The ability of the oil industry to optimize production of a reservoir relies on the possibility of evaluating the well effluent at regular intervals, in terms of quantity (flow rate) and of composition (the proportions of the various phases). This makes it possible to determine what corrective action may need to be taken. However, measuring the flow rate of oil well effluent is a problem that is complex because of the way effluents are usually made up of three phases, and because of the changes in flow conditions to which they are subject (pressure, temperature, shape of pipes). These factors give rise to a wide variety of flow regimes being observe, including some regimes of highly non-uniform and unstable character, with the proportions of the phases in the fluid mixture being capable of varying very considerably both in the flow direction (i. e. over time) and across the flow direction, in the form of phase stratification across the flow section. One extreme, but very common, case is slug flow i. e. that of a high gas content with the flow being made up of an alternation of portions that are essentially gas, known as"pockets", and portions that are constituted essentially by liquid, known as"plugs".

In the oil industry, the traditional practice is to separate the effluent into its component phases and to perform measurements on the phases separated in this way. However that technique requires separators to be installe on site, where separators are bulky and expensive items of equipment, and while wells are being tested, it also requires additional pipes to be put into place.

Numerus proposas have been put forward for developing techniques that would make it possible to avoid using such separators. A description of these developments is

to be found in SPE publication 28515 (SPE Annual Technical Conference, New Orleans, September 25-28,1994) by J. Williams,"Status of multiphase flow measurement research".

Most of the propositions suggest firstly a total flow rate sensor and secondly sensors for measuring the proportions of the phases in the mixture.

Amongst those proposas, US patent 4,788,852 describes apparats comprising a Venturi and a device for measuring gamma ray attenuation at three different energy levels, the device being situated at the constriction of the Venturi.

British patent application 2 128 756 explains that it is necessary to determine two of the following three magnitudes: total mass flow rate, total volume flow rate, and mean density of the fluid. In order to compensate for non-uniformities in the flow, which give rise in particular to differences of speed between the gas and the liquid phase, thereby making any flow rate measurement difficult, and also making density measurements inaccurate, it proposes homogenizing the fluid upstream from the sensors by means of an appropriate device. In addition to improving the quality of mean density measurement, homogenization has the effect of equalizing the speeds of the gas and liquid phases, and thus of enabling the gas flow rate to be measured.

An application of that principle is described in document WO 90/13859 which provides firstly a Venturi and a gamma ray density meter placed at the constriction of the Venturi, and secondly a mixer upstream from the Venturi for the purpose of homogenizing the multiphase fluid entering the Venturi. Nevertheless, the cost and the size of such apparats may limit its commercial applications.

The invention seeks to characterize three-phase oil effluents by means that are simple and cheap, and that are applicable to a wide variety of flow regimes, and in particular to slug flow.

In one aspect, the invention provides a flow rate measurement method adapted to oil effluents made up of multiphase fluid mixtures comprising water, oil, and gas, the method comprising the following steps: the effluent is passed through a Venturi in which the effluent is subjected to a pressure drop; a mean value of the pressure drop is determined over a period t, corresponding to a frequency fl that is low relative to the frequency at which gas and liquid alternate in a slug flow regime; a mean value is

determined for the density of the fluid mixture at the constriction of the Venturi over said period ti; and a total mass flow rate value <Q> is deduced for the period ti under consideration from the mean values of pressure drop and of density.

Appropriately, the density of the fluid mixture is measured by gamma ray attenuation at a first energy level at a frequency f2 that is high relative to said frequency of gas/liquid alternation in a slug flow regime, and the mean of the measurements obtained in this way over each period ti corresponding to the frequency f, is formed to obtain said mean density value.

Appropriately, the frequency f, is 0.1 Hz or less than 0.1 Hz, e. g. 0.01 Hz.

Appropriately, the frequency f2 is greater than 20 Hz, and preferably greater than 40Hz, e. g. being equal to 45 Hz.

The invention will be well understood on reading the following description made with reference to the accompanying drawings, in which: Figure 1 is a diagram of flow measuring apparats suitable for oil well effluents; and Figure 2 is a block diagram illustrating the data processing operations performed by the apparats of Figure 1, in two embodiments.

Oil effluents are usually made up of a multiphase mixture of liquid oil, of gas (hydrocarbons), and of water. Below we use the following notations: the symbols Q <BR> <BR> <BR> <BR> <BR> and q designate mass flow rates and volume flow rates respectively; the symbol p designates density; the symbols oc and y designate the static and dynamic proportions of the various phases; and the indices o, w, g, and l refer respectively to the oil, water, gas, and liquid phases (where the liquid phase is the oil and the water taken together), while the index in designates the fluid mixture.

The device comprises a pipe section 10 comprising a convergent Venturi 11 whose narrowest portion 12 is referred to as the throat. In the example shown, the section of the pipe 10 is dispose vertically and the effluent flows upwards, as symbolized by arrow F.

The constriction of the flow section in the Venturi induces a pressure drop Ap between level 13, situated upstream from the Venturi at the inlet to the measurement section, and the throat 12. This pressure drop is associated with the total mass flow rate Q and with the density pm by the following equation:

K.Q2 #p = + #m.g.hv #m<BR> m where g is the acceleration due to gravit, hv is the distance between the upstream level 13 and the throat 12, and K is a constant associated essentially with the geometry of the Venturi, and which is given by: <BR> <BR> <BR> <BR> <BR> 1-ß4<BR> <BR> <BR> K=<BR> <BR> 2C2.A2<BR> <BR> <BR> <BR> <BR> <BR> <BR> where ß is the constriction ratio of the Venturi, i. e. the ratio between the diameter of the throat and the upstream diameter of the Venturi, C is the discharge coefficient, and <BR> <BR> <BR> <BR> <BR> A is the section of the throat. The term pm. g. hv is generally small or negligible. By writing Ap* = Ap-p.. g. hv, equation (1) becomes: Q-gOp*,Pmlh where k = K-½.

In a preferred embodiment, the ratio p is 0.5. With a pipe having a diameter of 10 cm, the diameter of the throat is 5 cm. The discharge coefficient C is about 1. This coefficient depends to a small extent and in predictable manner on the properties of the fluid. Traditionally, this corrective effect is taken into account by the Reynolds number.

The pressure drop Ap is measured by means of a differential pressure sensor 15 connecte to two pressure takeoffs 16 and 17 opening out into the measurement section respectively at the upstream level 13 and in the throat 12 of the Venturi. In a variant, the measurement may also be performed by means of two absolut pressure sensors connecte to the pressure takeoffs 16 and 17, respectively.

The density pm of the fluid mixture is determined by means of a sensor which mesures the attenuation of gamma rays, by using a source 20 and a detector 21 placed on opposite sides of the Venturi throat 12. The throat is provided with"windows"of a material that shows low absorption of photons at the energies under consideration.

The source 20 produces gamma rays at two different energy levels, referred to below as the"high energy"level and as the"low energy"level. The detector 21 which comprises in conventional manner a scintillator crystal such as NaI and a photomultiplier produces two series of signals Whj and W, O referred to as count rates,

representative of the numbers of photons detected per sampling period in the energy ranges bracketing the above-mentioned levels respectively.

These energy levels are such that the high energy count rate Whj is essentially sensitive <BR> <BR> <BR> <BR> to the density pm of the fluid mixture, while the low energy count rate Wl,, is also sensitive to the composition thereof, thus making it possible to determine the water content of the liquid phase.

Preferably the high energy level lies in a range 85 keV to 150 keV. For characterizing oil effluent, this energy range presents the remarkable property that the mass attenuation coefficient of gamma rays therein is substantially the same for water, for sodium chloride, and for oil, being about 0.17 cm2/g. This means that based on the <BR> <BR> <BR> <BR> high energy attenuation, it is possible to determine the density pu ouf the fluid mixture without the need to perform auxiliary measurements to determine the properties of the individual phases of the fluid mixture (attenuation coefficients and densities).

The attenuation measured by the detector 21 is expressed by the following equation: A = D.Vm.m <BR> <BR> <BR> <BR> where Dv is the distance travelled through the fluid, i. e. in this case the diameter of the Venturi throat, and vm is the mass attenuation coefficient of the fluid mixture.

Since the mass attenuation coefficients of water and oil in the above-indicated energy range are substantially identical, and since the contribution of the gas is negligible because of its very low density, the mass attenuation coefficient vm, and thus the product Du. vu that appears in equation (3) can be considered as being substantially constant and independent of the densities pO and p, of the oil and water phases.

Under such conditions, the high energy attenuation Ahj is a very advantageous indicator of the density pm of the mixture.

A material that is suitable for producing high energy gamma rays in the energy range under consideration, and low energy rays is gadolinium 153. This radioisotope has an emission line at an energy that is approximately 100 keV (in fact there are two lines around 100 keV, but they are so close together they can be treated as a single line), and that is entirely suitable for use as the high energy source. Gadolinium 153 also has an emission line at about 40 keV, which is suitable for the low energy level that is used to determine water content. This level provides good contrast between water and oil,

since the attenuation coefficients at this level are significantly different, typical values being 0.228 cm2/g for oil and 0.291 CM2/g for sea water.

Figure 1 also shows a pressure sensor 22 connecte to a pressure takeoff 23 opening out into the throat 12 of the Venturi, which sensor produces signals representative of the pressure pv in the throat of the Venturi, and a temperature sensor 24 producing signals T representative of the temperature of the fluid mixture. The data pv and T is used in particular for determining gas density pg under the flow rate conditions and gas flow rate qg under normal conditions of pressure and temperature on the basis of the value for the flow rate under the flow rate conditions, determined in a manner described below. In this respect, it is preferable for the pressure to be measured at the throat of the Venturi. In contrast, it does not matter where temperature is measured.

The information coming from the above-mentioned sensors is applied to a data processing unit 30 constituted by a computer running a program for delivering the looked-for results by performing various treatments based on the principle explained below. In the explanation, reference is made to the block diagram of Figure 2.

1-Determinin total mass flow rate The need is to determine the total mass flow rate Q in a manner that is adapted to the particularly difficult case of the effluent flow regime being of the pocket-and-plug type, also called slug flow i. e. when the flow is in the form of alternating portions made up essentially of gas, known as"pockets", and portions made up essentially of liquid.

A mass flow rate value <Q>, referred to below as the mean mass flow rate, is determined (block 35) by applying above equation (2) to values defined as means <BR> <BR> <BR> <BR> <Ap*> and <pm> of the pressure drop Ap and of the density of the mixture pu over a time interval t, which is long relative to the gas/liquid alternation period of a pocket- and-plug flow: Q>-k(p*»Pml/z It has been found that this method gives results that are robust and of satisfactory accuracy, contrary to what might have been expected, given that equation (2) is non- linear and equation (4) is a priori only applicable to a single-phase liquid since the

values both of pressure drop Ap and of density pm are subject to extremely sudden variations in a slug flow made up of pockets and plugs. Such a method of determining mass flow rate transforms a multiphase flow, even a flow that is extremely non- uniform such as slug flow, into a virtual single-phase flow, in that it makes equation (4) above, that applies to single-phase flows, applicable to multiphase flows. The method can thus be depicted as performing a"virtual homogenization".

The period of gas/liquid alternations in a slug flow can vary widely, depending on circumstances: typically it lies in the range 0.1 to 10 seconds. The duration ti over which the above-mentioned mean values are defined must be long relative to said alternation period. It is appropriate to set it at a value that is not less than the upper limit of the above-mentioned range, i. e. 10 s, which corresponds to a frequency, written f1, of 0.1 Hz.

In the preferred implementation of the invention, these mean values are obtained as follows.

As explained above, the density is determined from the high energy attenuation Whj at a frequency f2 which is high relative to the gas/liquid alternation frequency in a slug flow (block 40). Starting from values p, n (i) respectively obtained for intervals t2 <BR> <BR> <BR> <BR> corresponding to the frequency f2, a mean density value <pm> is calculated over a duration corresponding to above-mentioned frequency f, (block 41). Such a method is adapted to the non-liner character of the equation relating the count rate Wh ; to the <BR> <BR> <BR> <BR> density pm of the fluid mixture. It is appropriate for the frequency f2 to be at least 20 Hz, preferably at least 40 Hz, and for example 45 Hz.

Measurements are acquired at such a sampling frequency by means of a specialized electronic circuit associated with the detector 21 but not shown in Figure 1. The circuit is implemented in accordance with the teaching of US patent 5,349,195, and it need not be described in detail herein. Such a circuit provides fast and reliable processing of the pulses sensed in each energy window.

The inherent response time of the differential pressure sensor 15 is considerably slower than the time corresponding to the frequency f2, for example it is about 0. 5Hz.

However, to simplify implementation, differential pressure measurement samples are

similarly acquired at the frequency f2 (45Hz in the example described) and the desired mean value <Ap> is obtained by averaging the measured samples #p (i).

The mean value <Ap*> used in determining the total mass flow in application of above equation (4) is obtained from the mean values <#p> and <#m>.

2-Determining water content The detector 21 mesures attenuation of gamma rays by the fluid mixture at the two above-mentioned energy levels: Whi = Whi,0exp(-Ahi,m) Wlo = Wlo,0exp(-Alo,m) where A stands for attenuation and W,, ;, o and Wlo,0 are the high and low count rates in the absence of attenuation, as obtained by calibration.

As already mentioned above, the attenuation of the gamma rays by the fluid mixture is as follows: Am = Dvvmepm where Dv is the diameter of the throat of the Venturi and vn, is the mass attenuation coefficient of the fluid mixture.

Given that the density of the gas phase is very low, it is possible to make the following approximations: #m # #1 and pm = (xi. pi where v, is the mass attenuation coefficient of the liquid (oil and water together) and α1 is the volume fraction of the liquid, and hence: Am # Dv.#1.#1.α1 The mass attenuation coefficient of the liquid v, is a function of the attenuation coefficients of oil and water vu and vw, and of the water/liquid mass ratio (WLR): #o+WLR(Vw-#o)#1= The water/liquid mass ratio WLR can be determined from the high and low energy attenuations in the mixture, given the attenuations in oil and in water as determined by calibration, by using the following expression: WLR = #o)/(#w-#o)-

where #m, #o, and 7v,", are the ratios of high energy attenuation over low energy attenuation for the fluid mixture, for oil, and for water, respectively: km=Ahi,.Al., m #o = Ahi,o/Alo,o Ahi,w/Alo,w#w= The values of the water/liquid mass ratio WLR (i) are thus calculated for each of the sampling intervals t2 (i) corresponding to the frequency f2 which is equal to 45 Hz in the example described, on the basis of the measurements Whi(i) and wlo (i) acquired over the interval under consideration (block 50). Thereafter (block 51) a mean value <WLR> is calculated over the interval t, corresponding to the frequency f, at which the mass flow rate Q is determined, with the values WLR (i) being weighted by respective confidence coefficients C (i).

The water ratio WLR is relative to the quantity of liquid in the fluid mixture. The measurement of WLR therefore applies to this quantity of liquid. When the gas content is high, the liquid content is small. Also attenuation is low because the density of the fluid mixture is low. Under such circumstances, the measurements relating to the proportions of water and oil in the liquid cannot be of good quality. The confidence coefficient should therefore decrease with increasing gas content. <BR> <BR> <BR> <BR> <P>Since the density pu ouf the fluid mixture is a sensitive indicator of its gas content, the confidence coefficient C (i) is advantageously determined as a function of the density pm. (This choice is justifie by the assumption that there is no systematic correlation between the composition of the liquid as expressed by the water fraction WLR, and <BR> <BR> <BR> <BR> the gas content as given by the density pm of the mixture.) Such a function can have several different forms. By way of example, the confidence coefficient may merely be <BR> <BR> <BR> <BR> equal to the density pm. Naturally, other types of function can be envisage. Also, the<BR> <BR> <BR> <BR> <BR> <BR> parameter used could be the variation in the density pn, rather than the density proper, or a combination of these factors could be used.

To sum up, each value of WLR (i) is associated with a confidence coefficient C (i) determined as a function of the density pm (i) obtained as described above, and the mean value <WLR> is a weighted mean of the values WLR (i):

<WLR> =(# C (i). WLR (i))/E C (i)<BR> where z symbolizes the sum of the values lying in the interval t, under consideration.

From the mean value <WLR> obtained for the mass fraction of water, it is possible to determine the density of the liquid phase p, and the mean value of the volume fraction of water <wlr>, knowing the densities p, and po of water and oil respectively in the effluent. The density values pw and pO and the density of gas pg (under the flow conditions) can be obtained in various ways by auxiliary measurements that are known per se. By way of example, for the liquid phases, mention can be made of the sampling device described in French patent application 96 14292 filed on November 22,1996.

The density of the liquid phase is obtained (block 52) from the following equation: 1/pl = <WLR>/pw + (1-<WLR>)/po and the mean volume fraction of water (block 53) is given by: <wlr> = #w/#l.<WLR> It should be observe that the water content of the liquid phase is a characteristic of the effluent which is stable (or at any rate which varies slowly), unlike the gas content of the mixture. It is therefore appropriate to determine pi at the frequency fl, as a function of the mean value <WLR>.

In addition, starting from the density of the liquid phase pi determined as described above, and from the density of the gas phase pg under flow conditions, it is possible <BR> (block 54) to determine the proportions of liquid cc, and of gas otg at the frequency f2 using the following equations: <BR> (#m(i)-#g)/(#l-#g)αl(i)= <BR> a, +a,=1 3-Determining the oil and the water flow rates The gas mass flow rate Qg is negligible relative to the liquid mass flow rate because of the low density of the gas. It can therefore be assumed that the liquid mass flow rate Q, is approximately equal to the total mass flow rate Q: Qi=Q

The oil and water mass flow rates can then be determined (block 60) approximately from the following equations: Qw = <WLR>. <Q> and QO = <Q>-Qw The water and oil volume flow rates qw and qO are deduced from the mass flow rates <BR> <BR> <BR> <BR> Qw and Qo (block 61) given the respective densities p, and pO of water and oil in the effluent: qw = Qw/r,, and q. = Q./p.

The volume flow rate of the liquid q, can then be determined as the sum of the oil and water flow rates: q-qo+qW or equivalently: q1 = Q/#1# The values obtained in this way are approximate since they assume that the gas flow rate is negligible. For more accurate determination, account is taken of the presence of gas, expressed by the volume fraction of the gas. Since the speed of gas is higher than that of the liquid (a phenomenon known as"slip"), it is necessary to make use not of the liquid and gas proportions (xi and αg as deduced in the manner given above from the gamma ray attenuation measurements since those proportions are static proportions, but instead to make use of dynamic proportions y, and yg (withy, + yg = 1) which take the slip phenomenon into account, which proportions are respectively lower than and greater than the static proportions oc, and αg.

This is expressed by the equation: αg+#g(αg)γg= where 8g (ag) is the slip term and represents a positive correction.

When dealing with values at the throat of the Venturi, it has been found that the dynamic proportions can be determined as a function solely of the gas proportion ocg and are therefore essentially independent of flow rate. A suitable example of the empirical equation making such determination possible is of the form: Pyg (i) = au (i) + A (γl(i))m(1 - γ1(i))n where the coefficient A and the exponents in and n are determined empirically. A suitable example corresponds to the following values: A = 1, in = 1, n = 6. This

equation makes it possible by iterative computation to determine the dynamic proportion yi from the static proportion (xi (block 55).

Using the values of γ1 (i) as determined at the high frequency f2 (45 Hz in this example), a mean value <^yl> is calculated (block 56) over each interval tl (i. e. 10 s in the example described) corresponding to the frequency fl.

The liquid volume flow rate q, can then be obtained at frequency fl from the total mass flow rate <Q>. The total volume flow rate, taking account of slip, is given (block 70) by the following equation: q = Q/PH where the magnitude PH is a value for the density of the mixture, known as the "homogeneous density", that takes account of the slip phenomenon, and that is obtained (block 57) from the equation: PH = <71>Pl + <γg>#g and the liquid flow rate q, is given (block 71) by: qm q'Y From which the water and oil flow rates are deduced (block 72): <BR> <BR> <BR> <BR> qw = <wlr>ql<BR> <BR> <BR> <BR> <BR> <BR> qO = (1-<wlr>) ql Determininggasflowrate4- To determine the gas flow rate, a suitable method, different from that described above, relies on an"instantaneous"value for the total mass flow rate Q (i) calculated (block 80) for each sample corresponding to the high frequency f2 (45 Hz in the example described) in application of the following equation: Q (i) = kf (P>-Pm (i). g. hv).#m(i)]½ with the notation being the same as above.

An"instantaneous"value for the total volume flow rate is then deduced (block 81) : q (i) = Q (i)/#H (i) where pH (i) is the"homogeneous density"as defined above, determined for each sample at the high frequency f2 (block 58) as follows:

PH (i) = 71 (i) Pl + yg (i) pg Based on this value for the total volume flow rate, an"instantaneous"value for the gas flow rate qg (i) is determined (block 82) for each sample at the high frequency f2: qg (i) =) tg (i) 1(i) and finally the gas flow rate qg is calculated by averaging the values qg (i) over each interval ti (10 seconds in the example described) corresponding to the low frequency f, (block 83).