Flicker Control for the ESRF Accelerator In collaboration with Good morning, my name is Mauro di vito (Ing at Power supply at ESRF) and in this presentation called flicker control for the ESRF accelerator, I will talk about the power converters that we are designing to feed our booster ring. I will give an overall description of these power supplies, its different components and subsystems, and how the behave regarding flicker generation on the power lines and how can we improve it with our digital control platform. Many of the automatic control components of this system were developed in collaboration with the LIC, a public control laboratory located in Argentina. ARGENTINA

Flicker Control for the ESRF Accelerator What is Flicker? Power quality parameter of a Point of Common Coupling (PCC). Low frequency variation of RMS value. Caused by high pulsed power consumption of a device connected to a PCC So, to introduce the topic, I will explain the definition of flicker in a power line. It is a power quality parameter of a Point of Common coupling (A point of common coupling is the physical point were we connect to a power supply network and we’ll call it PCC). In the case of flicker, it is a low frequency variation of the RMS value of the power line voltage. So flicker is caused by a high pulsed power consumption device connected to a PCC. To illustrate this, we have this generic schematic, where we can see the main voltage source of the power line, in this case, a 3-phase line with a output voltage E. A distribution line with both resistive and reactive characteristics (represented as an impedance), and in the end we have the PCC, where multiple devices are connected. One of them is our “High pulsed power consumption device”. This device draws a generally high current from the line. This current has low frequency characteristics, and thus, produces a pulsed power consumption. As we know from Ohm’s Law, when this current circulates through the line impedance, it will produce a voltage drop (delta U) on the PCC. This gives us a new U voltage that now has low frequency variations To measure the flicker, we take the peak-to peak variations of the RMS value divided by the mean RMS value. By definition: Flicker = 𝚫 𝐔 𝐩𝐩 𝐔 (RMS) l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator Flicker calculation It is desired to reduce the flicker produced by each power device. It is possible to calculate the individual flicker contribution. 𝚫𝐔=𝑰⋅𝑹𝒄𝒐𝒔 φ +𝑰⋅𝑿𝒔𝒊𝒏(φ) So now we will try to predict this flicker parameter. By calculating its value we could be able to think about strategies to reduce it. Our goal is to reduce internally the flicker produced by each power devices connected to the PCC. Its done with this formula here, which is obtained by doing a vectorial analysis of the situation, where we use amplitude and phase information of each AC signal In the graph we can see the original power source E and the PCC voltage U. The objective is to calculate the difference, DeltaU. Our power device Consumes a current with a phi angle in relation to its supply voltage U. This current circulating through a complex impedance produces a DeltaU with a new phase angle and a certain amplitude. By doing trigonometry and using some approximations we obtain this very simple equation. This tells us that the amplitude of the voltage drop is proportional to the amplitude of the consumption current I and is affected by its angle alpha. If any of these 2 parameters presents a low frequency variation, we will have flicker on the resulting U. l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator System under study: Booster at the ESRF Basic configuration of Ramped Injector Power Supplies (RIPS) For Quadrupole Magnets For Dipole Magnets So, having understood the problem of flicker, we can apply this analysis to the new power supplies we are developing for the ESRF booster. These ramping injection power supplies (RIPS) and they basically look like this. On this end we have the final charge, which is an electromagnet, and we need a ramped current circulating through them with a 4Hz repetition rate. We have two types of RIPS systems. One for Quadrupole Magents, and one for Dipole Magnets. The configuration presented on the image corresponds to the Quadrupole RIPS. The Dipole RIPS uses more energy and a more complex configuration. We’ll proceed to look in detail each of it’s components, l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator System under study: Booster at the ESRF (RIPS) PCCs At ESRF: 20kV 400V First, we have the power distribution line. The system takes its energy from this 3-phase line. This is our PCC and its here where we want to perturb as little as possible. We’ll be paying attention to its RMS value and its low frequency variations At the ESRF we have two main power lines : 20kV (for Dipole) and 400V (for Qpole) In the photo we can see the 20kv transformer and in the graph is a typical acquisition of the 3phase of the 400V network l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator System under study: Booster at the ESRF (RIPS) So, the next component is the first power converter of the system: a 12pulse thyristors rectifier bridge. In a thyristors rectifier, by controlling the conduction triggering angle alpha, it generates a variable DC output voltage. We are using an entirely digital control platform for this converter. This board was previously presented by my colleague Olivier Goudard in the 2014 POCPA. By synchronizing its internal PLL to the 3-phase network, it generates the trigger pulses on the gates of each thyristors. Thanks to the different measuring points of voltage and current around the system, we are able to design digital feedback loops that automatically adjust the alpha according to the desired operation mode. The control algorithms for flicker mitigation that I will be explaining in the following slides are all implemented digitally in this platform. “DEVELOPMENT OF A FULL DIGITAL THYRISTOR GATE CONTROL” - POCPA 2014 https://indico.bnl.gov/getFile.py/access?contribId=4&resId=0&materialId=slides&confId=687 l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator System under study: Booster at the ESRF (RIPS) Then, a passive filter is present. The capacitor bank is in charge of storing enough energy to sustain a full accelerator cycle on the booster. supply the magnets disturbing as least as possible the distribution line. l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator System under study: Booster at the ESRF (RIPS) Next, we have the Hbridge. A typical Hbridge allows us to change and/or invert output voltage using Pulse width Modulation. Its switching frequency is 6,4khz and the output filter has a cut-off frequency around 500Hz in order to reduce the ripple on the charge. l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator System under study: Booster at the ESRF (RIPS) Magnets at RIPS: Quadrupoles Dipoles So, we finally have the magnets. These are electromagnets, which generate a magnetic field to curve the electron beam inside. This is done at increasing energies, which goes along with the acceleration of the electrons. This explains the need for a ramped current on the magnets. This graph show an example of the shape of the current. The peak current in the Dipoles is much bigger than the quadrupoles, and explains why they need a RIPS capable of managing higher energies. l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator Flicker propagation Now that we know all of the systems components, we can analyze the waveforms of voltage and currents in different points of interest. I will also mention the values corresponding to the quadrupole RIPS, in order to get an idea of their magnitudes. We have the ramped current of the magnets, reaching a peak of around 500A. Its 4Hz repetition rate will be propagated to the rest of the system and waveforms, as this is the energy consumption rate of the system So, considering that the capacitor bank was charged to 1000V, as the ramping process begins, the energy is transferred from the capacitor to the inductor. As the load current goes back to zero, the energy is returned to the capacitor and in steady state, the capacitor goes back to the same voltage value, thus remaining stable through the rest of the pulsing. As the voltage on the right side of the inductor decreases, a current is drawn from the rectifier. This current naturally presents spikes at a 4hz rate. These spikes can be very high and easily reach a saturation state on the inductor. This variable consumption current is what ends up producing periodic variations on the RMs value of the PCC, and, as defined at the beginning of the presentation, we can say flicker is present on the system l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator Control strategies Constant rectifier output (Open Loop) Constant current control (Current Loop) Flicker control (Current + Flicker Loops) Now that we understand how the system behaves regarding to flicker and we can quantify it on the desired PCC, we can desing multiple strategies to deal with it. This three items are the tree main strategies that we have studied and evaluated and I will proceed to explain each one of them. Each of these presents an improvement over the previous case, but also means an increase of complexity of the control system. As I mentioned, this algorithms are all implemented in the digital gate control platforms used in the thyristors bridge. l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator Control strategies Constant rectifier output Constant current control Flicker control 𝚫𝐔=𝑰⋅𝑹𝒄𝒐𝒔 φ +𝑰⋅𝑿𝒔𝒊𝒏 φ 𝚫𝐔=𝑰⋅𝑹𝒄𝒐𝒔 𝜶 +𝑰⋅𝑿𝒔𝒊𝒏(𝜶) The first situation considers the rectifier with a fixed DC output voltage, without a feedback loop. The resulting Ith current drawn from the rectifier is established by the circuit (which would be the passive filter, and the Hbridge) As shown before, this current has abrupt 4Hz variations, which, as predicted by the flicker formula, will produce flicker. It must be noted that in the case of thyristors rectifiers, the alpha triggering angle of the thyristors is actually the resulting angle shift between the current and the voltage (phi). So, we can use the formula replacing phi by alpha. l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator Control strategies Constant rectifier output Constant current control Flicker control 𝚫𝐔=𝑰⋅𝑹𝒄𝒐𝒔 𝜶 +𝑰⋅𝑿𝒔𝒊𝒏(𝜶) Final flicker value: We ran some experimental testing of this situation in order to quantify the worst possible case, and then asses the improvement. We acquired the voltage value at the PCC. We calculated the RMS value, as we can see its close to the expected 400V. By zooming on it, we can see the 4Hz variations. In order to improve the situation, we had to implement a feedback loop 𝚫𝐔 𝐔 % =𝟎.𝟔𝟓% French Standard (@4Hz): <𝟎.𝟒𝟕% l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator Control strategies Constant rectifier output Constant current control Flicker control 𝚫𝐔=𝑰⋅𝑹𝒄𝒐𝒔 𝜶 +𝑰⋅𝑿𝒔𝒊𝒏(𝜶) In order to improve the situation, the first step is to control the rectifier output current with a closed loop. The objective is to reduce the 4Hz variations of this current. The strategy is to set a constant current during the cycle, that suffices to compensate the resistive losses on the rest of the system. Supplying this constant current guarantees that at the end of each 4Hz cycle, the capacitor regains its initial charge. By successfully eliminating the variations of the current amplitude, as we can see on the flicker formula, we took away the biggest contributor to the RMS variations on the line. But now a new problem appears. The angle between the consumption current and the PCC voltage is no longer constant. With a regulation loop in place, in order to change the rectifier output voltage, the thyristors triggering angle must vary. This alpha angle now oscillates along with the 4Hz variation, and the formula shows us that this must produce flicker. l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator Control strategies Constant rectifier output Constant current control Flicker control 𝚫𝐔=𝑰⋅𝑹𝒄𝒐𝒔 𝜶 +𝑰⋅𝑿𝒔𝒊𝒏(𝜶) Final flicker value: We measured the flicker produced with this new algorithm. We can see the 4Hz variations still present, but with a much lower value. We already gained a lot with this current loop, but in the final system there will be 2 Quadrupole RIPS running along with a Dipole RIPS (which is the biggest contributor) at the same time. This means that the overall 4Hz flicker present on the final system could be critical, and it is desired to find a strategy that would reduce it even more. 𝚫𝐔 𝐔 % =𝟎.𝟎𝟑𝟕𝟏% l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator Control strategies Constant rectifier output Constant current control Flicker control 𝚫𝐔=𝑰⋅𝑹𝒄𝒐𝒔 𝜶 +𝑰⋅𝑿𝒔𝒊𝒏(𝜶) The idea is to use the flicker formula to predict the total perturbations and include a new control loop that reduces this predicted value. We know the alpha value thanks to the TGC platform on the rectifier, and we know the current value as we had to measure it in the current loop control strategy. This loop must be designed so that is works in a frequency band around 4Hz. This way it will output a variable Iref that will be summed to the DC Iref used to compensate resistive losses. Together they produce a current that look like this, which effectively reduces variations of the PCC RMS value Initial experimental tests of this algorithm gave positive results, but we are still dealing with obtaining a precise measure of the R and X values of the PCC. Simulations show that flicker can be reduced at least to a third of the value for the constant current control strategy. l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator Future tests on the ESRF booster The future on this developments consists of further measurements on the Dipole power converter for the booster at the ESRF. As it manages more power than the Quadrupole RIPs, it consists of two sets of the same system connected in series, in order to achieve greater voltages on the magnets. The peak current that this system generates is up to 1500A. And the PCC is of 20kV Further tests and measurements are foreseen for the following months, while we prepare the system for beam generation for the first time. l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito

Flicker Control for the ESRF Accelerator Thank you l Flicker Control for the ESRF Accelerator l 24/05/2016 l Mauro di Vito