| Since the new round of reform on electric power system,China’s electricity market has been deeply developed with the gradual improvement of market mechanism,the expansion of transaction scale and the significant improvement of the operation ability of market entities.As the key link between power grid and users,electricity retailers are facing more incentive market competition and greater risks in diversified business segments.The traditional simple and extensive operation mode of electricity retailers,i.e.,the revenue-sharing model between electricity retailers and users based on the price difference between electricity wholesale and retail market,will be difficult to adapt to the market environment in the new stage.Therefore,electricity retailers urgently need to build up the concentrated drive-in market segments with new concepts,and achieve high-quality development by optimizing the decision-making of electricity purchase and sales,innovating retail varieties and improving service levels.Based on the electricity sale side reform under the new-type power system in China,the operation strategies of electricity retailers in diversified electricity market segments,such as mid-and long term electricity trading,electricity retail and demand response(DR),are studied in this dissertation.Economic theory,game theory,multiattribute evaluation,and data analysis technology have been widely introduced to optimize the decision-making of retailers in electricity markets so as to improve their operating income.The main work of this dissertation includes:(1)The optimal alliance strategy of deviation mutual insurance(DMI)is proposed for electricity retailers to deal with the energy deviation settlement mechanism in China’s mid-and long term electricity market.First of all,the alliance cooperation cost model involving the cost of information search and sharing,negotiation and risk management is constructed.Then,cooperative game and resource dependence theorybased indexes,such as marginal income contribution,deviation exemption margin,irreplaceability,and cooperation preference of electricity retailers,are proposed to allocate the alliance income.Entropy weight and analytic hierarchy process are used to comprehensively weight the indexes,so as to evaluate the bargaining power of different electricity retailers and clarify the income allocation scheme of the alliance.Finally,an alliance strategy optimization model considering the utility of electricity retailers and the cooperation stability of the alliance is constructed.Case studies illustrate that electricity retailers can reasonably participate in the DMI alliance according to their own scale and deviation rate,so as to reduce the deviation penalty.(2)For the needs of electricity retailers to improve market share and sustainable profitability,a peak-valley combination electricity plan is proposed as a new product and the corresponding optimization method is proposed with the bounded rationality of end-users considered.The new retail mode based on the combination of peak and valley time-sharing electricity quota and its electricity bills discount scheme are designed at first.Then,behavioral economics is introduced to explain the bounded rational behavior of users in reality,and the plan purchase and power response model of user are constructed based on herd effect and anchoring effect,respectively.Finally,the bi-level optimization model for the proposed peak-valley combination electricity plan design is constructed,aiming at maximizing the electricity purchase and sale revenue of retailers in the mid-and long term electricity market.Case studies illustrate that the proposed peak-valley combination electricity plan can reduce the electricity purchase cost of retailers in the mid-and long term electricity market by encouraging user to actively participate in load response,and strive for a win-win result.(3)Facing the market competition demand of electricity retailers to identify users’ differentiated energy consumption behavior and provide targeted electricity plans,an electricity plan recommendation method is proposed based on the implicit scoring of electricity plan and the portrait of users.Implicit scoring refers to user’s historical purchases of electricity plans,which reflects their preference for the detailed features of a specified plan.Combined with the labels of typical electricity plans in foreign electricity retail market,a label-based user portrait model is constructed with the decline of user preference over time considered.Pearson correlation coefficient and Euclidean distance are introduced for the two-scale similarity clustering of user load profiles.And the weighting model of electricity plan labels is constructed based on the clustering results and the silhouette coefficients of the labels,which reflects the relationship between different labels and user load.A user similarity evaluation model is established based on Euclidean distance with weighted labels,and then the recommendation method of electricity plan is proposed based on user portrait and collaborative filtering.Case studies illustrate that the proposed method can enable electricity retailers to explore users’ consumption preferences according to their historical purchase information,so as to improve the accuracy of electricity plan recommendation.(4)In order to improve the load integration value-added service and profitability of electricity retailers,an optimization method of demand response(DR)transaction strategy is proposed based on user load response characteristic recognition.Aiming at the current transaction mechanism of DR pilot in China,an optimization model of DR bidding and scheduling strategy of electricity retailers is constructed with user load rebound effect and the uncertainty of user response probability and capacity reduction considered.To realize the recognition of user load response characteristics and support the optimization of retailers’ strategy,a label-based demand resource pool(DRP)is proposed with user different characteristics covered.A DRP data completion algorithm based on Bias-singular value decomposition matrix factorization is proposed to preliminarily predict user unknown characteristics.Bayesian inference-based DRP data modification algorithm is proposed to realize the refined recognition of user load response characteristic according to the newly added DR samples and the experience knowledge of electricity retailers.Case studies illustrate that the proposed two-stage method has lower error in user load response characteristic recognition,and can provide reliable data support for the DR transaction strategy optimization of electricity retailers,so as to reduce their loss risk at the early stage of DR transactions.The main work of this dissertation is beneficial for electricity retailers to optimize their strategies in diversified business segments such as mid-and long term electricity trading,electricity retail and DR transactions.As a result,electricity retailers can reduce the electricity purchase cost in the wholesale market,increase the share of retail market,enhance revenue,broaden profit channels and improve the value-added service. |
