applied to several irrigation districts in other Spanish regions (Abadía et al., 2007;Córcoles et al., 2008). In other works, alternative protocols for energy audits were proposed, following similar methodologies and outputs (Ederra and Larumbe, 2007).
In this work, water and energy use were evaluated at the Fuente Palmera irrigation district, using the IDAE protocol for energy audits. Alternative management measures were adopted for achieve energy savings such as sectoring the network in several sectors according to homogeneous group. This measure was analyzed on the hydraulic model EPANET and the potential energy savings derived from this measure were evaluated.

Study area
The Fuente Palmera irrigation district, located in Córdoba (Southern of Spain) has a total irrigated area of 5,611 ha (Fig. 1). The climate in the region is predominantly Mediterranean, with rainfall concentrated mainly in autumn and spring, and dry spells in summer. The average annual rainfall in the area is 550 mm, and the average temperature is 17.9°C. In the analyzed irrigation season (2007) annual rainfall was 523 mm and potential evapotranspiration was 1,323 mm (Carrillo-Cobo, 2009). Consequently 2007 can be considered representative of the average year (Rodríguez-Díaz, 2003). There is a wide range of crops in the district, with cereals, citrus and olives trees covering more than 60% of the area (Carrillo-Cobo, 2009).
Irrigation water is diverted from the Guadalquivir River and conveyed to an elevated reservoir through a first pumping station. At the reservoir there is a booster pumping station feeding 85 hydrants. The pressurized collective network has a total length of 45 km. It was designed to supply 1.2 L s -1 ha -1 arranged on demand, with a minimum pressure head of 30 m at every hydrant. The topography is quite steep, and hydrant elevation ranges from 86 to 165 m.
The booster pumping station (altitude of 113.9 m) is equipped with six horizontal centrifugal pumps of 1,825 kW, two of 495 kW and one of 540 kW, equipped with variable speed drives, being the total installed power 2.2 kWha -1 . The pumps are activated sequentially according to manometric regulation. The pumping station has a telemetry system which records hydraulic parameters (pressure head and pumped flow) and electric intensity every minute.
Although most of the district is currently irrigated by drip irrigation systems, Fuente Palmera was originally designed for sprinkler irrigation, with higher pressure requirements.

Hourly irrigation water demand patterns
To classify the irrigation demand in homogeneous groups, the non-hierarchical clustering algorithm K-means (Cuesta, 2001) was used. The objective of this algorithm is to minimise variance within clusters and maximise variance between clusters (Jain, 2000). Its main limitation is that the number of clusters has to be fixed a priori.
The K-means algorithm is based on the minimization of a performance index, which is the sum of the squared distances of all the elements within the cluster to the centroid of the cluster. To measure the distance between elements, the Euclidean distance has been used (Rodríguez-Díaz et al., 2008).
Using the recorded flow at the pumping station, one vector was created for every week including the 24 ratios of the hourly average pumped water to daily average pumped water. Following this procedure, a daily water demand pattern was created for every week. Then, the K-means algorithm was applied to these demand patterns, corresponding to the whole irrigation season, and then they were grouped into homogeneous clusters. The analysis was performed for two, three and four clusters.

Performance curves
The information collected at the pumping station was used to energetically characterize the irrigation district. This analysis was carried out in two main steps. The f irst step was the generation of the frequency distribution of demanded flow for the irrigation season, at 50 L s -1 intervals.
The second step was the analysis of the hydraulic performance of the pumps installed at the pumping station. Reliable data on flows, pressures and power recorded at the pumping station were used to establish the pumping station characteristics curves (flow and pressure head; flow and power; flow and performance). Pumping performance (η) was determined as: where γ is the water specific weight (9,800 N m -3 ), F is the demanded flow rate (m 3 s -1 ), H is the pressure head at the pumping station (m) and W is the power consumption recorded at the pumping station (W).

Energy efficiency indicators
Energy indicators were selected from those suggested by the IDAE (2008) for conducting energy audits in irrigated areas. In this work indicators have been classified in four groups: -Descriptor indicators, informing about water use and irrigated areas within the irrigation district.
-Power indicators, analyzing power requirements. They also allow comparison between contracted power and recorded power use. These indicators can be used to assess whether the current energy contract meets the district's power demand, and to provide information to optimize the contract.
-Energy indicators, analyzing energy consumed for pumping and energy costs.
-Efficiency indicators. This is the most important group of indicators. They provide an energy assessment and a district classification. These are the indicators included in this group: • Energy dependency rate (EDR): Volume of water entering the system • Energy change index (ECI): Volume of water entering the system where V i and H i are the volume and pressure head supplied by pumping. Thus, ECI represents the average pressure head. • Pumping energy efficiency (PEE): where W s is power given to the water flow and W a is the electrical power consumed, determined as: where V is the voltage of each pump (V); I is the intensity (A) and cos ϕ is the pump power-factor. Ws was obtained from the following equation: where γ is the water specific weight (9,800 N m -3 ); F is the flow rate (m 3 s -1 ); and H m is the pressure head supplied by the pumping station (m).
• Energy supply efficiency (ESE): This index represents the ratio of the theoretical energy requirements and the energy supply. ∆E is the difference between the initial energy of the water (IE) before being diverted from the water source and the energy required for supplying the water and for operating the irrigation system (ER): • Overall energy efficiency (OEE), that takes into account the efficiency of the pumping station and the efficiency in the water supply:

Energy saving scenarios
In order to evaluate the impact of possible energy saving measures, two scenarios were developed taking into account different management strategies. Then, both scenarios were simulated using the EPANET hydraulic model (Rossman, 2000). The second scenario represent an alternative management strategy, not implying any change or upgrade in the hydraulic infrastructures. Their main characteristics are defined below: -Scenario 1. It represented current management. The network worked on-demand and the pressure head was fixed to 85 m. The hourly demand patterns, calculated using cluster analysis techniques, were used to establish the hourly base demand.
-Scenario 2. The irrigated area was organised in two independent sectors according to a homogeneous elevation criterion. The first sector included the hydrants under 127 m height, while in the second, hydrants above that elevation were included (Fig. 2). In this scenario the network was managed under semi-arranged demand and each sector could irrigate for 12 h per day only. The pressure head was fixed to 65 m and 85 m for scenarios 1 and 2, respectively. In order to ensure that every farm receives the same amount of water as in scenario 1, despite the reduction in the time allowed for irrigation, the base demand was doubled, assuming a uniform distribution pattern during the irrigation period. Therefore farmers had to apply higher flows in a shorter period of time.
These two scenarios were simulated for three days with different water demand levels, 6 th June and 15 th July (characterized by medium demand, 542.06 and 942.64 Ls -1 , respectively) and 14 th August (the peak water demand day, 1,478.40 Ls -1 ).

Hourly demand patterns of the irrigated area
The evolution of the average monthly water demand in 2007 is presented in Figure 3. The irrigation season started at the beginning of March and ended in the middle of October. The peak demand period occurred from June to August. Although the month with the largest demand was July, the peak daily irrigation demand was on 14 th August (1,478 L s -1 ). Between November and February farmers did not irrigate.
The weekly demand patterns were used to perform the cluster analysis. The irrigation season took 43 weeks. After repeating the analysis for two, three and four clusters, the best fit (minimizing the variance within clusters) was obtained for two clusters. One of the clusters included 35 weeks, with small variability among hours (cluster 1 in Fig. 4), and the second cluster included the eight remaining weeks, with significant variability between peak and off-peak hours (cluster 2 in Fig. 4). Cluster 2 included the low water demand weeks.
Cluster 1 covered the peak demand period and its standard deviation (0.64) was smaller than in cluster 2 (0.70). Therefore cluster 1 was the most representative irrigation pattern of the irrigation district. In this cluster, water consumption was very homogeneous during the day, with slight increases during the mornings and afternoons (from 18:00), regardless to the energy costs. In cluster 2, consumption was mostly concentrated from 7:00 to 14:00. Thus, most of the water demand occurred when the energy price is maximum. Figure 5 presents the frequency distribution of pumped discharge. Results confirmed that low flows were the most common. Actually, 40% of the instant flows were in the range 0-0.05 m 3 s -1 . However, when a discharge of 0.1 m 3 s -1 was exceeded, flow frequencies sharply dropped, always remaining below 3% frequency in each interval.

Energy analysis
The hydraulic behavior of a particular pump is specified in its characteristic curves, which relate discharge, pressure head, hydraulic performance and power. These  Figure 6 (a, b, c). The comparison of Figure 6c (performance of the pumping station) with Figure 5 (dis-charge histograph) shows that for the most common flow rates (0-0.1 m 3 s -1 ) the pumping station performance was extremely low (even lower than 25%). Flows above this range can be classified in two groups: the first group was composed of flows from 0.1 to 1 m 3 s -1 , where performance may exceed 90%; the second group included flows in excess of 1 m 3 s -1 , for which performance was around 70%.

Performance indicators
Descriptors (Table 1) The average applied depth was 1,783 m 3 ha -1 and the irrigated area during the studied irrigation season was 5,228 ha. The applied depth was signif icantly smaller than the irrigation water requirements, which were estimated as 4,760 m 3 ha -1 . The deficit irrigation is a common practice in this irrigation district and less than half of the total water requirements are applied. Comparing the volume of water diverted for irrigation (measured at the pumping station), and the volume of water supplied to users (measured at the hydrants), the conveyance efficiency was estimated as 96%, which implies adequate maintenance with very low water losses. (Table 1) Although the average recorded power was 1,989 kW, there was a significant variability among months. The peak power was 5,070 kW. Even during the peak demand season the power performance (ratio of the recorded power and contracted) was 68%. The contracted Energy audits in irrigation district S157 Figure 5. Frequency distribution of pumped discharges in the Fuente Palmera district. power could be reduced, achieving relevant savings in energy tariffs. In the off-peak months both recorded and contracted power were significantly lower, being the ratio for the entire season 71%. The ratio of the peak power consumption and the irrigated area was 0.9 kW ha -1 , which is small in comparison with the total installed power (2.2 kW ha -1 ). Thus the pumping capacity was too big even for the peak demand months. (Table 1) In the Fuente Palmera district 0.73 kWh were required to pump every cubic meter of water, implying energy consumption per unit of irrigated area of 1,360 kWh ha -1 . The average energy cost was €0.05 m -3 . As a consequence, energy represented about 30% of the total Management, Maintenance and Operation (MOM) costs. (Table 1) Since in the Fuente Palmera district all water is pumped, the EDR was 100%, being the ECI 70 m. The PEE indicator was around 70%, which in the classification proposed by IDAE (2008) is considered as excellent efficiency, included in Category A.

Efficiency indicators
The ESE indicator depends on the network's design and management as it represents the ratio between the minimum energy required by the system for supplying water to all the hydrants and the energy consumed in S158 M. T. Carrillo-Cobo et al. / Span J Agric Res (2010) 8(S2), S152-S161

Energy saving scenarios
The output of the EPANET simulations for the three studied days (6 th June, 15 th July and 14 th August) is summarized in Table 2.
Pumping performance was slightly higher for scenario 1 (where the network operated on-demand, ranging between 67% and 77.74%) than for scenario 2 (where irrigation was organized in two sectors, with an average performance of 71%).
When the average flows presented in Table 2 were analyzed in Figure 6c, on 6 th June and 15 th July, performance was around 80% (these average flows are on performance curve's maximum). For flows larger than 1 m 3 s -1 (as happened on 14 th August) performance was less than 70%.
Although the average power was very similar in the on-demand and sectored scenarios, the peak power was significantly reduced when sectoring was introduced. This is because in scenario 2 the water demand pattern was uniform, avoiding peaks in irrigation demand. This reduction in the peak power led the pumping station to work more time under high performance conditions and therefore implied a reduction in the daily energy costs. Actually, in scenario 2, reductions of up to 12% in energy costs could be achieved. The savings in relation to scenario 1 in both, peak power and energy costs, are summarized in Table 3.

Discussion
With the aim of improving irrigation efficiency, modernization of obsolete open channel distribution networks has been a common practice in Spain in the last decades. On-demand irrigation represents a step forward in flexibility for water users and an efficient way to reduce the water demand. On the other hand, it implies a significant increase in energy costs. In the coming years, irrigated agriculture will have to face the challenge of improving efficiency in all the resources involved in agricultural production, not only water. Thus, it is time to reflect on this and assess whether on-demand irrigation represents a clear benefit in terms of global sustainability and on the economic prof itability of irrigated agriculture. In this context, energy audits represent an important measure to evaluate energy use in irrigation districts and to detect inefficiencies.
In this work, the energy audits protocol has been applied to Fuente Palmera irrigation district. Analyzing the obtained indicators, the district was globally Energy audits in irrigation district S159  classified in group C (normal) according to the classification provided by IDAE. The OEE was estimated in 56%. These findings are consistent with other works in different Spanish regions, where the average OEE for several irrigation districts take value similar as 67% in irrigation districts in Murcia (Abadía et al., 2007), 41% in Castilla-La Mancha (Córcoles et al., 2008) or 59% in irrigation districts of Navarra (Ederra and Larumbe, 2007). Introducing sectors in network management (organizing farmers in two shifts) was proposed as an energy saving measure. Additionally, this energy saving measure improves ESE (Rodríguez-Díaz et al., 2009). The viability of this measure has to be analyzed in every case, checking that the on-farm irrigation systems can perform adequately when the irritation time is reduced. However, Carrillo-Cobo et al. (2010) reported that in the particular case of Fuente Palmera, even in the peak demand months and taking into account that flows are limited to 1.2 L s -1 ha -1 , farmers would be able to apply their irrigation water with small probabilities of supply failures, mostly when the local practices (deficit irrigation) are considered.
This management strategy was hydraulically simulated on EPANET, resulting in energy savings of approximately 12%. These energy savings may compensate the increment in energy tariffs. Further energy savings could be achieved by improving the hydraulic structures, such as the pumping station or the network layout and dimensions.