See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/302585801 Systematic Description of Dynamic Load for Cables for Offshore Wind Farms. Method and Experience Conference Paper · August 2016 CITATIONS READS 3 1,102 4 authors, including: Thomas Kvarts Rasmus Olsen Ørsted Offshore Wind Ørsted 9 PUBLICATIONS 111 CITATIONS 8 PUBLICATIONS 125 CITATIONS SEE PROFILE All content following this page was uploaded by Thomas Kvarts on 21 August 2018. The user has requested enhancement of the downloaded file. SEE PROFILE 21, rue d’Artois, F-75008 PARIS http : //www.cigre.org B1-303 CIGRE 2016 Systematic Description of Dynamic Load for Cables for Offshore Wind Farms. Method and Experience Thomas KVARTS, Ivan ARANA DONG Energy Wind Power Denmark Rasmus OLSEN, Poul MORTENSEN Energinet.dk Denmark SUMMARY With the increasing size of offshore wind farms, the number of array and export cables are also increasing. Since the cost of cables contribute significantly to the cost of offshore wind farms, there is an increased focus on improving the transfer capability of each cable, thus reducing the cross section or even the number of cables. It has been seen from Distributed Temperature Sensing (DTS) measurements that cables installed for offshore windfarms, designed for continuous load, rarely reach their expected service temperature calculated at steady state. This especially applies to areas with deep burial, which, to a large extent can be explained with the thermal capacity of the surrounding soil when the dynamic nature of the load in cables have not been considered [1]. This paper suggests a systematic method of describing the dynamic load from offshore wind farms ending up with a “worst-case dynamic load profile” that with a simple step load curve together with the thermal properties of the surroundings, provides an unambiguous design basis for cable sizing/rating. This can be further used in simple dynamic Finite Element Method (FEM) models by cable manufacturers or designers to model the cable installation conditions. The method is a development of the method presented in CIGRÉ TB610 [2] appendix D.5. To further justify the method, a Thermoelectric Equivalent (TEE) model, [3], is used to calculate the maximum cable temperature on the actual input load data for the worst year available as well as the found worst-case load profile. The TEE is used since it is an analytical numeric solution that can quickly calculate the temperature response even for detailed load data corresponding to several years. The TEE model has previously been compared to FEM and laboratory measurements [3]-[4] but a comparison between the TEE model and site measurements from an offshore wind farm is presented as a novelty in this paper. KEYWORDS Dynamic Load, HV Cables, Submarine Cables, Wind Power, Thermal Resistivity, Thermoelectric Equivalent, TEE. thokv@dongenergy.dk Background The work presented in this paper is a continuation of the method for describing the dynamic load of a wind farm, presented in CIGRÉ Technical Brochure 610 Offshore Generation Cable Connections, appendix D.5 [2]. The method has been used by the Danish TSO Energinet.dk to specify the cable for the “Horns Reef 3” 400 MW offshore wind farm, where a “worst case load even” was found by subjecting the load data to a moving averages analysis. The assessment of the data was mainly performed intuitively, based on knowledge of thermal constants of a buried cable system, and print out of a large number of running averages. This simple approach was possible since the amount of data was limited to only 3.5 years and because results could be compared and partly confirmed by service experience with the nearby offshore windfarm “Horns Reef 2”. In the following, the same methodology is developed further, giving a more automated and reproducible approach to choose appropriate length of time steps and how to exclude irrelevant data. This development gives the cable designer an easy way to find the length of a final full load period and a way to exclude short full-load periods in the running averages of preceding longer-low load periods, as well as to identify few critical years to investigate in depth. Worst-case dynamic load profile A load pattern of a cable for a wind farm can be directly related to the wind speed by the wind farm power curve, see example in figure 4-14 in CIGRÉ TB610 [2]. It has in the recent years become possible to obtain many years of detailed historical wind data, modelled or measured, even for offshore areas, with resolution down to 10 minutes. It is therefore now possible to analyse the cable load data in detail prior to specifying and selecting the cable. The advantage of using historical wind time series, rather than wind statistics, is that the nature of the wind is captured better. This means that both the yearly variations, as well as the short duration characteristic of the weather systems consisting of low and high pressure systems following each other in succession, are seen. It is for example easily concluded, from such wind data, that it is unlikely for an offshore wind farm to experience more than 5-7 days of continuous full production if such an event has never happened in 16 years. These aspects should be included in the design criteria used for dimensioning cables, in an unambiguous way, that when used together with the thermal characteristic of the surroundings enable different suppliers to achieve the same results. The “Worst case dynamic load profile” as found in the following, is an attempt to provide design criteria based on the dynamic load of a wind farm utilising the thermal capacity (inertia) of the surroundings of the cables for improved cable rating: Long term running RMS A base tool, used in the method is long term running RMS (Root Mean Square) of load values, that will be used instead of just running average load throughout this document in line with the CIGRÉ TB 610, [2] example. This is due to the fact the heat losses of a cable is mainly proportional to the current squared. These running RMS load values are found as a function of start time and length of the period according to (1). 1 π‘=π‘ +π 2 πΌ (π‘)ππ‘ πΌπ ππ (π‘π , π) = √ ∫π‘=π‘ π π π (1) 2 Where: t is the time, ts is the start time, T is the length of the RMS period over which the heating of the cable by the current must be found (e.g. 7 days, 14 days, 20 days, 40 days, 70 days…) and I(t) is the load current as function of time based on historical wind data and the park power curve. Reactive power (charging current) at the location of the cable route where hot spot is expected could be included. But since it can be shown that the heat loss contribution of a constant reactive power is independent of the load current, it is recommended to ad this late in the process, since it will vary with capacitance of the final cable design and solution for reactive compensation. It could even be left up to the cable manufactures to include the reactive power based on a given reactive power compensation setup. It should however be noted that the reactive power may not be fully independent of the park load. Step by step method for generating a Worst case dynamic load profile: 1. A useful first step in analysing the data has been found to be plotting the maximum running RMS load current (as function of the period T) using (1). a. Data on historical wind speed and wind direction of the farm w(t) are gathered for as many years as possible. b. The wind farm power curve P(w) including wake losses is found or estimated c. The load current in the cable is found as a function of the produced power, and the expected voltage, I(P, U). d. From w(t), P(w) and I(P,U) the running (RMS) load current is IRMS(ts,T) is calculated for different periods T (1 day, 2 days… up to 60-100 days), with starting times ts in each data point for the entire data set, e.g. each 10 min or 1 hour (data sets of 1 hour resolution should be sufficiently accurate). For each period T the maximum value is found. These maximum values can now be presented as a function of T, in Figure 1.1 7 days 10 days 40 days Figure 1; Example of maximum wind farm load in percentage for a future offshore windfarm in the UK as a function of the period T, based on 16 years of historical wind data modelled for the windfarm location. 2. From Figure 1 the relevant periods T to investigate further can be found by looking for step changes in the graph. This part of the analysis will require some experience, but 1 In the example the charging current has at this point not yet been included 3 the most important last step change of the “worst case load profile” (Figure 5 (b)) x days of 100 % load is found quite easy from this figure (step a. below): a. It is seen that it is unlikely to ever have more than T=7 days of 100 % load since the highest 7 day RMS value for 16 years is already fallen to 98%. b. The next significant step change is at T=10 days where the RMS drops to 96%, c. The final step change found from the graph is at T=40 days, d. No more T values have been deemed necessary since the final 3 steps will describe the last 57 days (40+10+7 days) of the worst case load profile, which exceed the length of the most windy part of the winter periods of the investigated years, and a RMS value for the entire data set of 16 years can be used for the preceding steady state initial value. e. The number of step changes to use could be refined to improve the effect of the method if needed, but in the example shown in Figure 1 identifying 3 step changes has for the worst case load profile, been seen to be sufficient to get a significant rating improvement. 3. The time periods T (here 7, 10 and 40 days) found in the previous procedure (2) are used to calculate and plot the RMS current for the entire data set provided (16 years), to identify the most windy years to be analysed further. If a few most windy years (winters) cannot be identified from this, additional periods could be analysed (e.g. 17 (7+10) or 47 (17+40) days). An example for T = 40 is shown in Figure 2 (i.e. running IRMS(ts,T=40 days) for 16 years), where the two peaks for year 10 (I4010) and 15 (I4015) can be selected for further analysis. This can be repeated for the other chosen periods T since additional worst years may need to be investigated. Figure 2; Running IRMS(ts,T=40days) for 16 years, the two peaks identified for year 10 (I4010) and 15 (I4015) to be analysed further (the charging current from an assumed worst case location on the cable route is included at this time but this could have been omitted and added later) 4. Since it is often the case for IRMS(ts,T) that a T=10 days maximum includes the 7 days maximum, and that the T=40 day maximum includes the T=10 and the T=7 days maximum, for each of the years suspected to be worst case years, the following “punch out procedure” is to be performed so that several events are not included more than once in the worst case load profile: f. The running IRMS p.u. values are plotted and the maximum values for each duration period T (7, 10 and 40 days) are identified (see Figure 3 A, B, C) 4 A B C Figure 3; Square 1 hour current (I2), 7 days (I7), 10 days (I10) and 40 days (I40) running RMS currents as function of days in year 15. The x-axis shows the days starting from the first year. g. The data preceding the maximum running RMS value for the lowest duration period (T=7 days) is removed (blue area of Figure 4) and the new running RMS apparent currents are plotted (dashed lines in Figure 4) as if the data for the 7 day period was not there, (both time and value removed from the data set so the RMS values can be found across/on both sides of the removed period). h. The new residual maximum values for each duration period are located; in Figure 4 it is seen that the T=10 days year-15 maximum is reduced from 0.98 to 0.96 (red Δ). i. The data preceding the maximum value for the second lowest duration period (T=10 days) is removed (red area in Figure 4) and the new running RMS apparent currents are plotted (dotted lines in Figure 4). j. The new residual maximum values are located; in Figure 4 it is seen that the T=40 days year-15 maximum is reduced from 0.93 to 0.88 (green Δ). Figure 4; Original 7 days (I7), 10 days (I10) and 40 days (I40) running RMS currents (solid lines) as function of days in year 15, after removing the worst 7 days (dashed lines) and after removing the worst 7 + 10 days (dotted lines). The x-axis shows the days starting from the first year. 5. Step 4 is repeated for every selected (windy) year and the maximum RMS current values found for each period in all selected years. 6. The steady state preload used is found as the RMS value for the entire period with data available, but cut to include only whole years. In this case 0.74 p.u. (not included in 5 Figure 4) It could be considered to specify the preload time to be limited to the lifetime of the windfarm e.g. 25 years, but the improving effect on cable rating is negligible except for very large burial depths. 7. Finally, the Worst-case dynamic load profile as a step change curve is produced from the previously found residual maximum load for each duration period and investigated windy year, see Figure 5 (b) Example of step 7 performed on year 15: Discussion of data Figure 5a) shows the RMS load current after the 7 days (blue area) and 10 days (red area) periods have been removed, and the 40 days period, used to calculate RMS current with T=40, is shown with a green area. Discussion of data Figure 5b) is the end result of the method, i.e. the worst case dynamic load profile that serves as input to manufacturer calculations used for cable dimensioning. In Figure 5b) it is possible to see how steady state preload (step 6) is followed by the 40 days period (green area), the 10 days period (yellow area) and finally 7 days period (blue area). This is a conservative approach in the example, since the order of events (the step changes) are shifted in time so that the 3 blocks representing 7, 10 and 40 days are in the most unfavourable order with the highest wind at the end of the Worst case dynamic load profile. This reverse order has shown to be the case in all analysed locations so far, where the wind has shown to be more consistent in the fall before the windy winter period, than in late winter/spring. By shifting the order, it is considered that the method includes the odd possibility of having a consistently windy event in late winter/spring, sometime in the future. But if this was not the case, it should be considered to include another margin of security in the method, if dynamic loading is not combined with e.g. online temperature and load monitoring. Wind farm load and RMS current in p.u. Square current and RMS current in p.u. T=7 T=10 1.00 0.96 T=40 0.88 Preload T=∞ 0.74 (a) (b) Figure 5; (a) Running RMS current during an investigated year and (b) Maximum values for all years as a sums up to a Worst case dynamic load profile in p.u. as a function of days (red curve). (Charging current of a specific location on the cable route is included). The coloured blocks in (a) and (b) represent the same areas. In (a) the x-axis shows the days starting from the first year. Discussion of data For this example, the wind farm power curve including aerodynamic wake losses have been used. Depending on the site weather conditions, wind turbines and wind farm layout the wake losses become irrelevant after 16-17m/s as all turbines operate at rated capacity. Additionally, no power curtailment or turbines out of operation were assumed. 6 The historical wind speed and direction of the farm data sets have 1 hour resolution, this has been deemed sufficiently accurate to represent the load variations from a large wind farm and temperature variations to accurately calculate the long term running RMS values. Other faster temperature or loading variations do not need been accounted for since any maximum final temperature will be only seen after several days off full load due to the long thermal constants of the cable surroundings and represented by the last step of the Worst case dynamic load profile. Thermoelectric equivalent For analysing the thermal response, of an export cable, to the Worst case dynamic load profile the Thermoelectric Equivalent (TEE) method may be used. TEE is a simple but very time efficient method, to model the thermal response of a power cable to an applied load. The method is based on the resemblance between heat flowing in a thermal system (such as a cable installation) and the current flowing in an electric circuit. The analyses of the temperature at different positions in cables thus rely on simple algebra, which is why TEE calculations are fast compared to other methods. Different studies, [3]-[8] have shown that the TEE method, even though it is simple, can model the dynamic temperature in a power cable very accurately, i.e. within a couple of degrees centigrade. By applying the TEE method, to a static load and to the load of Figure 6b), it can therefore be compared how large a conductor is needed when 1) dimensioning export cables based on steady state requirements and 2) dimensioning export cables based on the Worst case dynamic load profile. In the simulations the parameters in Table 1 have been used. Parameter L ρtherm,soil θamb ctherm,soil Value 2.0 0.7 10 2β106 Unit m K·m/W °C J/(K·m³) Description Installation depth, centre of cable Thermal resistivity of seabed Ambient temperature Specific heat of seabed (low estimate!). Table 1; Thermal parameters of surroundings As the dimensions and designs of three phased HV cables may vary from manufacturer to manufacturer, the dimensions in Table 2 have been used in the simulations. Parameter dc di ds dj da De Wd λ1 λ2 ks kp Value 41.5 91.0 101.0 107.4 248.9 261.9 0.8 0.3607 0.3573 1.0 1.0 Unit mm mm mm mm mm mm W/m - Description Conductor diameter Insulation outer diameter Lead sheath outer diameter Outer diameter of each cable core Armour outer diameter Cable outer diameter Dielectric losses Screen loss factor. As per IEC 60287 Armour loss factor. As per IEC 60287 Skin effect coefficient Proximity effect coefficient Table 2; Cable dimensions used in the simulation of static and dynamic thermal response to loads. 7 Even though it is a simplification, the dimensions in Table 2 have been used for all conductor cross sections, where adaption in relation to cross section has been made only in the electrical resistance of the conductor. I.e. the thermal simulations have taken into account the lower electrical resistance of a larger conductor by utilising the resistances stated in IEC 60228, [9]. The projected current profile for a 220 kV export cable from a specific wind farm location looks as shown in Figure 6a), where the maximum current is 1052 A. The worst case load for this profile is shown in Figure 6b). a) b) Figure 6; Current used in the TEE calculations with the (a) actual current for the worst year found (year 15) and the given (b) Worst case load profile. Both loads include charging current for a specific location on the cable route. Under conventional, steady state, dimensioning of the export cable, a 1052 A requirement would necessitate the installation of a 2000 mm2 aluminium conductor cable whereas the dynamic loading profile in Figure 6b) only necessitates the installation of a 1600 mm2 aluminium cable as shown in Figure 7. Figure 7; Thermal response of a 1600 mm2 3-phased submarine cable to worst case dynamic load profile in Figure 6b). Comparison between worst case dynamic load profile and dynamic load by TEE The simplified worst case dynamic load conditions of Figure 6b) was, as stated, derived from Figure 6a) as it is easier for the cable buyer to compare the performance of cables from different manufacturers, during tendering phase, when the load conditions are simplified. 8 In order to show that the simplification approach is also thermally valid, the thermal response of a three phase export cable to the load profiles in Figure 6a) and Figure 6b) is compared in the following. The thermal analyses of Figure 8 show that the Worst case dynamic load profile seems to be on the safe side when comparing to the original load profile as the maximum temperature of the conductor in the two cases is 88 °C (Figure 8b)) and 85 °C (Figure 8a)) respectively. This suggests that it might be possible to use an even smaller conductor and still respect the general 90 °C limitation of XLPE cables. a) b) Figure 8; Thermal response of a 1600 mm2 3-phased submarine cable to a) originally predicted load profile and b) worst case dynamic load profile Furthermore, if the final step of 1p.u. (full load) for 7 days in the Worst case dynamic load profile would last less days, it could be possible to decrease the cross section of the cable even further. The thermal response of a 1400 mm² aluminium conductor cable, to the originally predicted load of Figure 6a), results in a maximum conductor temperature of approximately 95 °C which is not acceptable. However in case the buyer allows the use of an off-standard cross section, a conductor of 1500 mm² aluminium would result in 90 °C. By changing the dimensioning criteria from a steady state load to the varying load it is therefore possible to reduce the conductor cross section by 25 %. Comparison between TEE and FEM As the TEE method by nature is a very simplified 1D model of the 3D reality, it should be tested how the method performs compared to more extensive simulations. The present section is therefore dedicated to comparing the results of TEE simulations with 2D FEM simulations. As shown in Figure 9 the TEE method is accurate within ± 2 °C, for almost the complete simulated time period, when comparing to the FEM simulations, and this is even though the simulation time of the TEE method is 10 seconds which is approximately 50 times faster than the FEM simulations. For additional information regarding FEM and TEE calculations, please refer to [3]. 9 a) b) Figure 9; Comparison of TEE method and FEM simulation. a) Absolute comparison and b) difference between simulation results. It should be acknowledged that it is difficult to obtain a more accurate result with a method which simulates the complex structure of a three phased export cable in one dimension. In its nature the TEE method simulates the thermal system as if it consisted of cylinders, which of course is difficult when the cable consists of three phases configured in a triangular geometry. Furthermore, it should be acknowledged that in order to utilise thermal simulations not only in dimensioning but also in the operation of the three phased cables, i.e. utilising Dynamic Line Rating, it is necessary to have simple and fast simulation tool which can be implemented in IT systems already existing at power grid operators (such as SCADA systems). The TEE methodology is ideal for such applications as it requires a minimum of computational resources and relies on well-known algebra. This could be interesting for wind farm operators, since implementing the dynamic load method for cable design, may also encourage to design the cables with the possibility to curtail the power from the wind farm, since the method easily could be used to assess how often curtailing would be necessary over the life time of the wind farm. Relevant DTS measurements This section is dedicated to comparing the TEE simulation of a three phase cable with measurements obtained from the transmission grid. Figure 10a) shows measurements of the highly varying current output of an offshore wind farm from when the first turbines are installed and the following almost two years. Figure 10b) shows measurements of the temperature in the fibre (red, dotted line) and the TEE method simulated temperature (blue, solid line). 10 a) b) Figure 10; a) Measurement of load b) Measurement and TEE method simulation of temperature of the fibre. It is seen that there are some deviations between the measurements and the TEE method simulations; however this should be seen in the light that it has not been possible to gain access to credible data for the surrounding temperature. The simulations have therefore been performed by assuming constant 10 °C surroundings, and this of course affects the accuracy of the simulations when comparing to the measurements. On this background it should be recognised that the TEE method simulations are accurate within ± 8 °C, even with the lack of information regarding the ambient temperature, and this must be considered very satisfactory. From the above it is obvious that greater knowledge of the ambient conditions would lead to a better forecast of the loadability of the cable as mentioned as well in [10]. Measurements of the ambient temperature in the seabed, as well as measurements of the thermal resistivity and heat capacitance, are therefore recommended to be performed wherever possible. It may even be that some meteorological databases (or Geographical Information Systems) already contain such information and such datasets should therefore be sought for. Conclusions This paper proposes a step-by-step methodology to sort several years of data of wind speed and direction, to calculate RMS currents and finally a worst-case dynamic load profile. Here, a sequence of loads with different durations are increased gradually until the maximum load is reached. This simple load profile allows the supplier to design the cable. Subsequently, the conductor temperature is calculated using the predicted load profile and worst case dynamic load profile. Here, the final conductor temperature shows small differences, using both loads. The paper continues to compare the TEE model with FEM simulations. The paper finishes with a comparison of measurements and TEE simulation; the results to show fair agreement. It is shown that application of dynamic loading is relevant when large export cable systems are operated with varying load, allowing for a reduction of the conductor cross section in the range of 25 % compared with a continuous loading. In principle the procedure can be used for both underground and submarine cables; however the installation conditions along the route would determine the dynamics of the cable system and hereby the effect of utilising dynamic loading for cable rating and especially cable routes with deep burial will benefit of from dynamic loading. 11 BIBLIOGRAPHY [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] L. Colla and M. Marelli, Dynamic rating submarine cabes. Application to offshore windfarms. EWEA Offshore 2013 Conference paper 162 [TB 610] CIGRÉ B1.40 Technical Brochure 610, “Offshore Generation Cable Connections”, February 2015 R.S. Olsen, J. Holboell and U.S. Gudmundsdóttir, 2012, "Dynamic Temperature Estimation and Real Time Emergency Rating of Transmission Cables", IEEE General Meeting 2012. R. Olsen, G.J. Anders, J. Holboell and U.S. Gudmundsdóttir, 2013, “Modelling of Dynamic Transmission Cable Temperature Considering Soil Specific Heat, Thermal Resistivity and Precipitation,” IEEE Trans. Power Delivery, vol. 28, no. 3, 1909-1917. R. Olsen, J. Holboll, and U.S. Gudmundsdottir, 2013, "Electrothermal Coordination in Cable Based Transmission Grids", IEEE Trans. Power Systems, vol. 28. No. 4, 4867 – 4874. R. Olsen, 2013, Dynamic Loadability of Cable Based Transmission Grids, PhD-Thesis, Denmark. M. Diaz-Aguiló, F. de León. S. Jazebi and M. Terracciano, “Ladder-Type Soil Model for Dynamic Thermal Rating of Underground Power Cables”, IEEE Power and Energy Technology Systems Journal, vol. 1, 2014. M. Diaz-Aguiló and F. de León, “Introducing Mutual Heating Effects in the Ladder-Type Soil Model for the Dynamic Thermal Rating of Underground Cables”, IEEE Transaction on power delivery, vol. 30, no. 4, Aug 2015 IEC, Standard 60228 Conductors of insulated cables, 3rd edition, 2004 R. Huang, J.A. Pilgrim, P.L. Lewin, D. Payne, “Dynamic Cable Ratings for Smarter Grids”, IEEE PES Innovative Smart Grid Technologies Europe (ISGT Europe), 2013, Copenhagen. 12 View publication stats