Optimization model of new energy consumption based on load demand-side response

As a technically and economically feasible and potentially resource-rich regulation tool, demand-side response can effectively strengthen the guarantee of new energy consumption, and is expected to play an enhanced key role in the construction of new power systems. This paper analyzes the factors affecting the linkage between demand-side and new energy generation side, the promotion of new energy consumption by demand-side response, and proposes a method to respond to the change of new energy output from the customer side, using interruptible load, transferable load, shiftable load, and curtailable load strategies to respond to the peak regulation problem brought by new energy access to the grid. The optimal demand-side response strategy is obtained by constructing a new energy consumption optimization model and solving it with the direct non-inferiority solution method, taking into account the economic and new energy consumption objectives. Finally, the effectiveness of the method is verified by simulating the IEEE 37-node system with multiple new energy configurations.


Introduction
With the continuous development of new energy, the installed capacity of new energy is increasing, but the volatility and uncertainty limit the possibility of its direct replacement of traditional fossil energy, and with it comes the dilemma of new energy generation and consumption.In the context of the irreversible replacement of fossil energy by new energy, new energy consumption has become an urgent problem to be solved [1] .As the proportion of installed power of new energy with high volatility and intermittency continues to rise, the shortcomings of China's power system with insufficient flexible regulation resources are increasingly prominent.Demand-side response, as a technically and economically feasible and potentially resource-rich regulation means, can effectively strengthen the guarantee of new energy consumption and is expected to play a key role in the construction of new power systems [2][3] .At present, the main challenges in demand-side response for new energy consumption include the following aspects: (1) Source-grid load-storage characteristics analysis and accurate prediction.
With the increasing penetration of new source-load resources in the power grid, their impact on the power grid is becoming more and more significant.It is necessary to conduct in-depth analysis of the regulation methods and operational characteristics of the power grid itself, especially around the sourcegrid load-storage interaction, and study the role of energy storage system in the grid connection of renewable energy to suppress power fluctuations and the safe and stable operation of the power grid [4- 6] .At the same time, it is necessary to make accurate multi-time scale prediction of new energy output to provide basic support for new energy access to the grid.
(2) User-side flexible load demand response potential assessment.To give full play to the role of user-side flexible load in the multi-faceted interaction of smart grid, quantitative assessment of user demand response potential is required.Usually, when studying the active participation of responsive loads such as smart home appliances in the system active power balancing process, power companies or load aggregators use the quantitative assessment of demand response potential to develop load response instructions to avoid system under-response or over-response [7] .However, current research is often based on several assumptions, such as the potential assessment assumes that the proportion of participating responding loads is a certain fixed proportion, but in practice the operating power of the load is variable, making the user demand response participation also variable, and also affected by a variety of factors, so it is necessary to establish an accurate quantitative model of the short-term potential of demand response [8][9][10][11] .
(3) Coordinated source-network-load-storage control technology considering demand response.Both source and load sides are stochastic and related to time scale.The demand response resources on the user side are scattered and varied, while small and medium-sized loads lack channels to participate in demand response and it is difficult to regulate the grid, i.e. there are obstacles to implement demand response strategies from groups to individuals.Therefore, it is necessary to establish a multiuser interaction model for demand response considering both time and space levels, and to realize reasonable and effective cooperative control of multiusers considering the differences of different load operating characteristics [12] .
To address the above problems, this paper studies the mechanism of demand-side response participation in supply-demand balance and the promotion of demand-side response for new energy consumption, and proposes a method to offset the uncertainty of new energy with demand-response loads.Different types of demand-side response, such as interruptible load size and palletizable load, are used as decision quantities to establish a new energy consumption optimization model considering new energy consumption objectives and economic objectives, and a multi-objective Pareto optimal solution is solved by genetic algorithm to obtain the optimal demand response scheme [13][14] .

Customer response resources and new energy grid-connected power
The response resources of users can be divided into response power and response capacity.In the long term, the user response power should be able to balance the new energy grid-connected power.In the short term, in order to ensure the balance of supply and demand and stable operation of the system, the new energy access capacity should not exceed the expected user response capacity.Analyzing the influencing factors of response resources is helpful to explore the methods that can control the balance power and capacity in the long and medium term (quarter, month, week), short term (1~24h) and ultra short term (10min~1h).

Classification of demand response
According to the FBBC (Federal Energy Regulatory Commission) taxonomy, demand-side response is divided into incentive-based demand response and price-based demand response.Incentive-based demand response is the direct use of incentives and compensation to motivate and induce customers to participate in the various load reduction programs required by various systems.Price-based demand response allows end-consumers to directly face price signals based on time and spatial location, and to make their own arrangements and adjustments in the timing and manner of electricity consumption.Incentive-based demand response includes direct load control programs, interruptible or curtailable load programs, demand-side bidding programs, power buyback programs, emergency demand-side response programs, capacity market programs, and ancillary service market programs; price-based demand response includes time-of-use tariffs, key peak-load tariffs, and real-time tariffs.

Factors influencing user response resources
In order to ensure user response power and capacity, the implementation of demand-side response can start from the following aspects: ensuring user information security; designing reasonable incentives and penalties; and introducing load aggregators.
1) Ensure user information security.User response information involves its cost management, and keeping user's load information confidential is the most basic requirement for obtaining reliable response resources.
2) Design reasonable incentives and penalties.Reasonable rewards can ensure the power and capacity of user response, while reasonable penalties can play the role of forced response to ensure the reliability of user response resources.
3) Introduce Load Aggregator (LA).In foreign countries, a load aggregator is a third-party enterprise that provides load services for scattered customers and is independent of the grid.The customer pays a small fee and the load aggregator manages the load for him, ensuring his production on the one hand and gaining benefits for him through demand response on the other.The load aggregator can provide various types of stable load response resources (long and medium-term stable, short-term stable and ultra-short-term fast).

Rolling balance of responsive resources and new energy generation
In order to accommodate the intermittent and fluctuating nature of new energy sources, it is necessary to establish a power balancing mechanism for long and medium term (quarter, month, week), short term (1~24h), and capacity balancing mechanism for ultra short term (10min~1h).
For the long and medium term, the parties involved in demand-side response need to figure out the possible grid-connected power of wind power and PV power generation in the area under the jurisdiction of grid dispatch for this quarter/month/week, and plan in advance the required response power for this quarter/month/week.Similar to the previous day's market, the demand-side participants play games to determine the required response power and the related incentive and penalty rules for this quarter/month/week.
The short-term (4~24h) balancing is based on the long-term and medium-term balancing.Short-term balancing requires consideration of the day's generation plan and customer load schedule to maintain the day's power balance, so sufficient load that can be curtailed or interrupted during the longer time period (4~24h) needs to be reserved.
The ultra-short-term (10min~4h) capacity balance is based on the short-term balance.Ultra-shortterm response resources need to vary according to the short term fluctuations of new energy generation to maintain capacity balance, thus requiring users to respond to the incentive signal promptly and rapidly.

New energy consumption optimization model
The model considers 3 levels of source-grid-load in collaboration for the distribution network, where the source layer includes conventional power plants and new energy power plants, and conventional power plants consider thermal power plants, new energy power plants consider wind farms and photovoltaic power plants.The load layer includes active load and passive load.Active load considers interruptible load, transferable load, shiftable load, and curtailable load, while passive load, i.e., basic load, is the load that must be satisfied under all circumstances.

Objective function
With the objective of minimizing the total cost and maximizing the new energy consumption rate, the new energy consumption optimization model is established with the interruption load size, transferable load and levelable load strategies as the decision quantities in different time periods, namely: min =   +   +   = ( IL,1 ,  IL,2 , ⋯  IL,T ) max =  consumable  rated_output * 100% Where:  is the total cost;   is the new energy purchase cost;   is the conventional power purchase cost;   is the demand-side response compensation cost;  , ,  , ,...,  , is the interrupted load power in time period 1,2,...,T; η is the new energy consumption rate; is the actual new energy consumption power; is the rated power output of new energy.The cost of each item is calculated as follows: Where: ΔT is the adjacent time interval; EDG is the unit price of new energy purchase; EVCC is the unit price of conventional power purchase; and is the conventional energy power.

Constraint conditions and solution methods
1) New energy capacity constraints.
2) Power balance constraint.That is, the new energy output, conventional power output, total load, interrupted load, transferred load, and panning load should achieve real-time dynamic balance.Where: T TL,1 , T TL,2 are the time of continuous operation before and after load leveling, respectively.7) Base load constraint.
≥  ,  , =  , −  ,, −  ,, −  ,, Where: P fix,t is the basic load, which is the part of the demand-side load that is not involved in interruption, transfer, panning, and must work.That is, the equivalent output power must meet the important power requirements, and the extra part may participate in the demand-side response.

𝑃 𝑙𝑖𝑛𝑒 ≤ 𝑃 𝑙𝑖𝑛𝑒,𝑚𝑎𝑥
Where:P line is the actual line transmission power; P line,max is the maximum power allowed to be transmitted by the line.9) Node voltage constraint.

𝑈 𝑖,min ≤ 𝑈 𝑖,s,t ≤ 𝑈 𝑖,max
Where: U i,min , U i,max are the lower and upper limits of the node voltage, respectively.For the above dual-objective optimization problem, the decision quantities are usually limited to integers and contain a large number of 0-1 variables according to the actual requirements of the demandside response contract, which will result in: (i) a finite number of combinations of values of the decision quantities, but the number is too large to be enumerated; and (ii) usually multiple sets of better solutions, while satisfying the objective of lowest total cost.
In single-objective optimization problems, there is usually only one optimal solution, and the optimal solution can be found by relatively simple and common mathematical methods.However, in multiobjective optimization problems, the mutual constraints among the objectives may make the improvement of the performance of one objective often at the expense of the performance of other objectives, and there cannot exist a solution that makes the performance of all objectives optimal, so for multi-objective optimization problems, the solution is usually a set of non-inferior solutions -the Pareto solution set.
In the presence of multiple Pareto-optimal solutions, it is difficult to choose which solution is preferable if there is no more information about the problem, so all Pareto-optimal solutions can be considered to be equally important.It follows that for a multi-objective optimization problem, the most important task is to find as many Pareto-optimal solutions as possible about that optimization problem, and in this paper, the solution algorithm for solving the Pareto front is used, which is solved by MATLAB using a genetic algorithm based on genetic algorithm.

Example analysis
In order to verify the effectiveness of the new energy optimization and consumption model, the IEEE37 node test system is selected for simulation analysis, and the network topology and numbering are shown in Figure 4.The simulation connects a single-phase PV power generation system with rated capacity of 0.2 MW at nodes 704 and 708, respectively.Where 704 and 708 are PV access points.The network is an asymmetric distribution system with 4.8 kV voltage level, and the simulation results are expressed in the standard minimum value.The voltage reference value is 4.8kV and the power reference value is 1MVA.
In Figure 1, the nodes configured for thermal power plant and wind farm are marked with triangles, and the connected PV systems are denoted as PV1 and PV2, respectively.the specific configuration scheme of the system is shown in Table 1 and Table 2.
OpenDSS software is applied to model the IEEE37 node system, based on which, the new energy optimization consumption model algorithm proposed in this paper is implemented by writing a program in Matlab.The state estimation convergence target ε = 10 3 .Taking a 24-hour day with a 1-hour time interval, the new energy rated output curve is shown in Figure 2, and the typical daily load curve is shown in Figure 3.The fixed compensation cost of interruptible load is taken as 5000 RMB, the unit price of interruptible load compensation is taken as 0.9 RMB/kWh, the unit price of transferable load compensation is taken as 0.4 RMB/kWh, the unit price of panning load compensation is taken as 0.3 RMB/kWh, the unit price of conventional power purchase is 0.4499 RMB/kWh, and the unit price of new energy purchase is 0.3749 RMB/kWh.
The characteristics and roles of transferable load and levelable load are relatively similar, and these 2 loads are considered in aggregate as one type of resource, and interruptible load as another type of resource.According to the arithmetic system configuration, the following 4 case scenarios are designed: case1: no demand-side response; case2: interruptible load is considered; case3: transferable load and panning load are considered; case4: interruptible load, transferable load, panning load, and curtailable load are considered.The interruptible load, transferable load, shiftable load, and curtailable load are configured according to Table 2, and the simulation is solved to obtain the cost of power purchase and new energy consumption curve for a typical day at 24 hours under four scenarios, as shown in Figure 4   Table3 New energy consumption rate and power purchase cost.According to Table 3, the analysis leads to the following conclusions: 1) Comparing case1 and case2, after the introduction of interruptible load, at each time node, the total system cost is reduced, and the model can be considered to meet the economic requirements, while the new energy consumption rate is increased in both cases, indicating that the model plays the role of real-time optimal dispatching, which helps to consume new energy with stochastic power output characteristics.
2) Comparing case2 and case3, the optimization effect of case3 is more obvious at the time nodes where transferable load and levelable load are introduced, while the optimization effect of case2 is more obvious at the rest of time nodes, which indicates that interruptible load is a more flexible demand-side load resource than transferable load and levelable load, which are only suitable for allocation at load nodes with specific demand, while interruptible load is more adaptable and scalable, and can be considered to be configured in all load nodes.
3) Comparing case2 and case4, after the introduction of panning load and transferable load, it does not optimize the total cost and new energy consumption rate at each time node, but it can improve the total cost and consumption rate indexes in one cycle.In some time nodes, the system tide distribution can be made more reasonable by shifting and transferable load, which can play the role of peak-shaving and valley-filling.The reduced power purchase cost is larger than the demand-side response compensation cost, so the total cost is further reduced, while the consumption rate is further improved.4) Comparing the total cost and the consumption rate, the trend of the 2 curves is opposite, the higher the consumption rate, the lower the total cost; the lower the consumption rate, the higher the total cost.The scale of demand-side response is not large, and the impact on the total system load is limited, and the compensation cost of demand-side response is also limited.However, demand-side response can optimize the power supply structure, and improve the new energy consumption rate while significantly reducing the power purchase cost, thus meeting the economic requirements.Therefore, the economy index and the consumption rate index can be considered and optimized simultaneously.

Table 2
IEEE37 node system load side composition.