Power Quality and EMC State of the Art and new Developments

Transcription

1 Power Quality and EMC State of the Art and new Developments Math Bollen 1,2, Mats Häger 3, Frans Sollerkvist 1 1 STRI AB, Ludvika, Sweden; 2 Luleå University of Technology, Skellefteå, Sweden; 3 Banverket, Borlänge, Sweden math.bollen@stri.se 1 INTRODUCTION The increasing interest in power quality is visible among others from the large number of publications on the subject. Figure 1 gives the number of papers per year containing the term "power quality" in the title, as a keyword, or in the abstract, in the INSPEC database. The increase in production of papers, especially since 1995, is clearly visible Number of papers Year Figure 1 Number of publications per year in power quality. A lot has been achieved the last 10 years and power quality has become a mature subject [1], [2], [3]. But there remain some unresolved issues, some of which will be addressed in this paper. It should be noted that this paper does not aim at giving a complete overview of power quality; instead only a short description is given of some of the ongoing trends. 2 STANDARDISATION A number of interesting power quality standards have been developed during the last several years and more are on the way. Currently the main developments are ongoing within IEC, with IEEE activities ongoing in a number of working groups but in general lagging compared to IEC. However, where power quality and distributed generation is concerned, IEEE has published a 1

2 number of interesting documents and is taking the lead in this area. In this paper, two developments are discussed in some more detail: performance indices and compatibility between equipment and supply. 2.1 Power-quality indices and objectives Power-quality indices are quantifiers for the quality of supply at a specific location in the system, or for a (large) number of locations. The definition of suitable indices is an important standardsetting activity. The work resulting in IEC has provided some of the basic measurement definitions for calculating power-quality indices. The method for calculating power-quality indices differs for variations and for events, as shown in Figure 2. Power-quality variations are slow and small changes in the shape, amplitude and/or frequency of the waveform. Examples are harmonic distortion, voltage amplitude variations (typically referred to as "voltage variations") and dc components in ac systems. Basic characteristics, such as the 10/12-cycle 5th harmonic voltage, are obtained from sampled voltages over pre-defined time windows. Basic characteristics are aggregated into values covering longer time periods, e.g. 1 minute. From the aggregated or basic characteristics, site indices are calculated, e.g. the 95-percentile of the 1- minute values over 1 week. The site indices from several sites are combined into system indices, e.g. the percentage of sites for which the site index does not exceed 4%. sampled voltages sampled voltages basic characteristics (features) Segmentation Characteristics versus time aggregated characteristics pattern recognition Single-event indices site indices site indices system indices system indices Figure 2 Calculation of power-quality indices for variations (left) and for events (right). Power-quality events are sudden and large deviations from the ideal waveform, such as voltage dips, interruptions and transients. What distinguishes events from variations is that events require a triggering action. From the waveform data, characteristics versus time are obtained: rms voltage for voltage dips; single-events indices for voltage dips are residual voltage and duration; different site and system indices are in use, with SARFI (number of dips per year below a given threshold) and the voltage-dip table (number of dips within given ranges of duration and residual voltage) being the most popular ones. Features, indices and objectives are defined in a number of standard documents; an overview is given in Table 1. The current situation can be summarized as follows: features are defined for most disturbances; site and system indices exist for power-quality variations, for long 2

3 interruptions and are under development for dips; objectives exist for variations but not for events. For transients, not even the features are defined, although an informative annex till IEC lists some features. Table 1 Status of power-quality indices and objectives in international standards and related documents. Voltage variations Harmonics Flicker Voltage dips Interruptions Features IEC IEC IEC IEC IEC IEEE Std.1453 Site indices IEC IEC IEC CIGRE C4-07 IEEE P1564 System indices CIGRE C4-07 CIGRE C4-07 IEC CIGRE C4-07 IEEE P1564 Voltage characteristics EN EN EN Planning levels IEC IEEE Std.519 IEC EMC levels IEC IEC IEC IEEE Std.1366 Some network operators move a step further than the standard and define their own quality objectives. The planning levels and voltage characteristics as used by the Swedish transmissionsystem operator (TSO) are presented graphically in Figure 3 [4][5]. The voltage characteristics are what the network operator "promises" to their customers; whereas planning levels aim at deciding on the connection of distorting installations. The voltage characteristics are higher or equal to the planning levels. 3 2,5 Harmonic order 2 1,5 1 volt char planning level 0, Objective [%] Figure 3 Voltage characteristics and planning level for the Swedish transmission grid. 2.2 Definitions The discussion on definitions for different types of disturbances used to be an important part of any power-quality study. Recently, these discussions have been put aside as the general idea has developed that disturbances cannot be uniquely defined. Instead, the waveform is characterized 3

4 through a number of accurately defined features. It is not possible to distinguish between harmonics and flicker but it is possible to define THD (Total Harmonic Distortion) and Pst (short-term flicker severity) for any measured or simulated waveform. In the same way powerquality events are defined through a triggering method and a threshold setting. The same voltage waveform may cause triggering on multiple criteria, e.g. dip, swell and transient. Discussion is ongoing about naming complex events, like the one in Figure 4; the event would be described as a swell of duration T 1, an interruption of duration T 3, a dip of duration T 1 +T 2 +T 3, or all three at the same time. rms voltage swell threshold dip threshold interruption threshold T 1 T 2 T 3 time Figure 4 Example of an event that can be classified as a dip, as a swell and as an interruption. With most monitors the event will also cause a trigger on voltage transients. 2.3 Equipment and network performance A number of power quality disturbances have a negative impact on equipment connected to the power network. Both end-user equipment and generator units may be affected. The risk for maloperation or damage can be limited by improving equipment performance or by improving network performance. Disturbance level Immunity level Immunity limit Voltage characteristic (Electromagnetic) compatibility level Planning level Electromagnetic enviroment Figure 5 How to achieve compatibility between equipment and power supply for power-quality variations. 4

5 This concept is well developed and understood for power-quality variations; the general approach is summarized in Figure 5. The immunity limit for equipment is equal to or higher than the compatibility level. The compatibility level is a disturbance level that is not exceeded for most locations for most of the time (95% of time and 95% of locations are often mentioned in this context). Each device is next tested against the immunity level so that its actual immunity level is higher than the immunity limit and thus higher than the compatibility level. An automatic consequence of this is that the equipment performs as intended most of the time for most locations. There are some problems with this approach, e.g. the max 5% of locations where equipment will not operate at all, and the max 5% of time that equipment will not operate at other locations. In practice this is less of a concern than it may appear at first as the vast majority of locations has a quality of supply significantly better that the compatibility level. Another concern is that equipment is tested under strictly defined laboratory circumstances whereas the conditions may differ significantly during the application of the equipment. This is especially a concern for higher frequencies. The voltage characteristic is defined as the level that is not exceeded for any location most of the time; again 95% is commonly used. The discussion on the 95%-limits in EN is well known and will not be repeated here. Planning levels are internal quality objectives that a network operator uses to ensure that the quality of supply does not deteriorate. Planning levels are used, for example, to assess the connection of fluctuating or distorting equipment. It is obvious from the above reasoning that the planning level should be more strict than both compatibility level and voltage characteristic. More recently regulators in different countries have been using power-quality indices as performance requirements that the network operators have to full-fill. Unfortunately the voltage characteristics in EN have often been used as a base. The resulting requirements are unacceptable for many costumers due to the fact that there are no requirements during 5% of time and due to the 10-minute averaging used to calculate the indices. For power-quality events, like dips and interruptions, the approach in Figure 5, based on probability-distribution functions, is not valid. Instead limits may (have to) be placed on event characteristics and number of events per year. System indices for interruptions are available and widely used. Most countries require a reporting from the network operators on the number of voltage dips experienced by customers. In most cases only the average number of interruptions over the whole network is reported. Recently some countries have introduced additional requirements: this can be either a limit on the number of interruptions for individual customers (e.g. not more than 3 per year) or a limit on the maximum duration of an interruption (e.g. 18 hours). Such requirements are typically associated with a penalty system; where penalties have to be paid to the affected customers or to all customers through a reduced tariff. Indices and objectives for voltage dips are still under development. Proposals for site and system indices are being discussed but the discussion appears to be drifting away from its original aim, compatibility between equipment and supply, towards indices that can be easily used for comparing sites and systems. There is, as far as the authors are aware, no work ongoing on voltage-dip objectives. A possible approach towards defining suitable objectives is shown in Figure 6. The left-hand diagram shows the requirements placed on equipment in IEC Any dip within the light-grey area on the left should not lead to equipment mal-operation. The dark-grey on the right indicates long 5

6 interruptions, with durations longer than 3 minutes. These are the events for which the regulator puts requirements on the network operator: these events should be limited in duration and in number. It is clear from the figure that there is a huge gap between the requirements placed on equipment and the requirements placed on the network operator. Any event between the grey areas will not have to be tolerated by equipment but there is also no requirement on the network operator to limit their number. A more desirable situation is shown in the right-hand diagram. Instead of two curves, one for equipment and one for the network operator, there is only one curve left. Equipment should be immune for any dip above this curves; the network operator should limit the number of dips below the curve. In terms of EMC standards: "Compatibility has been achieved." Figure 6 The compatibility gap between equipment and supply (left) and a proposal to close the compatibility gap (right). Note the logarithmic horizontal scale. The situation is even more complex in reality. Equipment immunity requirements not always are related to the process driven by that equipment. An adjustable-speed drive with automatic restart may comply with the standard requirement, but the process may not be able to tolerate this restart. Another issue that has to be taken up in standard development is that the characterisation of dips through residual voltage and duration only does not fully describe the equipment performance due to the dip. Especially the voltage unbalance during the dip, i.e. the three phase voltages are different during the dip, is of concern. Recently a new CIGRE/CIRED working group (JWG C4.110) has been started to address voltage-dip compatibility issues between equipment, processes and the power supply [6]. 3 POWER-QUALITY MEASUREMENTS AND MONITORING The developments in power-quality monitoring have gone quick the last several years, driven by development in computing, data storage and communication, but also by customers demand on information on power quality levels. There are no longer serious technical limits in the sampling speed, data storage capacity and analysis speed of the monitors. The limitations are in the implementation of the customers' demand for more information, on the presentation of the results 6

7 in an easy-accessible and understandable way, and in the bandwidth limitation of the instrument transformers. A number of reasons exist for performing power-quality measurements. To estimate the dip frequency and dip characteristics with a customer. This is often an industrial customer with sensitive equipment and/or processes and the results are used to decide about the choice of mitigation measures. Estimating the dip frequency requires long measurement periods, several years at least, so that permanent monitoring at strategically chosen locations may be useful to collect the data for future requests from customers. With such measurements it is important to present the results in such a way that they can be compared with the immunity of equipment against voltage dips. Additional information on the source of the disturbances may be an aid in choosing mitigation methods. Next to long-term measurements, stochastic prediction methods are suitable to give accurate prediction of the expected number of disturbances without the need for a long monitoring period. To estimate the network performance. The deregulation has resulted in a higher demand from regulators for data to quantify the performance of the network. Information on powerquality disturbance levels may be a future demand. Some network operators use such measurements for internal reporting and to prioritise investments. To find the source of disturbances. This is part of the trouble-shooting that used to be the main application of power-quality measurements. Detailed information on disturbances is needed for such measurements. The main interpretation is still done by power-quality experts through visual inspection of the recordings. The implementation of automatic recognition tools would simplify this work, especially when used in combination with permanent monitoring. To monitor the network. Once a number of power-quality monitors are installed, they not only provide information on power-quality disturbances but also diagnostics for the network. The purpose of power-quality monitors overlaps with that of disturbance recorders. For monitoring of the network it is no longer sufficient to give residual voltage and duration of the events. Instead automatic disturbance recognition is needed. In the forthcoming sections the recent progress in research towards automatic methods for classification and recognition of power-quality disturbances is summarized. 3.1 Automatic information extraction Methods for analysis and classification of power-quality disturbances consist of a number of steps, where each step requires specific tools; see the block diagram of Figure 7. 7

8 v(t) i(t) event segments Feature extraction features Segmentation Classification class transition segments Additional processing Figure 7 The process of classification of power-quality disturbances Segmentation and triggering are essential steps in deciding which time window of the voltage and/or current waveform requires further processing. Standard methods for feature extraction are heavily based on root-mean-square (rms) and discrete Fourier transform (DFT), but a number of interesting alternatives have been proposed in the literature, with emphasis on wavelet-based methods. The extracted features are the input to the actual classification. This includes finding the origin of the disturbances and extracting additional information from the signal. Standard methods are rather straightforward, often based on the comparison between a feature with a threshold value. Classification of power-quality disturbances is in itself nothing new; distinction has been made between harmonic distortion, voltage dips, flicker, etc for many years already. Both IEEE 1159 and EN give methods for classification of events into dips, swells, (short and long) interruptions, under- and overvoltages, and transients. Further classification of voltage dips into events with different severity is done, e.g. by the South-African standard NRS In all cases very simple classification criteria are used, based purely on residual voltage and duration of the event. The recent developments are the use of advanced classification methods and classification based on origin of the event [7]. The latter is also referred to as the extraction of additional information from the event recording. A number of papers on automatic classification of voltage disturbances have been published during the last several years. These can be roughly classified into two groups: Classification of event waveforms, with typical classes including: "voltage dips", "interruptions", "transients" and "distortion". This work has its importance for the development of classification tools but has limited practical value, as standard methods are available. Classification of events based on their origin, with typical classes including "faults", "transformer energizing" and "capacitor energizing". This work has huge practical value. The classification methods used and under development roughly fall into the following groups: Visual inspection by a power-quality expert. This is the method most commonly used in practice for classification after origin of a disturbance. Rule-based systems, often implemented as an expert system, to distinguish between different classes of event. An expert system for classification of voltage dips and interruptions based on their origin is presented in [8]. The method distinguishes between 9 classes and has been applied to a large set of measured disturbances obtained in a medium-voltage distribution system. Statistical-based methods using advanced signal-processing techniques. The artificial neural network is by far the most popular method in literature often combined with a set of wavelet filters for feature extraction and fuzzy logic for the decision-making. Such a method is used in [9] 8

9 for a classification based on waveform and in [10] to distinguish between transformer energizing and faults. An origin-based classifier, based on neural networks, is presented in [11]. A classifier based on support vector machines is presented in [12]. The latter two classifiers are tested using large amounts of measured voltage recordings. Many papers concentrate on waveform classification and use synthetic data (i.e. data obtained from simulations). It remains unclear if training by using synthetic data will result in a classifier that can be applied to measurement data. 3.2 Triggering, Segmentation, Flagging Next to the development of automatic classification methods and partly as a pre-processing for such methods, research is ongoing on methods to detect the start and end of power-quality events as well as sudden changes in character during the event. Three different terms are used for what is essentially the same problem: "triggering"; "segmentation" and "flagging". Triggering is needed to start the recording and analysis of an event; segmentation is used to split an event recording into time windows during which the features are reasonably constant; flagging is used to prevent double counting for events and variations. The methods are treated separately in most documents, but they can all be traced back to the detection of non-stationarity in a signal Standard Methods Triggering methods for voltage dips, swells and interruptions are prescribed in IEC : the one-cycle rms voltage shall be compared with predefined threshold levels. Triggering methods for voltage and current transients are less standardized. A number of methods are being used, some of which are mentioned in an informative annex with IEC , but none of them is dominating. A flagging method is prescribed in IEC in which a feature is labelled ("flagged") when a dip, swell or interruption was present during the time window over which the feature is obtained. For example, a 100-ms dip will require the flagging of the 10/12-cycle values, the 150/180-cycle values and the 10-minute values. Note that the document does not state that the values should be removed during further analysis, but the general practice appears to be to remove all flagged data Stationary and non-stationary signals A signal is stationary when its statistical properties do not change as a function of time. The concept of feature extraction is based on the assumption that features are constant over the analysis window. For analyzing measurement signals, the term "quasi-stationarity" is often used to indicate that a small variation in feature values is acceptable. When signal statistics vary significantly with time, the signal is referred to as non-stationary. Note that stationarity is not the same as constant, e.g., the calculation of the flicker severity (Pst) is based on the assumption that the statistical properties do not change during a 10-minute window, but the voltage magnitude is far from constant. It should also be noted that there is no strict border between stationary and non-stationary signals. The difficulties in triggering, segmentation and flagging are related to the absence of such a strict border. 9

10 400 Voltage [V] Time [cycles] 10 Harmonic group [%] Harmonic order Figure 8 A voltage dip of 5 cycle duration within a 10-cycle window (top) and the harmonic groups obtained over the 10-cycle window. An example of a (simulated) non-stationary signal is shown in Figure 8 over a 10-cycle window. The pre-dip and post-dip rms voltage is 230 V; during the dip the rms voltage is equal to 70 V; the dip duration is 5 cycles. Both before and during the dip, the distortion of the waveform is zero. The spectrum of the signal is calculated over the 10-cycle window. The harmonic groups are shown as defined in IEC Despite the absence of waveform distortion, the spectrum shows high harmonic distortion, with 8% second-harmonic voltage. As the waveform is no longer quasi-stationary, the resulting features (in this case the harmonic spectrum) do not relate to any properties of the actual waveform. This voltage waveform should however be classified as a voltage-dip event, not as a high harmonic distortion. The harmonic spectrum over this 10-cycle window will be flagged and most likely removed during the further analysis. Triggering is a way of detecting the non-stationarity; this both starts the processing of the event and flags the features. It is thus no surprise that in IEC the same method is used for triggering as for flagging. In some cases the voltage waveform for an event contains other non-stationarities apart from start and end of the event. The result is that the feature extraction method for the event (e.g. the rms voltage or the phase angle) cannot be applied to the whole event recording. Segmentation is a method to detect these additional non-stationarities and split up the signal into a number of stationary segments to which feature-extraction methods can be applied. 10

11 Voltage [kv] Voltage [kv] Time [Cycles] Time [Cycles] Figure 9 Multi-stage voltage dip: voltage waveforms (left) and rms voltages versus time with transition segments indicated in grey (right). An example of a complex event is shown in Figure 9: a measured dip due to a developing fault that is cleared in two stages. The one-cycle rms voltage is plotted as a function of time. The grey bands indicate the so-called transition segments during which the signal features suddenly change and the quasi-stationarity assumption no longer holds. The transition segments are often directly associated with sudden changes in the power system; in this case (from left to right): initiation of a non-symmetrical fault; development of the fault to a three-phase fault; fault clearing by circuit breaker nearest to the fault; fault clearing by the second circuit breaker. The time windows in between the transition segments are referred to as event segments. Any of the feature extraction methods can be applied to the event segments, including the standard methods DFT and rms Overview of Methods Three different approaches for segmentation, triggering and flagging can be distinguished: Changes in voltage magnitude are detected, typically the rms voltage over one or one half cycle. The standard triggering method from IEC is based on this method. An absolute threshold may be used, or a relative threshold where the current value is compared with the values obtained during the past (typically the last few milliseconds up to seconds). The slidingwindow reference method in IEC is an example. The output of a high-pass filter can be used to detect sudden fast changes in voltage or current waveform. This method is used for detecting transients in some power-quality monitors. Most of the work on wavelet-based triggering uses the highpass filter property of wavelets to detect an event, e.g. dip, transient [9][14][15]. One of the problems in practical implementation of waveletbased triggering method is the noise in the filter outputs when using actual measurements. This ``noise'' is a combination of harmonic waveform distortion, cycle-to-cycle variations in the fundamental and harmonic contents as well as high-frequency components in the signal. Being able to reduce the noise would make the method also applicable for the detection of very small changes and for the detection of transients in heavily distorted signals. Methods for reducing the noise level of the filter output are proposed in [16][17][18]. These three proposed methods can be applied to any signal with a zero expected value before the event, not just to the output of a wavelet filter. All three methods have been applied to measured signals but none of them has been applied to a sufficiently large set of measurements to test the performance of the method. 11

12 The residual from a model-based method indicates how much the current value deviates from the model obtained from the values obtained in the past. This is an important advantage of the use of model-based methods: they not only estimate the signal features but also indicate when the signal becomes non-stationary. In [8] the Kalman-filter residual is used for segmenting voltage dips, swells and interruptions and for the detection of transients. In [16] the residual of a predictionerror filter is used to detect the start and ending points of voltage dips. 4 MITIGATION METHODS AND EQUIPMENT Voltage variations and voltage dips are the main power-quality problems for many large and small industrial customers. The economical consequences for individual customers and for the industry as a whole, due to these deficiencies is complex to determine, but it is clear that the costs can be very high and therefore need increased attention. More and more effort is spent in many organisations and projects in order to quantify economical and financial consequences due to PQrelated problems for industrial and public customers. A clear and open mapping of the financial consequences is needed to make a decision on the need for mitigating measures. Information from power-quality surveys and stochastic prediction methods is also needed for such an assessment. For low-voltage customers problems due to transients and overvoltages have increased. A possible explanation is the increased use of electronic equipment, like computers, internet communication equipment, radio, and TV. The problems appear to occur especially in areas and geographical locations with different electrical environments, network impedance and network topology. The economical consequence for an individual customer is in most cases small compared to the huge sums associated with interruption of plant operation for large industrial customers. However the total number of low-voltage customers affected makes that the total societal costs are still significant. Harmonic distortion is a concern for large industrial customers but is in most cases taken care of rather well in the design of the installation. Growing levels of distorting equipment with small and domestic customers has been mentioned as a concern for several years, but no serious widespread problems have been reported to date. The main disturbances leading to problems are either a shortage of energy (dips and interruptions) or a surplus of energy (transients and overvoltages). These two problems are associated with two different mitigation approaches: - Energy injection, either as active or reactive power in order to improve the voltage profile. - Energy removal, i.e. limiting voltage magnitude in transients. Modern power electronic devices such as Static Var Compensators (SVCs) contain components capable of fast switching of high currents and blocking high voltages. This has resulted in a new generation of SVCs based on voltage source converters, with ratings up to several hundreds of MVA. With a short response time these devices can be used for mitigation of both slow and fast voltage variations and voltage unbalance by injecting varying amounts of reactive power. Since the currents switched by the valves also can control active power, the possibility to improve voltage quality by energy injection is determined by the energy storage capacity. Improved battery performance, super capacitors, high-speed flywheels, and similar devices in combination 12

13 with voltage-source converters, form an interesting base for new applications. Possible applications include mitigation of flicker and waveform distortion with large industrial installations; power-factor or reactive-power control for wind turbines and large wind parks; voltage-dip mitigation through injection of active and reactive power. An important part in the long-term solution of power-quality problems is played by international standards. As mentioned in Section 2.3, these documents are a part of the responsibility sharing between the network operator and the customers. In the IEC standards on electromagnetic compatibility the responsibility is mainly placed with the equipment manufacturers. Relevant immunity and emission requirements for equipment and systems will improve the compatibility and provide a clearer foundation for selecting system availability and reliability. Hopefully, and if correctly used, the outcome of standardization work will in the long run also lead to less power quality related problems. The application of communication and improved protection algorithms in relay protection makes new protection schemes possible with improved selectivity and in many cases also shorter overall fault-clearing times. This in turns limits the duration of voltage dips. The rapid change in many countries replacing existing overhead lines with underground cables will limit some power-quality-related problems (dips and interruptions associated with adverse weather), but may introduce different system behaviour. With cables instead of overhead lines the number of faults will most likely decrease during the first years or even decades, but at the expense of longer outages once there is a fault. This is a well-understood consequence of cable network and normally taken into consideration in the design. However, the ageing of solid cable insulation is quite different compared to outdoor air insulation utilized in overhead lines, and will most likely require new and reliable diagnostic methods in order to maintain the initial benefits. Without such techniques a high fault rate may appear suddenly somewhere in the future. All the above discussion, and in fact the vast majority of work on power quality, is directed towards known problems with existing equipment and network. The challenge is to build up a sufficient level of knowledge on power engineering and power quality to detect unforeseen problems at an early stage. Such is the more important as these as-yet-unknown problems may not have an appropriate mitigation method available. Power-quality education is an important mitigation method against both currently known problems and against as-yet-unknown future problems. 5 OTHER EMERGING ISSUES A number of other areas can be identified on which work is ongoing or needed. An admittedly incomplete list is given below. The work on light flicker due to voltage fluctuations ("voltage flicker" in short) has resulted in the flickermeter standard, which is seen as an important achievement of early power-quality work. The method has recently been adopted by IEEE as well. The flicker severity Pst gives an indication of the severity of flicker with incandescent lamps due to measured voltage fluctuations. The index does however not give an indication for the flicker due to nonincandescent lamps like the increasingly popular fluorescent and energy-saving lamps. The development of suitable models is needed to guarantee the future applicability of the flickermeter method. 13

14 Most of the past and ongoing work on power quality has been directed towards voltage dips and waveform distortion up to about 2 khz. Work has also been performed on voltage flicker, transient overvoltages due to lightning and on capacitor-energizing transients. As a result of this there exists a workable understanding and knowledge of these disturbances. The situation is different with disturbances at higher frequencies, like distortion in the range khz. The increasing use of active interfaces with equipment, e.g. in modern fluorescent lamps, will lead to an increased emission in this frequency range [19][20][21]. The jury is still out on the potential impact of this, but the need for additional knowledge on these disturbances cannot be denied. Power-line communication also takes place in this frequency range. A further understanding of disturbances and damping in this frequency range is also needed to understand the limitations of e.g. remote meter reading via the grid. Another area on which knowledge lacks are voltage and current transients. Some increase in their number is expected with an increasing use of capacitor banks and reactors for voltage control. Also the increased use of underground cables will give an increase in the number of transients. At low-voltage level switching of end-user equipment will give more transients. As the character of the transients is not changing significantly, no new wide-scale problems are expected. An increased knowledge on this area will however not hurt, especially as information on the origin of transients may be an important part in the automatic analysis of power-quality events. Overvoltages with durations up to several minutes are a neglected part in power-quality standard. Requirements are typically set for 10-minute averages, but not for overvoltages of shorter duration. This is a place in which the increased attention for power quality may actually have resulted in a deterioration of the supply quality. Overvoltages during single-phase faults and transient overvoltages are normally not a concern for end-user equipment, with the exception of lightning overvoltages. The concern is in longer-duration overvoltages due to load switching, capacitor and cable switching and miscoordination between transformer tap-changers. Methods are needed to include these in standards and regulations. The integration of distributed generation and renewable energy sources in the power grid will lead to new phenomena and possibly new problems. Many, but not all, of the potential problems are associated with power-quality disturbances like voltage variations, flicker, harmonic distortion and energizing transients. A major concern for the system operator is the immunity of the generators against voltage dips and frequency swings. The sudden loss of a large amount of generation is a serious threat to the transmission system security and should be avoided. Immunity requirements set by transmission-system operators may contradict however with protection and anti-islanding requirements set by distribution-system operators. 6 CONCLUSIONS Power quality has become a mature area on which a lot has been achieved in research as well as in product development and standardisation. But enough challenges remain to make this an interesting area, not in the least because of the need for cooperation between network operators, industrial customers, equipment manufacturers, representatives of domestic customers, and researchers from different disciplines. 14

15 7 REFERENCES [1] R.C.Dugan, M.F.McGranaghan, S.Santoso, and H.W.Beaty, Electrical power systems quality. New York: McGraw Hill, [2] M.H.J.Bollen, Understanding power quality - voltage sags and interruptions. New York: IEEE Press, [3] M.H.J. Bollen and I.Y.H.Gu, Signal processing of power-quality disturbances. New York, IEEE Press, [4] TR6-01, Tekniska riktlinjer för elkvalitet, Del 1: Spänningens egenskaper i stamnätet. [5] TR6-02, Tekniska riktlinjer för elkvalitet, Del 2: Planerings- och emissionsnivåer, mätmetoder och ansvarsfördelnimng avseende elkvalitet i stamnätet. [6] CIGRE/CIRED JWG C4.110, Voltage dip immunity of equipment used in installations, [7] M.H.J.Bollen and I.Y.H.Gu, Analysis and Classification of Power Quality Disturbances: Ideas, Methods and Techniques, submitted to International Conference on Harmonics and Quality of Power, October [8] E.Styvaktakis, M.H.J.Bollen and I.Y.H.Gu, Expert system for classification and analysis of power system events, IEEE Trans Power Delivery, Vol.17, No.2, 2002, pp [9] Zwe-Lee Gaing, Wavelet-based neural network for power quality disturbance recognition and classification, IEEE Trans Power Delivery, Vol. 19, no. 4, 2004, pp [10] T. X. Zhu, S.K. Tso, K.L. Lo. Wavelet-based fuzzy reasoning approach to power-quality disturbance recognition, IEEE Trans Power Delivery, Vol.19, No.4, 2004, pp [11] S. Santoso, E.J. Powers, W.M. Grady and A.C. Parsons, Power Quality Disturbance Waveform Recognition Using Wavelet-Based Neural Classifier, IEEE Trans Power Delivery, vol.15, no.1, 2000, pp [12] P.G.V. Axelberg, I.Y.H Gu, M.H.J. Bollen, Classification of power-network disturbances by using the support vector machine method, this conference. [13] O. Poisson, P. Rioual and M. Meunier, Detection and Measurement of Power Quality Disturbances using Wavelet Transform, IEEE Trans Power Delivery, vol.15, no.3, 2000, pp [14] L. Angrisani, P. Daponte, M.D'Apuzzo and A. Testa, A Measurement Method based on the Wavelet Transform for Power Quality Analysis, IEEE Trans Power Delivery, vol.13, no.4, October 1998, pp [15] S.Mallat, W.L.Hwang, Singularity Detection & Processing with Wavelets, IEEE Trans Information Theory, vol.38, no.2, 1992, pp [16] J. Chung, E.J. Powers, W.M. Grady and S.C. Bhatt, New Robust Voltage Sag Disturbance Detector using and Adaptive Prediction Error Filter, IEEE Power Engineering Society, Summer Meeting, [17] O. Sen, Z.-X. Song, D. Chen and J. H. Wang, Application of Wavelet Soft-Detection Threshold Denoising Technique in Power Quality Detection, International Power Quality Conference, Singapore, October [18] H.-T. Yang and C.-C. Liao, A De-Noising Scheme for Enhancing Wavelet-Based Power Quality Monitoring Scheme, IEEE Trans Power Delivery, vol.16, no.3, 2001, pp [19] E.O.A. Larsson, C.M. Lundmark, M.H.J. Bollen, Measurement of current taken by fluorescent lights in the frequency range khz, IEEE PES General Meeting, June [20] C.M. Lundmark, E.O.A. Larsson, M.H.J. Bollen, Unintended consequences of limiting high-frequency emission by small end-user equipment, IEEE PES General Meeting, June [21] M.H.J. Bollen, F. Sollerkvist, A. Larsson, M. Lundmark, Limits to the hosting capacity of the grid for equipment emitting high-frequency distortion, this conference. 15