SELF-ADJUSTING REGULATOR

摘要

1437035 Regulators ALLMANNASVENSKA ELEKTRISKA AB 20 July 1973 [21 July 1972 13 Nov 1972 24 Jan 1973 3 May 1973 4 May 1973] 34604/73 Heading G4A A self adjusting regulator in which the relationship between the input and output signals is of the form y(t+k)=AZ(t)+e(t+k) (where y(t) is the output at time t, A=[b, #] where b, # are transfer coefficents, ZT(t) = [u(t), QT(t)] where u(t) is the input at time t, QT(t) is an observation vector containing inter alia old values of x, y and e(t) is a sequence of stochastical variables constituting disturbance in the system), operates in accordance with the following algorithm:- The calculations are performed using the computing circuitry of Fig. 2 in which from output signals y(t), y(t-1) . . . stored in shift registers 1071 and input signals x(t-1), x(t-2) ... stored in shift register 1072, the vector ZT is derived and entered into memory A 10612, The parameth ‰(t-1) is stored in memory 10602 and is fed together with the vector ZT to calculating unit 103 which derives the scalar product ‰(t-1)ZT and feeds it to calculating unit 105 which derives V(t) and feeds it to memory 10613. A vector R=Q 2 is calculated and stored in memory 10606 from which it is fed to calculating unit 105 together with ZT to form the scalar product R.ZT. Calculator 105 then forms the sum D of this product and the constant W, (W, which lies between 0 and 1, being introduced so that newer observations have greater weight than older ones). D is stored in memory 10614 and used in the calculation of K=R/D which is entered in into memory 10605 and used in the computation of the product Ki.V. This product together with the parameter ‰i is fed to calculating unit 105 to derive ‰i+KiV which is then stored in memory 10602 as the updated parameth. By forming the computation (Qi 1 -Ri.Ki)/w in calculating units 104, 105 and storing the result in memory 10601 the element Qi 1 is also updated, this process being repeated for all the elements of the matrix Q. The magnitudes #(t) and #(t-1) are then formed and stored in memories 10615, 10616 respectively. Calculating unit 104 computes the product of the con- tents #q c 2(t) of memory 10604 which represent the estimated loss in the previous cycle and the contents #(t-1) of memory 10616. To this is added, in calculating unit 105, V2/D, the sum after multiplication by w being divided, in calculating unit 104, by #(t) to produce the new estimated loss q e 2(t) for storage in memory 10604, and after division by #b for storage in memory 10613. The sum ##i+q 12i #q e 2(t)/#b is then derived by operating on the contents of memories 10613, 10601, 10602, the sum #b+q 11 q2 e (t)#b being analogously derived. From these values and the observation vector #(t) the input control valve u(t) is computed. It may be checked to see if it falls within a permitted range, being given a predetermined value if it does not. Several such computing devices may be combined (Fig. 2, not shown) to control a multivariable system.