The development of sustainable generation, a reliable electricity supply and affordable tariffs are the primary requirements to address the uncertainties in different future energy scenarios. Due to the predicted increase in Distributed Generation (DG) and load profile changes in future scenarios, there are significant operational and planning challenges facing netwrok operators. These changes in the power system distribution network require a new Active Network Management (ANM) control system to manage distribution constraint issues such as thermal rating, voltage, and fault levels. The future smart grid focuses on harnessing the control potential from demand side via bidirectional power flow, transparent information communication, and contractual customer participation. Demand Side Management (DSM) is considered as one of the effective solutions to defer network capacity reinforcement, increase energy efficiency, facilitate renewable access, and implement low carbon energy strategy. From the Distribution Network Operator's (DNO) perspective, the control opportunity from Demand Response (DR) and Decentralized Energy Resource (DER) contributes on capacity investment reduction, energy efficiency, and enable low carbon technologies. This thesis develops a new decentralized control system for dealing effectively with the constraint issues in the Medium Voltage (MV) distribution network. In the decentralized control system, two novel control approaches are proposed to autonomously relieve the network thermal constraint via DNO's direct control of the real power in network components during the operation period. The first approach, Demand Response for Power Flow Management (DR-PFM), implements the DSM peak clipping control of Active Demand (AD), whilst the second approach, Hybrid Control for Power Flow Management (HC-PFM), implements the hybrid control of both AD and DER. The novelty of these two new control algorithms consists in the application of a Constraint Satisfaction Problem (CSP) based programming model on decision making of the real power curtailment to relieve the network thermal overload. In the Constraint Programming (CP) model, three constraints are identified: a preference constraint, and a network constraint. The control approaches effectively solve the above constraint problem in the CSP model within 5 seconds' time response. The control performance is influenced by the pre-determined variable, domain and constraint settings. These novel control approaches take advantages on flexible control, fast response and demand participation enabling in the future smart grid.