Due to the difficulties of handling non-linearities in many large systems to which on-line optimal control is being applied, many applications have had to be restricted to the linear model-quadratic cost case. In particular, if the calculations are performed in a decentralized manner, the sub-system problems must yield a rapid solution and in the simple linear-quadratic case analytic solutions to these sub-problems may be obtained. The method of quasilinearization for the resolution of boundary-value problems arising in the solution of non-linear differential equations has been widely developed. This paper examines the use of quasilinearization algorithms for the solution of sub-problems arising in a problem decomposition using Lagrangian duality. The good convergence properties of the algorithms make them particularly useful for the solution of the sub-system problems. The actual improvement in operating costs obtained by handling more general sub-system non-linearities is compared with the increased computational burden for an actual on-line water control scheme.
[1]
R. Bellman.
Dynamic programming.
,
1957,
Science.
[2]
F. Fallside,et al.
Hierarchical optimisation of a water-supply network
,
1975
.
[3]
J. B. Rosen.
The Gradient Projection Method for Nonlinear Programming. Part I. Linear Constraints
,
1960
.
[4]
R. Bellman,et al.
Quasilinearization and nonlinear boundary-value problems
,
1966
.
[5]
Robert E. Larson,et al.
State increment dynamic programming
,
1968
.
[6]
Stuart E. Dreyfus,et al.
Applied Dynamic Programming
,
1965
.
[7]
Ronald A. Howard,et al.
Dynamic Programming and Markov Processes
,
1960
.
[8]
Angelo Miele,et al.
Modified quasilinearization algorithm for optimal control problems with nondifferential constraints
,
1974
.