DE3146374A1 - Method and arrangement for detecting and indicating cooling malfunctions in reactor cores - Google Patents

Abstract

A method is described which is available even in the event of transient processes in the reactor, and can detect cooling malfunctions with high sensitivity. The local fuel or coolant temperature, which cannot be directly measured in a sodium-cooled fast breeder reactor is determined for this purpose from gross output, primary throughput of coolant, coolant inlet temperature and coolant outlet temperature at each fuel element with the aid of a recursive parameter and status estimating method, preferably using a Kalman filter. The determined estimated value is monitored for deviations from the desired value. Deviations initiate a fault indication. The noise of the estimated value of the local fuel temperature is substantially smaller than that of the previously employed coolant outlet temperatures, as a result of which the high sensitivity is achieved. Moreover, the simultaneous availability of the fuel and coolant temperatures improves the diagnosis of the causes of the malfunctions. Due to the continuous determination of all relevant parameters, the method is available without limiting the sensitivity even in the event of changes of state. A correction circuit also enables monitoring of short-term malfunctions of the core parameters such as occur, e.g., in the event of the passage of bubbles.

Description

Procedure and arrangement for the determination and reporting of

Cooling Disturbances in Reactor Cores The present invention relates to a method and an arrangement for the detection and reporting of cooling faults in reactor cores, especially of sodium-cooled fast breeder reactors the main patent application P 31 19 045.6. In sodium-cooled fast breeder reactors various cooling disturbances in the reactor core are conceivable to avoid major damage should be detected as quickly as possible. In the event of a cooling fault, for example, the partial or total clogging of one of liquid sodium traversed channel in a fuel assembly or the occurrence of larger gas bubbles in the coolant, it can lead to overheating of the fuel and thus to a more or less severe damage to the fuel assembly concerned.

For the early detection of such cooling disturbances, up to now the coolant temperature measured at the fuel element outlet and with a setpoint compared. If the deviation of the output temperature from the setpoint exceeds one certain size, a cooling disturbance is assumed and a corresponding one Error message output or, if necessary, a shutdown of the reactor triggered. The main problem with this method is the constant fluctuations in the Coolant outlet temperature, the so-called noise of the measured value, which the Limits the sensitivity of this procedure. Either must namely the response threshold can be set very high so that the statistically occurring peaks in the Noise spectrum no false alarm is triggered, or one must through time integration the measured values a Make smoothing, which is based on the time constants leads to delayed error messages and falsifications in the case of transient processes.

An improved method for reducing the noise in the measured value the coolant outlet temperature was determined at the Specialists Meeting on Reactor Noise (SMORN II) in Gatlinburg, Tennessee (USA), 1977 by M. Edelmann under the topic "Noise and DC Balanced Outlet Temperature Signals for Monitoring Coolant Flow in LMFBR Fuel Elements proposed. An attempt is made to reduce the fluctuations in the coolant outlet temperature through further measurements and corrections to compensate for the sensitivity and improve the response time of the cooling monitoring system. The one suggested there The procedure actually compensates for a considerable part of the fluctuations, however, it is primarily suitable for stationary operation, as some parameters are used as must be entered constantly. A drifting of these parameters may lead to false messages, if the response threshold for the alarm is not selected high enough.

The monitoring function is based exclusively on the coolant temperature derived. Simultaneous monitoring of the fuel temperature is not possible In DE-OS 30 08 198, a similar set of problems is dealt with. It becomes a process described to determine the proportion of the volume throughput in a fuel assembly yestad. However, there is no provision for monitoring the fuel temperature.

The object of the present invention is therefore to provide a sensitive Detection of cooling faults in reactor cores, especially those that are sodium-cooled fast breeders, to enable false reports to be largely excluded without the need to restrict the sensitivity of the monitoring. In particular should also a cooling monitoring in transient processes, i. H. during normal Changes to operations in the reactor system are possible without restricting the early warning potential be what is excluded in the previous procedures.

In addition, the display of short-term faults, such as Bubble passage through individual fuel assemblies is aimed for. Diagnosing the causes of cooling disturbances should by the simultaneous monitoring of the fuel and Coolant temperature can be improved.

To solve the problem described, a method according to the Main claim proposed. This is no longer just the coolant outlet temperature monitored on each fuel assembly, but the not directly measurable local fuel or coolant temperature from easily measurable quantities using a thermohydraulic Derived from the model. This local fuel or coolant temperature is one Quantity, the value of which is afflicted with very little noise. The problems, which resulted from the noise of the coolant outlet temperature, arise no longer after a determination of the local fuel temperature. Therefore can a deviation of the local fuel or coolant temperature from the setpoint high accuracy, so that an error message for the relevant fuel assembly can be triggered. Since the local fuel temperature is not directly direct can be measured, it was, according to the previous opinion of experts, as an indicator not suitable for a cooling fault. It was assumed that a determination of the fuel temperature with a high level of instrumentation, for which there would be no room at all in a fast breeder reactor would have to.

However, it has been shown that a determination of the local fuel temperature from other easily measurable quantities with the help of a simple thermohydrauliscllen Model is possible.

In the second claim it is therefore proposed that the local fuel or coolant temperature by measuring total power, primary throughput of coolant, Coolant inlet temperature and coolant outlet temperature of each fuel assembly, as it is done in a known manner in nuclear reactors anyway determine. This method does not require any additional instrumentation on the nuclear reactor and still delivers enough measured values for all free parameters of the terhmohydraulic Model. With the help of these measurements, the fuel temperature in the Fuel assembly, the coolant temperature in the fuel assembly, the power fraction of the Fuel assembly, the proportion of the mass flow rate of the fuel assembly, the specific Heat of the coolant in the fuel assembly, the heat capacity of the fuel and the Coolant in the fuel assembly and the integral heat transfer coefficient between fuel and determine coolant. In particular, the fuel temperature is now suitable for monitoring cooling faults in each individual fuel assembly. The remaining Sizes are used to classify those indicated by the fuel temperature Anomaly, d. that is, they can provide information about the cause and risk of an anomaly give.

In the third and fourth claim it is proposed that the fuel or coolant temperature using a recursive parameter and state estimation method, in particular using a Kalman filter, from the measured values as a function of to ascertain from the time. This method, which in particular with the aid of of a computer can be run at great speed, also works at Changes in the state of the reactor flawlessly and considerably faster than the previous ones Procedure.

In addition, the availability of the monitoring is not made by semi-manual Calibration impaired, which is another advantage over the prior art is.

In claim 5 an arrangement for performing the method suggested. To do this, the transducers for total power, primary throughput, Coolant inlet temperature, coolant outlet temperature of each fuel assembly connected to evaluation electronics in which the local fuel or coolant temperatures can be derived from the measured values according to a recursion formula.

The local fuel temperatures now take the place of the previous ones monitored coolant outlet temperatures. Their noise is important lower than that of the previously measured coolant outlet temperature.

In the case of sodium-cooled fast breeder reactors, the following measurands are available through measurements that are necessary anyway: P = total output of the core TI = temperature of the coolant at the core inlet W = total throughput of the coolant in the cooling area Furthermore, the coolant outlet temperature is measured for each fuel element: T0 = coolant outlet temperature with the aid of a simple thermohydraulic A fuel assembly can be described as follows: The letters have the following meaning: Tf = mean fuel temperature of the fuel assembly Tc = mean coolant temperature in the fuel assembly tp = proportion of the fuel assembly output in relation to the total output P = = proportion of the fuel assembly throughput in relation to the total throughput wk = heat transfer coefficient between fuel and coolant h = specific heat of the coolant Cf = Heat capacity of the fuel in the fuel assembly Cc = heat capacity of the coolant in the fuel assembly Assuming a linear axial temperature profile in the fuel assembly, the following applies: T = 2 (T + TI) or To - 2TC -c 21 (T0 + T1) or T0 c 1 For a Linearization around the working point, the. is time-dependent, and a discretization of the linear model, equations (1) and (2) can be summarized in a system of linear difference equations: where + t, H and D are parameter matrices and means the state vector of the fuel assembly at time tk.

is the measurand and is the vector of the also measured input (Disturbance) quantities.

The system of equations (3), (4) describes the effect of the disturbances uk on the core behavior at the respective discretely selected measurement times bk in the environment of the working point.

One of the problems dealt with here is that it is not to determine directly measurable state vector Xk from the measured values Bk and uk. this can be in a particularly suitable manner by a state estimation method, a so-called Kalman filter, accomplish. The related theory is e.g. B. described in the book: K. W. Schrick, "Applications of Kalman filter technology", Oldenborg Verlag, Munich Vienna 1977.

In the following, a superscript index, introduced only for better understanding, is used, which indicates the point in time up to which the measured values have flowed into the quantity indexed in this way. For example: Estimated value of the state vector X for the time tk + 1 using measured values up to and including the time tk.

A predicted estimated value for X at time tk + can be calculated according to (3): The more imprecise this estimated value is at the beginning, the greater the difference between the measured value Yk + 1 and the estimated value will be (see (4)) Therefore the actual Kalman filter uses the difference weighted with a factor Kk + 1 as a correction to the predicted estimate As a result, measured values from time tk + 1 flow in, and the result is an improved estimated value according to the recursion formula: It has been proven in the literature that approaches the actual value X with increasing number of measurements, ie with increasing index k.

For the invention presented here, however, it is crucial that the fuel temperature, which cannot be directly measured, can be made available and that this variable is extremely low-noise. The reason is on the one hand the large heat capacity Cf of the fuel, which has the effect of a 'Z low-pass filter', so to speak, on the noise of the various parameters, and on the other hand the fact that the estimated value is considerably less noisy than the actual value of the fuel temperature Tf also the designation Kalman filter.) A block diagram for the evaluation according to (5) including this filter is shown in the drawing, FIG Circuit processed in such a way that the non-measurable core state respectively.

at the exit.

The circuit exactly reproduces the recursion formula (5).

The estimated quantity present at the output The estimated quantity present on the becomes amplified with and fed to an adder 1. From the measured value present at the input uk + 1, the measured value uk delayed by At is formed via a delay circuit, which is amplified with r and also fed to the adder 21.

The measured value Yk + 1 also present at the input is fed to an adder #z. This is where the result value Xk, amplified by the c'k factor -H, comes from the adder and also the measured value amplified with -D. The result (corresponds to Yk + 1 - Ykt1) is weighted in an amplifier with Kk + 1 and fed to the adder , at the entrance too is present. The estimated value comes from C3 as the result which takes the place of after the time dt t via a delay circuit occurs when in the next step is appreciated. The continuously measured variables uk + 1 and Yk + 1 are queried and processed by the circuit via a clock generator at time intervals bt (discretization), whereby the delay by # t can also take place by storing the previous value. In the block diagram, it should be noted that the individual quantities are vectors or matrices, so that a circuit with a breakdown according to individual components would contain the individual components several times, which has been avoided here for a better overview and schematization, but is immediately apparent to the person skilled in the art.

The very low noise estimate of the fuel temperature Tf of the respective The fuel element is now monitored, and a deviation from the target value triggers an error message the end.

It is also crucial that this monitoring process also for transient processes (e.g. load maneuvers) is available since all relevant parameters are continuously redefined from the measurements (automatic Calibration).

For this purpose, the quantities Yk and Uk measured at time tk are used with the aid processed by a second circuit in such a way that the estimated values + and r occur at the output and can be inserted into the Kalman filter if necessary. It is essential that that the accuracy of the core parameter estimation does not depend on the estimation errors in depends on the fuel and coolant temperatures. (There is no feedback from the Kalman filter required on the parameter estimation arrangement.) It is from the literature known (e.g. E. V. Bohn, M. K.

De Beer "Consistent Parameter Estimation in Multi-input Multi-output Discrete Systems ", Automatica, Vol. 13, pp 301-305, 1977) that the state x in the equations (3), (4) eliminated analytically and that thereby an algebraic Relationship between the elements of the system matrices +, P and the measured values Y, u can be made. The resulting equations are standard linear Regression format.

The unknown parameters can therefore be checked using the method of least squares can be determined recursively.

The arrangement is structurally identical to that in the drawing illustrated Kalman filter. The state vector x is replaced by the Elements of and r formed parameter vector. Thus, through this coupling the availability of the Fuel element monitoring increased sharply, while false alarms increased despite increasing Sensitivity can be reduced significantly.

In the system described so far, however, there are still cases in which the monitoring with a Kalman filter is not applicable, or the estimation results are not interpretable. This is especially true for short-term disturbances, as they arise, for example, when bubbles pass through a fuel assembly.

So far, attempts have been made to pass bubbles through the core with the aid of a reactimeter, but this is only possible if all fuel elements fairly evenly be traversed by bubbles while the passage of a small bubble through just one fuel assembly to one undetectable with the reactimeter Effect leads. In the case of the Kalman filter described so far, the recurring Periodic calibration is only effective if the system parameters (such as B.

Heat transfer coefficient, radial flow rate and power distribution) are sufficient change the estimation error slowly compared to the convergence speed, d. H.

remain constant for at least a minute. Disturbances of the radial Throughput distribution can also be determined with good estimation accuracy, if the disturbance lasts only briefly (fractions of a second). But if at the same time also change core parameters with the fuel element throughput disruption, then the Estimation accuracy of the Kalman filter may not be sufficient for short-term processes. This is the case with the passage of the bubbles.

Since with the Kalman filter based on the model, the measured values k again are estimated, the "detuning" of the Kalman filter can be recognized by that the estimated values for the entered measured values differ considerably from these. These deviations, in the following measured value -Residuen bY- called Yk- yk, but contain additional information that is required when compensating for the described Disadvantage can help. Simulation calculations have shown that with passage of a bubble through a fuel assembly, values of the measured value residue that differ greatly from zero appear. This can only be a consequence of the parameter errors occurring during the passage of the bubbles be. It should be remembered that the parameters (K = heat transfer number between Fuel and coolant, Cc = heat capacity of the coolant in the fuel assembly, æw = proportion of the fuel element throughput in relation to the total throughput) during the flow of the bubbles of about 0.5 seconds depending on the bubble size. After the Bubble from the core, these parameters take on the original values again, but the estimation errors in T caused by this "detuning" only disappear after about 20 seconds

An arrangement is therefore proposed in claim 6 which also enables the passage of bubbles to be detected and diagnosed and which cancels the detuning of the Kalman filter. For this purpose, a correction circuit is added to the Kalman filter. To understand this correction, the thermohydraulic model is to be used again: The fuel temperature Tf can be represented as a function of measured quantities and the residual h Y and the estimated value for Tc: A corrected value is now sought that makes AY zero and must therefore satisfy the following equation: Since this correction can be carried out by processing measured variables and the variable Tc which is not affected by brief parameter disturbances, this offers an improvement in monitoring.

The claim 6 therefore provides a correction circuit that this Formula exploits to follow the Kalman filter. For example, in the case of bladder passage, where the Kalman filter is briefly detuned, the one determined by formula (7) gives Value # Tf, COR shows the current value of the fuel temperature.

Since the input variables u are noisy, but otherwise only move slowly change, these can, as proposed in claim 7, by means of low-pass filters be smoothed.

The time constants can be chosen to be large compared to the noise frequencies are and still remain below those that are significant for shifts in the operating point Time. The time constants are on the order of a few seconds.

The output variable of the correction circuit Tf, COR is also somewhat noisy and can be smoothed according to claim 8. The time constant of the proposed However, the low-pass filter is very critical, since too large a time constant causes a phase error in Tf COR so that this value does not follow changes quickly enough. A there is an acceptable compromise between smoothing effect and phase error Time constants of about 1 sec.

The heat capacity of the fuel itself acts on changes in the Fuel temperature like a low-pass filter with a certain time constant. Their size should be kept as an upper limit.

Another aspect of the correction circuit is dealt with in claim 9. The output variable STc of the Kalman filter can be replaced by the mean value of the coolant inlet and outlet temperature with sufficient accuracy for many cases: Therefore, the correction circuit can also have pure measured variables applied to it at the input and does not require the output values of the Kalman filter. In this case, the correction circuit works even if the Kalman filter fails and contributes to the redundancy of the monitoring system. For this design, a low-pass filter would have to be used to smooth the noisy input variable are provided.

In the drawing, exemplary embodiments of the evaluation and correction circuit shown. Figure 1, already in connection with claim 5, shows the basic structure of a Kalman filter for solving the given task. The selected display type is based on the Kalman filter technique usual drawings.

Fig. 1 does not show a circuit diagram in the usual sense, but a kind Block diagram represented by an analog circuit or also, for example, with a Microprocessor can be realized.

In Fig. 2 is a corresponding plan for an embodiment the correction circuit shown. The circuit forms through adders and amplifiers the formula (7) according to.

The input variables are each smoothed by a low-pass filter TP1, TP2, TP3. The time constants T1 72 T3 r3 of the three filters are not very critical. In FIG. 2, amplifiers are marked by squares in which the gain factor is entered. Triangles with inscribed Dt mean a delay with a time difference of bt between the input and output values. Adders are drawn as circles, with the signs of the summands next to the inputs. The correction circuit shown shows an embodiment that occurs when the measured values are discretized, but this is not absolutely necessary. The continuously measured values at the input are taken over by the electronics at intervals # t. The delay stores the applied value for the time At, so that the previous measured value is present at their output when the new one arrives at the input.

The considerations outlined here for monitoring the cooling of a Reactor by means of "indirect measurements" (i.e. not measurable by means of the estimates Sizes) using a Kalman filter and recursive parameter estimation methods can be generalized and allow e.g. B. also a determination of the contributions to the reactivity balance. It therefore seems likely that the most special Example of cooling disorders, tried and tested methods are gaining in importance and they will lead to global core surveillance.

The ongoing determination of all relevant, but not necessarily directly measurable sizes of a reactor from the measurement data improves both the early warning potential as well as the possibilities of identifying the causes of malfunctions.

Claims (9)

Procedure and arrangement for the detection and reporting of cooling faults in reactor cores patent claims $ 11 method for the detection and reporting of Cooling disturbances in reactor cores, especially of sodium-cooled fast ones Brutreaktorent according to the main patent application P 31 19 045, characterized by the following Features: a) The not directly measurable local fuel or Coolant temperatures are derived from easily measurable quantities. b) The local fuel and coolant temperatures are monitored. c) A deviation in the local fuel or coolant temperature from the setpoint triggers an error message for the fuel assembly in question. 2. The method according to claim 1, characterized by the following features: a) The local fuel or coolant temperature is determined from measurements of total power, Total primary coolant flow rate, coolant inlet temperature and coolant outlet temperature on each fuel assembly, which are carried out in a manner known per se. b) The measured values are at predetermined time intervals A t from a Evaluation or computing electronics are called up that provide an estimated value for the local fuel temperature or coolant temperature determined. 3. The method according to claim 1 or 2, characterized by the following Features: a) To determine the fuel or coolant temperature, a recursive Parameter and state estimation methods applied. 4. The method according to claim 3, characterized by the following features: a) The state estimation method includes a Kalman filter. 5. Arrangement for performing the method according to claim 1, characterized by the following features: a) The transducers for total power, primary coolant flow rate, Coolant inlet temperature and coolant outlet temperature of each fuel assembly are connected to an evaluation unit in which the local fuel and coolant temperatures can be derived from the measured values. b) The evaluation unit forms the formula by amplifying the measured or estimated values with appropriate factors and adding them using adders according to this formula. c) The local fuel or coolant temperature, which is used as an output signal the evaluation unit is supplied to a setpoint comparator, which is connected to large deviation triggers an error signal. d) There is another circuit that is based on the measured values the parameter matrices t, (each re-estimates and, if necessary, in the evaluation unit feeds. 6. Arrangement according to claim 5, characterized by the following features: a) There is a correction circuit that follows the Kalman filter is. b) The measured values g and sTa are applied to the correction circuit as input variables at. c) Another input variable is that from the output of the Kalman filter incoming estimate STc. d) The correction circuit forms the formula according to: 7. Arrangement according to claim 6, characterized by the following features: a) For smoothing the (quasi-stationary) input variables low-pass filters (TP1, TP2, TP3) are available. b) The time constants (# 1 # 2 # 3) of the low-pass filters (TP1, TP2, TP3) are large compared to the noise frequencies and small compared to those for operating point shifts necessary times. 8. Arrangement according to claim 6 or 7, characterized by the following Features: a) To smooth the output of the correction circuit Tf COR is a Low-pass filter (TP4) available. b) The time constant (# 4) of the low-pass filter (TP4) is in the order of magnitude of 1 sec and is limited by the heat capacity of the fuel given time constants. 9. Arrangement according to claim 6, 7 or 8, characterized by the following features: a) The input variable is instead from the output of the Kalman filter, the size which is formed from measured variables. b) It is a low-pass filter for smoothing available.