FR2769402A1 - NUCLEAR REACTOR DRIVING TECHNIQUE - Google Patents

Abstract

The invention relates to the control of nuclear reactors. According to the invention, the measurements made from sensors lead to the development of possible responses to the observed disturbances induced in the operation of the reactor, in particular by the variations in network load. From these responses, according to the invention, the different initial states are determined by following a simplified treatment which makes it possible to take into account the complexity of the models comprising the use of the generalized disturbance method. The details of the final states thus obtained then give rise to a choice of the response chosen, which is the one which best responds to the previously chosen strategy. The invention allows control in real time with light computing means. This faculty allows to operate according to models closer to physical reality and responding better to the operator's choices.

Description

The invention relates to a technique for piloting nuclear reactors, and

  more particularly a technique that allows more adequate monitoring of

  load variations imposed by the network.

  The piloting of nuclear reactors of the pressurized water type comprises, in a known manner, the use of traditional measurements in particular those relating to neutron fluxes, the concentration of neutron poison, the inlet and outlet temperatures, the pressure, etc. Piloting also includes the implementation of means allowing the modification of the operating conditions, essentially the movement of control bars, and the variations in poison concentration.

neutron.

  The measures indicated above are normally transformed into derived parameters called "observables", which represent the operation of the reactor more closely to operating requirements. These are in particular the anti-reactivity and the reactivity reserve, the axial imbalance, local reaction rates, etc. These derived parameters result from a treatment according to programmed methods of the measurements carried out directly on the reactor at means

many sensors.

  The multiplicity of simultaneous measurements and the interdependence of the phenomena controlling the overall progress make the determination of the incidence of a change made by the control means particularly complex. This determination is however the essential element on which the

piloting of the reactor.

  The processing, according to programmed methods, of the measurement results which allows real-time knowledge of the "observable" parameters can be more or less sophisticated. However, the complexity of the calculations necessary to determine the final state resulting from a disturbance in the operating conditions results in an inadequate data processing time.

  piloting reactors in real time.

  The complexity, and therefore the processing time, are all the more important as the physical model used as a basis for data processing is more elaborate to get as close as possible to physical reality. The complexity is such that the implementation of the models developed is not possible by the calculation means currently used for piloting the reactors. In particular, the use of three-dimensional models is practically impossible in

  piloting conditions used so far.

  This results in control which does not best follow the variations in network load while respecting the instructions chosen by the user to favor certain preferred operating conditions, such as the

  maintenance of an optimal reactivity reserve.

  The invention proposes to provide a technique for piloting nuclear reactors, in particular of the pressurized water type, which meets this objective more satisfactorily, and this without requiring the

  use of extremely powerful computing means.

  The control technique according to the invention uses: - sensors for measuring the operating conditions, - the determination from these measurements of specific quantities of the operation of the so-called observable reactor, - the comparison of the quantities thus determined with data previously established to guarantee the operation of the reactor according to methods chosen by the operator, the programmed development of responses, from the result of this comparison, - the processing by means of calculation of the responses established according to the deviations observed, to predict the consequences of these responses on the final state following such a response, the use by the means of calculation of the generalized disturbance method in which, in the expressions comprising a poltz of Boltzman, only the terms of degree 0 and 1 are kept, assigned a coefficient that accounts for the contrib mean mode of the higher modes for the asymptotic values of the parameters, - the comparison of these forecasts with the previously established data, and the leading choice, among the possible answers, of that answering the best

  to the methods previously selected.

  The invention results from the demonstration of the fact that, in the calculation step leading to a forecast of the final state following a perturbation, the contributions of the higher modes of the Boltzman polynomial expansions, which generate the heaviest calculations , can be replaced by simple coefficients which, with experience, allow to obtain results without significant deviation from those of

more complete calculations.

  In other words, the invention makes it possible to obtain, in a short treatment time and with light calculation means, without significant loss of precision, the forecasts on the behavior of the reactor in response to a modification of the control conditions. A consequence of this improvement is the possibility of implementing a much more precise control which lends itself to better implementation

defined objectives.

  The operator can in particular drive the reactor in accordance with an optimized strategy, for example on the safety margins, the economy of effluents or the limitation of the movements of the control rods with the idea of minimizing the wear of the mechanisms

commanding them.

  The improvement of the means available to the operator makes it possible to use the most sophisticated models, in particular those in 3D for a better representation of the phenomena concerned. Processing a 3D model considerably increases the complexity of calculations, and for this reason has so far been excluded from control techniques

actually used.

  The models used for implementing the invention can involve several types of measurement. These are the measurements made directly

  on the reactor from the signals received from the sensors.

  In practice, piloting necessarily includes neutron flux measurements at different points of the reactor. Other measures are advantageously taken into account, in particular the inlet and

outlet, pressure ...

  With regard to the prediction of situations resulting from controlled modifications, it is necessary to start from a well defined initial state. This state is itself known by instantaneous measurements. These measurements are integrated into an expert-type model beforehand

established.

  In slightly disturbed conditions, the expert system is sufficient for operating the reactor. For a larger disturbance, when at least one measured parameter leaves a predetermined range, either by the operator or automatically in response to a modification of the network load, by the technique according to the invention, the system predicts the evolution of the reactor, establishes a diagnosis on this predicted situation, and draws up a control strategy which allows an evolution that better meets the chosen objectives. The prediction mechanism according to the invention is particularly useful in the case of frequent disturbances which render ineffective the only recourse to an expert system which can only account for events spaced in time. Load monitoring requires frequent modifications that the invention

  allows to treat satisfactorily.

  These frequent modifications can eventually lead to a drift of the "initial conditions" used in the predictive calculation requiring the

recourse to periodic readjustment.

  The periodic registration according to the invention is advantageously carried out on the basis of the measurements of the parameters, by following a technique for calculating

minimization of deviations.

  The calculation leading to this registration is advantageously carried out using also a generalized disturbance method. This method allows the identification, among the various internal parameters of the model, of those which contribute significantly to the deviations which it is necessary to correct. The following registration can then be restricted to these parameters, which limits the processing used

to the registration itself.

  In the following description it is made

  reference to examples established for the fundamental observable parameters of the reactor, showing the reliability of the technique according to the invention using a simplified calculation method with respect to the results of

complete calculations.

  In establishing the forecast of an end state following a perturbation, the expression of the Boltzman operator for the observable parameters, used for this forecast, shows that it is necessary to take into account not only the terms of degree 0 and 1, but also the higher terms in order to approach asymptotic values sufficiently. The contribution of terms greater than 0 and 1 to the asymptotic value of the observables depends on the perturbation itself. These calculations are performed for each parameter and must follow the succession of multiples

  disturbances quickly leading to extreme complexity.

  The technique according to the invention initially comprises the determination of the coefficients which make it possible to dispense with terms of degree greater than 0 and 1 in the calculation of the forecast values of the

parameters after disturbance.

  These coefficients, generally designated by A, are obtained by carrying out one or more complete determinations including not only the terms of order 0 and 1, but also the higher orders

(3, 4, 5, ...).

  The coefficients P correspond to a formula of the type: = SO - St

S - S1

  in which SO, S1, and S are respectively the values of the terms of order 0 and 1 of the parameter, and its

asymptotic value.

  Since Boltzman's polynomial expansions are rapidly converging, as shown in the examples which follow, taking into account terms 3, 4 or 5 is sufficient to approach the value So The value S. can be assimilated to the average of the values obtained by reproducing several times the calculation from different initial situations taking into account only a limited number of these terms

higher order.

  Once the coefficients p have been determined as has just been said, they can be used in practice for determining the forecast values of the final states following a disturbance in the operating conditions of the reactor. The operation is then equivalent to calculating the asymptotic value: SM = So + pS1 In other words, it is sufficient for this forecast value to calculate only the terms SO and S1 of

polynomial expansion.

  Practice shows that the use of this approximation is perfectly justified for disturbances in which it is necessary to take account of higher orders. For minimal disturbances, the higher orders practically not intervening, the approximation does not play. In practice, the proposed treatment will therefore only be carried out for disturbances exceeding a certain threshold.

previously established.

  Examples of determining the extrapolation value of typical quantities: axial imbalance, local power at selected points, anti-reactivity and reactivity reserve. These examples lead, as indicated above, to the

determination of the coefficients À.

  The operation is carried out on a pressurized water reactor model. The neutron flux is the parameter studied in this model. The initial disturbance corresponds to a reduction in the power of the reactor. This reduction modifies the catch sections and generates a modification of reactivity of the system which must be controlled by a movement of the

absorption bars.

  The following tables show the result of the calculations for the four quantities named above and this going up to the fifth order. The determination according to the application of the generalized perturbation technique leads to a convergence evolving in an oscillating manner. Each ascending term introduces a contribution which is

smaller and smaller.

  At the end of the fourth or fifth order the difference between the target value and the calculated and extrapolated value is very small, and it is superfluous to

continue beyond.

  Axial imbalance Error Reference value - 0.0231327 Target value 0.0404286 42.78% Order 1 - 0.0462136 14.31% Order 2 - 0.0384415 4.91% Order 3 - 0.0411215 1.71% Order 4 - 0.0401921 0.58% Extrapolated value 0.0404988 0.17% Local power Error Reference value 0.00536841 Target value 0.00345967 55.17% Order 1 0.00241041 30.33% Order 2 0.00403218 16.55% Order 3 0.00315063 8.93% Order 4 0.00362505 4.78% Extrapolated value 0.00333702 3.54% Anti-reactivity Error Reference value 0 Target value 1054 Order 1 1690 60, 58% Order 2,688 34.64% Order 3,1254 19.12% Order 4,958 8.98% Order 5 1,099 4.45% Extrapolated value 1,046 0.60% Reactivity reserve Error Reference value 3659 Target value 2,607 40.53% Order 1 1969 24, 47% Order 2 2971 13.99% Order 3 2405 7.72% Order 4 2701 3.63% Order 5 2559 1.80% Extrapolated value 2613 0.27% Application of the technique according to invention, using a simplified formula comprising only the use of terms of order 0 and 1 and of the coefficient making it possible to account for the contribution of the terms of higher order, is carried out for a series of 30 disturbances corresponding to a range of value of reactivity going from 66 to 1153 cfm

(1 pcm = 10-5).

  The statistical results of these determinations are reported in the following table which compares the values of the deviations for the calculations comprising respectively terms of order 3, order

  4, and the value determined according to the invention.

  Error Error Relative relative error relative 3rd order% 4th order% according to the invention axial offset 3.06 1.12 1.08 power 4.36 2.01 1.79 local relative anti 3.74 1.78 0.79 reactivity reserve of 0.73 0.40 0.13 reactivity The preceding results confirm the effectiveness of predicting the results of the disturbances, and consequently of the operation of the reactor according to the technique of the invention. The comparison of the values obtained using the simplified operator comprising the coefficient 3 determined beforehand, in fact leads to smaller relative errors than those which are obtained by a much heavier processing integrating the terms of order 3 and even d order 4. The use of this simplified technique is therefore perfectly suited to steering which best follows the evolution of the network load while meeting the priorities chosen in terms of the objective of the reactor operating conditions (minimization of movements control bars, optimization of

safety margins, etc.).

  Furthermore, this simplification is such that it makes it possible to increase the number of parameters monitored while using calculation means.

relatively light.

l1

Claims (8)

  1. Method for controlling nuclear reactors comprising: - the use of sensors for measuring operating conditions, - the determination from these measurements of specific quantities of the operation of the reactor, known as observables, - the comparison of the quantities thus determined with data previously established to guarantee the operation of the reactor according to methods chosen by the operator, - the programmed development of responses, from this comparison, - the sorting by an expert system of responses exceeding a pre-established threshold, leading at the next stage of processing by means of calculation, - processing by means of calculating the responses established as a function of the differences observed between the measured values and the quantities previously established, in order to predict the consequences of these responses on the state final following such a response, - the use, in the establishment of forecasts ions, by the means of calculation of the generalized perturbation method in which, in expressions comprising a polynomial Boltzman expansion, only the terms of degree 0 and 1 are preserved, assigned a coefficient which accounts for the average contribution of higher modes for   the asymptotic values of these parameters.   - the comparison of these forecasts with the previously established data, and the leading choice, among the possible answers, of the one answering the best to the selected terms.   2. A control method according to claim 1, in which the coefficient accounting for the contribution of the higher orders of the Boltzman operator is obtained for each parameter by a prior determination of the value of the parameter corresponding to order terms greater than 0 and 1, terms in limited number, making it possible to approach the asymptotic value of the parameter concerned, this coefficient being expressed by the expression: p = So - Se S - S in which S is the asymptotic value of the parameter SO and S1 are respectively the values of the terms of   degree 0 and 1 of the Boltzman operator.   3. A control method according to claim 1 or claim 2 in which a periodic adjustment is established consisting of the determination of the actual operating conditions of the reactor, from instantaneous measurements, this adjustment constituting the initial state from which the forecasts are established.   4. A control method according to claim 3 in which the registration reference values are determined from the measurements of the parameters by a   technique for minimizing deviations.   5. Piloting method according to one of   previous claims in which the model   representing the running of the reactor is a model three-dimensional.   6. Piloting method according to one of   previous claims in which at least one   measured parameter is constituted by the neutron flux at different points in the reactor.   7. Control method according to one of   previous claims wherein the introduction   in the establishment of forecasts, an element representing the terms of degrees greater than 0 and 1, is limited to disturbances whose magnitude exceeds one   certain threshold previously established.   8. Control method according to one of   claims 4 to 7 wherein the registration comprises   the use of a generalized disturbance method, to identify in the model, the parameters contributing significantly to discrepancies noted.