Real-time Pricing Method Research Based On Improved Dual Decomposition For Smart Grid

Smart grid is the development direction of future power grid with reliability, highquality, high efficiency, compatibility, interaction and so on. Real-time pricing is anideal pricing mechanism, which can save power and protect environment, clip peakand fill valley, guarantee the users’ and power provider’s maximum benefits, and cancomplete the demand side management, encourage users consuming wisely, and caneffectively solve the balance problem of the smart grid. However, usually subgradientalgorithm used based on the dual decomposition has difficulty in adjusting step lengthwhen solving the real-time pricing model. What’s more, when the network scale islarge, the algorithm’s convergence rate is slow and even not convergent. Theimproved dual algorithm given in this paper, that is, proximal center algorithm notonly has the separability, but also does not need to adjust the step length. When thenetwork scale is large, our algorithm has faster convergence rate and betterconvergence performance.The main contents and the innovation points are as follows:1) Establish the real-time pricing model with independent time slots, that is,we maximize the minus between the users’ total utilities and the cost of the powerprovider subjected to the total user’s consumptions are not more than the total powersupplied. We use the proximal center algorithm to solve this model and also analyzethe algorithm’s convergence. Matlab simulation results validate that the proximalcenter algorithm has faster convergence rate, compared with subgradient algorithmbased on the dual decomposition.2) According to some users’ power requirement, we introduce the factor of slotscoupling which make smart grid provide users with certain requirements of qualityassurance. Then we give a corresponding real-time pricing model. We also use theproximal center algorithm to solve this problem. By Matlab simulation to validate thealgorithm, when the network scale is large, it also has the advantage of fasterconvergence rate than subgradient algorithm.3) The users’ utility function is usually concave. In fact, the users’ utilities maybe not concave. If we use the logarithm to realize their utilities,“0” utility will occurwhen its consumption is not zero. In order to avoiding this situation, we establish utility-oriented fairness real-timepricing model. Use the concept of “pseudo utility” to realize utility-fairness. We useproximal center algorithm to solve this model. Matlab simulation results show that theusers’ utilities and power consumption level under utility-fairness in each slot arebetter than those under proportional-fairness.