A Method for Optimizing the Operation of a Data Center Microgrid Based on Demand Response

Abstract

The present invention discloses a method for optimizing the operation of a data center microgrid based on demand-side response, belonging to the field of data center microgrid optimization, including: obtaining the day-ahead offers of the load side and the power generation side respectively; the load providers on the load side include DCMG; establishing and solving a two-stage electricity market clearing model based on CVaR; in the first stage, maximizing social welfare based on the day-ahead offers of the load side and the power generation side, and the decision variables include the power generation plans of each unit of the power generation side at each time period after clearing, and the load reduction plans of each time period on the load side after clearing; after scenario sampling in the second stage, solving the adjustment amounts of the decision variables in the first stage under each scenario, comprehensively obtaining the decision variables in the second stage after adjustment, and calculating the clearing electricity price therefrom; aiming at minimizing the operation cost of DCMG, performing day-ahead scheduling optimization on DCMG to obtain the power output plans of each device therein. The present invention can reduce the operation cost of the data center and reduce the risk brought by system uncertainty.

Description

Technical Field

[0001] The present invention belongs to the field of data center microgrid optimization, and more specifically, relates to a method for optimizing the operation of a data center microgrid based on demand response. Background Art

[0002] According to statistics, the data scale in China reached 64 ZB in 2020 and is still expanding rapidly at an annual growth rate of 50%. As an important infrastructure, the data center can provide physical support for the calculation of massive data. Along with the high-speed development of its industry, the power consumption of the industry is also growing at a high rate of more than 10% per year, which makes the data center generate high electricity costs every year.

[0003] As an industrial-level large power user, the data center itself has a strong flexible adjustment ability for energy consumption. Therefore, it can interact with the power market by participating in demand response, thereby effectively reducing its own operating costs. At present, in terms of the interaction between the data center and the power market, existing research has shown that the interaction between the data center - power market based on the demand response (DR) mechanism plays an important role in reducing the operating costs of the data center and improving the economic stability of the power grid system.

[0004] However, in the interaction between the data center and the power market, there are still some difficult problems to be solved. On the one hand, at the level of the power market clearing mechanism design, while utilizing the flexible adjustment ability of the data center's own energy consumption, the problem of load uncertainty will occur, and the existing power market clearing models do not consider the risk aversion problem under uncertain environments; on the other hand, at the level of the interaction between the data center and the power market, how the data center, as a load provider, combines its own load curtailment situation to submit an optimal bidding strategy to the power system operator, and how to schedule and optimize the operation of the data center microgrid under the market clearing framework are still problems to be solved urgently. Summary of the Invention

[0005] Aiming at the defects and improvement requirements of the existing technology, the present invention provides a method for optimizing the operation of a data center microgrid based on demand response, aiming to reduce the operating costs of the data center while effectively reducing the risks brought by system uncertainty.

[0006] To achieve the above object, according to one aspect of the present invention, a method for optimizing the operation of a data center microgrid based on demand response is provided, including:

[0007] Obtain the day-ahead bids of the load side and the generation side respectively; the day-ahead bids of the load side include the total load demands, the load curtailment electricity and the corresponding prices of each load provider at each time period, and the day-ahead bids of the generation side include the generation plans and the corresponding prices of each unit at each time period; the load providers on the load side include DCMG;

[0008] Establish a two-stage electricity market clearing model based on CVaR and solve the decision variables. In the two-stage electricity market clearing model, the optimization objective in the first stage is to maximize social welfare based on the day-ahead bids of the load side and the generation side. The decision variables include the generation plans of each unit of the generation side at each time period after clearing, and the load curtailment plans of each time period on the load side after clearing. After scenario sampling in the second stage, solve the adjustment amounts of the decision variables in the first stage under each scenario, and adjust the decision variables in the first stage to obtain the decision variables corresponding to each scenario. Multiply them by the corresponding scenario probabilities and sum them up to obtain the decision variables in the second stage. Extract the load curtailment plans of each time period of the DCMG after clearing from them, and calculate the clearing electricity price corresponding to the decision variables in the second stage.

[0009] Aim at minimizing the operating cost of the DCMG, and perform day-ahead scheduling optimization on the data center microgrid according to the clearing electricity price and the load curtailment plans of each time period of the DCMG after clearing to obtain the output plans of each device in the DCMG.

[0010] Furthermore, in the two-stage electricity market clearing model, the objective function in the first stage is:

[0011]

[0012]

[0013] where x represents the set of decision variables T represents the total number of scheduling time periods; i represents the index of the generation-side units, i = 1, 2, …, I, and I represents the number of generating units; j represents the index of the price-electricity pair bidding segments on the generation side; represents the generation plans of each unit of the generation side at each time period after clearing, respectively represent the planned output results of electricity, reserve, and frequency regulation of the j-th bidding segment of generating unit i in time period t, respectively represent the prices of electricity, reserve, and frequency regulation of the j-th bidding segment of generating unit i, respectively represent the total number of bidding segments of electricity, reserve, and frequency regulation of generating unit i; is a 0 / 1 variable. 0 means that generating unit i does not provide frequency regulation in time period t, and 1 means that generating unit i provides frequency regulation in time period t; n represents the index of the load provider, k represents the index of the load-side bidding segment, and K n represents the total number of bidding segments of load merchant n; represents the load curtailment plans of each time period on the load side after clearing, represents the non-curtailable load planned volume on the load side in time period t after clearing, P L represents the price of the non-curtailable load on the load side, Indicates the planned amount of load that can be curtailed within the k-th bidding segment of load provider n during the post-clearing period t, LP n,k Indicates the price of the load that can be curtailed within the k-th bidding segment of load provider n during the post-clearing period t.

[0014] Furthermore, the constraint conditions of the first stage include:

[0015] R1-1: The total power, reserve, and frequency regulation plans of each generating unit do not exceed its generating capacity;

[0016] R1-2: The power, reserve, and frequency regulation planned output of each unit in each period do not exceed the upper and lower limits of its bid electricity quantity;

[0017] R1-3: The power generation on the generation side is balanced with the load on the load side;

[0018] R1-4: The load on the load side does not exceed the upper and lower limits of its bid electricity quantity;

[0019] R1-5: Frequency regulation constraint, the expression is as follows:

[0020]

[0021]

[0022] R1-6: Reserve constraint, the expression is as follows:

[0023]

[0024]

[0025] Among them, Indicates a predefined lower limit value, Indicates a predefined upper limit value; λ1 represents a preset risk adjustment coefficient, and λ2 represents a preset proportionality coefficient.

[0026] Furthermore, in the two-stage electricity market clearing model, the objective function of the second stage is:

[0027]

[0028]

[0029]

[0030]

[0031] Among them, S′ represents the total number of scenarios, s represents the scenario index; ρ represents the risk aversion factor, ρ ∈ [0,1); η represents the threshold of the expected social welfare value, α represents the confidence level; ζ sdenotes the auxiliary variable in the \(s\)-th scenario, \(p\) s denotes the occurrence probability of the scenario; the set of decision variables denotes the second-stage adjustment amount of the first-stage decision variable in the \(s\)-th scenario respectively denote that in the \(s\)-th scenario, for the in the first-stage decision variable

[0032] Furthermore, in the two-stage electricity market clearing model, the constraint conditions in the second stage also include:

[0033] R2-1: After adjustment, the total electricity, reserve, and frequency regulation plans of each generating unit do not exceed its generating capacity;

[0034] R2-2: After adjustment, the electricity, reserve, and frequency regulation planned output of each unit at each time period do not exceed the upper and lower limits of its bid electricity quantity;

[0035] R2-3: After adjustment, the electricity generation on the generation side is balanced with the load on the load side;

[0036] R2-4: After adjustment, the load on the load side does not exceed the upper and lower limits of its bid electricity quantity;

[0037] R2-5: Frequency regulation constraint, the expression is as follows:

[0038]

[0039]

[0040] R2-6: Reserve constraint, the expression is as follows:

[0041]

[0042] Furthermore, before solving the two-stage electricity market clearing model, it also includes:

[0043] Linearize the frequency regulation constraints in the two stages.

[0044] Furthermore, for any load provider on the load side, in its day-ahead bid, the load curtailment electricity quantity and price at any time period \(t\) are respectively:

[0045]

[0046]

[0047] Among them, \(u\) represents the index of the traditional unit, \(u = 1, 2, \ldots, U\), and \(U\) represents the total number of traditional units managed by the load provider; is the conservative prediction value of the net load in time period \(t\), represents the output of the traditional unit \(u\) in time period \(t - 1\), represents the upper limit of the power output of the traditional unit \(u\), \(R_U\) u represents the upward ramp rate of the traditional unit \(u\), \(\beta\) u represents the fuel cost parameter of the traditional unit \(u\).

[0048] Furthermore, the objective function for the day-ahead scheduling optimization of the data center microgrid is:

[0049]

[0050]

[0051]

[0052] Among them, \(\alpha\) u represents the no-load cost of the traditional unit \(u\); represents the electric power output of the traditional unit \(u\) within time period \(t\); \(S_U\) u represents the start-up cost of the traditional unit \(u\), \(S_D\) u represents the shutdown cost of the traditional unit \(u\), \(\tau\) u，t and \(\tau\) u，t-1 are both 0 / 1 variables. Being 1 means the traditional unit \(u\) is started within the corresponding time period, and being 0 means the traditional unit \(u\) is not started within the corresponding time period; represents the clearing price of time period \(t\), \(P\) t grid represents the electricity purchase quantity within time period \(t\).

[0053] Furthermore, the constraint conditions for the day-ahead scheduling optimization of the data center microgrid include:

[0054] Power balance constraint:

[0055]

[0056] Output upper and lower limits and ramp rate constraints:

[0057]

[0058]

[0059] Electric energy storage system ESS operation constraints:

[0060]

[0061] When the data center purchases electricity from the grid by connecting to the local area grid, it is restricted by the capacity of the transmission line:

[0062]

[0063] Among them, P t disc represents the discharge power during period t; is a 0 / 1 variable, where 1 means discharging during period t and 0 means not discharging during period t; ER t represents the new energy power generation during period t; P t DC represents the power load of the data center during period t; P t char represents the charging power during period t; is a 0 / 1 variable, where 1 means charging during period t and 0 means not charging during period t; and respectively represent the lower and upper limits of the electric power output of the traditional unit u; RD u and RU u represent the downward and upward ramping powers of the traditional unit u, represents the electric power output of the traditional unit u during period t + 1; ES t and ES t+1 respectively represent the charging states of the electric energy storage system during periods t and t + 1, ES min and ES max respectively represent the minimum and maximum charging states of the electric energy storage system, η char and η disc respectively represent the charging efficiency and discharging efficiency of the electric energy storage system;; and respectively represent the maximum charging power and maximum discharging power of the electric energy storage system; represents the transmission capacity of the line.

[0064] According to another aspect of the present invention, there is provided a computer-readable storage medium, including a stored computer program; when the computer program is executed by a processor, it controls the device where the computer-readable storage medium is located to execute the above-mentioned data center microgrid operation optimization method based on demand response provided by the present invention.

[0065] Generally speaking, through the above technical solutions conceived by the present invention, the following beneficial effects can be achieved:

[0066] (1) The present invention takes the data center microgrid as the load side and establishes a two-stage electricity market clearing model based on CVaR. When establishing the model, the generation plans of each unit at each time period on the generation side and the electricity quantity of the load that can be curtailed and the corresponding price at each time period on the load side are considered simultaneously. Thus, the demand-side response can be considered, and the flexible adjustment ability of the data center microgrid's own energy consumption can be fully utilized. At the same time, in the two-stage electricity market clearing model, after the decision variables that maximize the social welfare are obtained in the first stage, in the second stage, based on the decision variables in the first stage, the adjustment amounts of the decision variables in each scenario are optimized and solved, and the decision variables in the first stage are adjusted in each scenario. Finally, the final decision variables are obtained by synthesizing the decision variables in each scenario, and the cleared time-of-use electricity price and the load curtailment plan are provided to the data center microgrid to optimize its operation. Therefore, the risk brought by uncertainty can be effectively reduced. Generally speaking, the present invention effectively reduces the risk brought by system uncertainty while reducing the operating cost of the data center.

[0067] (2) The present invention establishes a two-stage electricity market clearing model based on CVaR, and its risk aversion degree can be flexibly controlled by the risk aversion factor ρ to meet different application requirements.

[0068] (3) In the preferred embodiment of the present invention, before solving the two-stage electricity market clearing model, the frequency modulation constraints related to the 0 / 1 variables are linearized, which can improve the efficiency and accuracy of model solving.

[0069] (4) In the preferred embodiment of the present invention, in the day-ahead quotation provided by the load side, the maximum generation capacity of the traditional units in the data center microgrid at each time period is determined as the electricity quantity of the load that can be curtailed at the corresponding time period, and the linear marginal cost of the traditional units (i.e., the fuel cost parameter of the traditional units) is determined as the price of the load that can be curtailed at the corresponding time period. This quotation mechanism is consistent with the load situation in the data center microgrid, effectively ensuring that the electricity market clearing model established based on this quotation information can fully consider the demand-side response and fully utilize the flexible adjustment ability of the data center microgrid's own energy consumption. Description of the Drawings

[0070] Figure 1 Schematic diagram of the operation optimization method for the data center microgrid based on demand-side response provided by the embodiment of the present invention;

[0071] Figure 2 Predicted power generation of the wind farm and the photovoltaic power station provided by the embodiment of the present invention;

[0072] Figure 3 Cleared electricity price of the electricity market based on demand-side response provided by the embodiment of the present invention, including the cleared electricity prices of three resources: electricity, frequency modulation, and reserve;

[0073] Figure 4 The curtailment amount of the controllable load of the data center microgrid after market clearing provided by the embodiments of the present invention;

[0074] Figure 5 Comparison of social welfare and conditional value at risk (CVaR) under different risk aversion factors ρ provided by the embodiments of the present invention;

[0075] Figure 6 The scheduling results of the electric energy storage system (ESS) under different electricity price backgrounds provided by the embodiments of the present invention; among them, (a) is the ESS scheduling result under the fixed electricity price background, and (b) is the ESS scheduling result under the time-of-use electricity price background;

[0076] Figure 7 The scheduling results of each unit in the data center microgrid (DCMG) under different electricity price backgrounds provided by the embodiments of the present invention; among them, (a) is the scheduling result of each unit in the DCMG under the fixed electricity price background, and (b) is the scheduling result of each unit in the DCMG under the time-of-use electricity price background. Detailed implementation manners

[0077] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0078] In the present invention, terms such as "first" and "second" in the present invention and the accompanying drawings (if any) are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence.

[0079] In order to reduce the operating cost of the data center while effectively reducing the risks brought by system uncertainties, the present invention provides an operation optimization method for a data center microgrid based on demand response. The overall idea is as follows: regarding the data center microgrid as the load side of the power market, when establishing a power market clearing model, consider the day-ahead offer provided by the data center microgrid to the power market based on the situation of its internal controllable load, so that the data center microgrid actively participates in and even affects the power market clearing process. At the same time, establish a two-stage risk aversion power market clearing model based on CVaR, so that the data center microgrid can reduce the operating cost during operation optimization and reduce the risks brought by uncertainties.

[0080] In the present invention, the main English abbreviations involved are as follows:

[0081] DR: DemandResponse, demand-side response;

[0082] DCMG: Data Center Micro Grid, data center microgrid;

[0083] ESS: Energy Storage System, energy storage system;

[0084] CVaR: Conditional value at risk, conditional value at risk;

[0085] PSO: Power System Operator, power system operator.

[0086] The following are examples.

[0087] An operation optimization method for a data center microgrid based on demand-side response, as Figure 1 shown, includes: obtaining the day-ahead offers of the load side and the generation side respectively on the PSO side; the day-ahead offers of the load side include the total load demand of each load provider in each time period, the load curtailment electricity quantity and the corresponding price, and the day-ahead offers of the generation side include the generation plan of each unit in each time period and the corresponding price; the load providers on the load side include DCMG.

[0088] In this embodiment, the day-ahead offers provided by the load provider implement a DR offer mechanism based on load curtailment. Taking the data center microgrid as an example, the load curtailment electricity quantity and the corresponding price in each time period are determined as follows:

[0089] First, considering the output of new energy units, in this embodiment, the new energy units of the data center microgrid mainly include wind turbines and photovoltaic units; the conservative estimate of the net load of DCMG for each time period is:

[0090]

[0091] Among them, represents the total power load of the data center in time period t under scenario s; and respectively represent the output of the wind turbine and the photovoltaic unit in time period t under scenario s; scenario s represents the specific scenario obtained after scenario sampling. Optionally, in this embodiment, scenario sampling is performed by the Monte Carlo sampling method. Through scenario sampling, multiple scenarios and the probabilities of each scenario occurring can be obtained. To reduce the exponentially growing computational burden (i.e., the curse of dimensionality) caused by the increase in the number of scenarios, the scenario reduction method is used to extract the scenarios.

[0092] DCMG determines the load curtailment amount is the maximum power generation capacity of the self-owned generator set during period t, that is

[0093]

[0094] where represents the conservative estimate of the net load during period t, represents the output of the traditional unit u during period t - 1; represents the upper limit of the power output of the traditional unit u; RU u represents the upward ramp rate of the traditional unit u.

[0095] The DCMG declaration corresponds to the price is the linear marginal cost of unit u, and its formula is as follows:

[0096]

[0097] where β u represents the fuel cost parameter of the traditional unit u.

[0098] After the above calculations, the day-ahead offer provided by the data center is a piecewise cuttable load electricity - price pair, that is In this embodiment, the cuttable load amount and the corresponding price for each period on the load side are determined in the above manner, which is consistent with the load situation in the data center microgrid and reflects the demand-side response.

[0099] Considering that the plan for cuttable load on the load side will lead to load uncertainty, after obtaining the day-ahead offer information on the power generation side and the load side, this embodiment further includes: establishing a two-stage electricity market clearing model based on CVaR and solving the decision variables, and the corresponding model is as follows:

[0100]

[0101]

[0102] where F(x) is the social welfare of the first-stage optimization with respect to the decision x, that is, the difference between the cost paid by the load side for purchasing electricity and the income obtained by the power supply side; Δx represents the decision adjustment amount related to the random scenario set ξ, and F(Δx|ξ) is the expected social welfare of the decision under the scenario set ξ; ρ is the risk aversion factor within the range of [0, 1); ξ s represents the s-th scenario in the scenario set ξ, p s represents the scenario ξ s of the occurrence probability; η represents the threshold of the expected social welfare value, represents the set of real numbers; the CVaR parameter measures the lower expected social welfare value in an uncertain environment.

[0103] Specifically, in the two-stage electricity market clearing model, the optimization objective in the first stage is to maximize social welfare based on the day-ahead bids of the load side and the generation side. The decision variables include the generation plans of each generator unit at each time period after clearing, and the load shedding plans of each time period on the load side after clearing. After scenario sampling in the second stage, the adjustment amounts of the decision variables in the first stage under each scenario are solved, and the decision variables in the first stage are adjusted to obtain the decision variables corresponding to each scenario. After multiplying by the corresponding scenario probabilities and accumulating, the decision variables in the second stage are obtained.

[0104] In this embodiment, in the two-stage electricity market clearing model, the objective function in the first stage is:

[0105]

[0106]

[0107] where \(x\) represents the set of decision variables \(T\) represents the total number of scheduling time periods; \(i\) represents the index of the generator units on the generation side, \(i = 1, 2, \ldots, I\), and \(I\) represents the number of generator units; \(j\) represents the index of the price-quantity pair bidding segments on the generation side. The bidding segment represents the quantity-price segments that the generation side can bid within a time period. Each quantity-price segment includes the electricity quantity that can be provided and the corresponding price; represents the generation plan of each generator unit at each time period on the generation side after clearing, respectively represent the planned output results of electricity, reserve, and frequency regulation for the \(j\)-th bidding segment of generator unit \(i\) within time period \(t\), respectively represent the prices of electricity, reserve, and frequency regulation for the \(j\)-th bidding segment of generator unit \(i\), respectively represent the total number of bidding segments for electricity, reserve, and frequency regulation of generator unit \(i\); is a 0 / 1 variable. Being 0 means that generator unit \(i\) does not provide frequency regulation within time period \(t\), and being 1 means that generator unit \(i\) provides frequency regulation within time period \(t\); \(n\) represents the index of the load merchants on the load side, \(k\) represents the index of the bidding segments on the load side, and \(K\) n represents the total number of bidding segments of load merchant \(n\); represents the load shedding plan of each time period on the load side after clearing, represents the non-sheddable load planned quantity on the load side within time period \(t\) after clearing, \(P\) L represents the price of the non-sheddable load, represents the planned quantity of the load that can be shed within the \(k\)-th bidding segment of load merchant \(n\) corresponding to time period \(t\) after clearing, \(LP\) n,k represents the price of the load that can be shed within the \(k\)-th bidding segment of load merchant \(n\) corresponding to time period \(t\) after clearing; in the objective function of the first stage, Denotes the total price at which the PSO wholesales electricity to the load merchants, \(g\). t Denotes the total cost of the PSO purchasing electricity, reserve, and frequency regulation from the power generation side. That is, it represents the social welfare.

[0108] The constraint conditions in the first stage include:

[0109] R1 - 1: The total amounts of electricity, reserve, and frequency regulation plans of each generating unit do not exceed its generating capacity. The relevant expression is:

[0110]

[0111] Among them, \(O\) i Denotes the generating capacity of the \(i\)-th generating unit on the power generation side.

[0112] R1 - 2: The output of electricity, reserve, and frequency regulation plans of each unit in each time period does not exceed the upper and lower limits of its bid electricity quantity. The relevant expression is:

[0113]

[0114] Among them, Denote the upper limits of the electricity, reserve, and frequency regulation electricity quantities of the \(j\)-th bid segment of the generating unit \(i\) respectively.

[0115] R1 - 3: The power generation on the power generation side is balanced with the load on the load side. The relevant expression is:

[0116]

[0117] R1 - 4: The load on the load side does not exceed the upper and lower limits of its bid electricity quantity. The relevant expression is:

[0118]

[0119] Among them, Denotes the upper limit of the planned amount of load that can be curtailed within the \(k\)-th bid segment of the load merchant \(n\) corresponding to the time period \(t\), \(SNC\) t Denotes the upper limit of the non - curtailable load plan amount on the load side within the time period \(t\).

[0120] R1 - 5: Frequency regulation constraints, specifically including: (i) When each generating unit provides frequency regulation, its power generation plan amount is more than the frequency regulation plan amount, and the difference between the two is not less than a predefined lower limit value (ii) The sum of the power generation plan amount and the frequency regulation plan amount of each generating unit is less than a predefined upper limit value The expressions are as follows:

[0121]

[0122]

[0123] R1-6: Spare constraints, with the expressions as follows:

[0124]

[0125]

[0126] Among them, λ1 represents a preset risk adjustment coefficient, and λ2 represents a preset proportionality coefficient; different from the power generation and frequency regulation constraints, the spare plan quantity should be sufficient to make up for the power loss caused by some generator failures; it can be further adjusted by the risk adjustment coefficient λ1 (set by PSO) to ensure a safer power source. On the contrary, the spare provided by each unit shall not exceed λ2 of its power generation output.

[0127] After the solution of the first stage is completed, the decision variables that maximize the social welfare without considering uncertainty can be obtained, including the power generation plans of each unit on the power generation side at each time period after clearing and the load shedding plans of each time period on the load side after clearing

[0128] The optimization decision in the second stage is to adjust based on the scheduling decision in the first stage after realizing the load uncertainty based on the probability scenario set, considering maximizing the expected social welfare under the scenario set ξ. For each scenario ξ s introduce an auxiliary variable ζ s , then the risk aversion market clearing optimization objective in the second stage can be transformed into the following solvable form:

[0129]

[0130]

[0131]

[0132]

[0133] Among them, S′ represents the total number of scenarios, s represents the scenario index; ρ represents the risk aversion factor, ρ ∈ [0, 1); η represents the threshold of the expected value of social welfare, α represents the confidence level; ζ s represents the auxiliary variable under the s-th scenario, p s represents the occurrence probability of the scenario; the decision variable set represents the second-stage adjustment amount of the first-stage decision variables under the s-th scenario, respectively represent the adjustment amounts of in the first-stage decision variables under the s-th scenario; The adjustment amount of the electricity wholesale price caused by the second-stage adjustment represents the deviation of the generation cost due to the decision adjustment in the second stage.

[0134] The constraint conditions in the second stage also include:

[0135] R2-1: After adjustment, the total amount of electricity, reserve, and frequency regulation plans of each generating unit does not exceed its generating capacity. The relevant expression is:

[0136]

[0137] R2-2: After adjustment, the electricity, reserve, and frequency regulation planned output of each unit in each time period does not exceed the upper and lower limits of its bid electricity quantity. The relevant expression is:

[0138]

[0139] R2-3: After adjustment, the power generation amount on the generation side is balanced with the load amount on the load side. The relevant expression is:

[0140]

[0141] R2-4: After adjustment, the load amount on the load side does not exceed the upper and lower limits of its bid electricity quantity. The relevant expression is:

[0142]

[0143] R2-5: Frequency regulation constraint, specifically including: (i) After adjustment, when each generating unit provides frequency regulation, its power generation plan amount is more than the frequency regulation plan amount, and the difference between the two is not less than the predefined lower limit value (ii) After adjustment, the sum of the power generation plan amount and the frequency regulation plan amount of each generating unit is less than the predefined upper limit value The expressions are as follows:

[0144]

[0145]

[0146] R2-6: Reserve constraint, the expression is as follows:

[0147]

[0148] Among the above constraint conditions, related to In the relevant frequency modulation constraints, there are non-linear constraints where a real variable is directly multiplied by a quadratic variable (0 / 1 variable). To improve the efficiency and accuracy of model solving, in this embodiment, before solving the model, the relevant non-linear constraints will be linearized first; optionally, in this embodiment, the BigM method is specifically used for linearization, that is, the constraints are relaxed, and a maximum auxiliary variable Inf is introduced. The frequency modulation constraints in the first stage are relaxed as follows:

[0149]

[0150]

[0151]

[0152] The frequency modulation constraints in the second stage are relaxed as follows:

[0153]

[0154]

[0155]

[0156] After adjusting the decision variables in the first stage according to the model solving results in the second stage, the obtained decision variables in the second stage can effectively reduce the economic risks brought by load uncertainty.

[0157] On the PSO side, after clearing based on the two-stage market clearing model, it further includes: extracting the load shedding plan of each time period of the cleared DCMG from the decision variables in the second stage, and calculating the clearing electricity price corresponding to the decision variables in the second stage;

[0158] With the goal of minimizing the operating cost of DCMG, according to the clearing electricity price and the load shedding plan of each time period of the cleared DCMG, the day-ahead dispatch optimization of the data center microgrid is carried out to obtain the output plan of each device in DCMG;

[0159] The objective function for the day-ahead dispatch optimization of the data center microgrid is:

[0160]

[0161]

[0162]

[0163] Among them, α u represents the no-load cost of the traditional unit u; represents the electric power output of the traditional unit u within the time period t; SU u represents the start-up cost of the traditional unit u, SDu represents the shutdown cost of the traditional unit u, τ u，t and τ u，t-1 are both 0 / 1 variables. Being 1 means the traditional unit u is turned on during the corresponding time period, and being 0 means the traditional unit u is not turned on during the corresponding time period; represents the clearing price of time period t, P t grid represents the electricity purchase quantity during time period t; i.e., represents the start-up and shutdown costs and fuel costs of the traditional generator set; i.e., represents the electricity purchase cost;

[0164] The constraint conditions for the day-ahead dispatch optimization of the data center microgrid include:

[0165] Power balance constraint:

[0166]

[0167] Output upper and lower limits and ramp rate constraints:

[0168]

[0169]

[0170] Electric energy storage system ESS operation constraints:

[0171]

[0172] When the data center purchases electricity from the grid by connecting to the local regional grid, it is subject to the capacity constraint of the transmission line:

[0173]

[0174] where, P t disc represents the discharge electricity quantity during time period t; is a 0 / 1 variable. Being 1 means discharging during time period t, and being 0 means not discharging during time period t; ER t represents the new energy power generation quantity during time period t. In this embodiment, ER t =P t wind +P t PV , P t wind and P t PV respectively represent the output of the wind turbine and the output of the photovoltaic unit during time period t; P t DC represents the power load of the data center during time period t; P t charIndicates the charging power within time period t; Is a 0 / 1 variable, where 1 indicates charging within time period t and 0 indicates no charging within time period t; And Respectively represent the lower and upper limits of the electric power output of the traditional unit u; RD u And RU u Represent the downward and upward ramping powers of the traditional unit u, Represents the electric power output of the traditional unit u in time period t + 1; ES t And ES t+1 Respectively represent the charging states of the electric energy storage system in time periods t and t + 1, ES min And ES max Respectively represent the minimum and maximum charging states of the electric energy storage system, η char And η disc Respectively represent the charging efficiency and discharging efficiency of the electric energy storage system;; And Respectively represent the maximum charging power and maximum discharging power of the electric energy storage system; Represents the line transmission capacity;

[0175] After the data center microgrid solves the above operation optimization model, the output plans of each device that minimize the operation cost of the data center microgrid can be obtained; Since the load curtailment plan and clearing price based on which the data center microgrid performs operation optimization are obtained by solving the two-stage power market clearing model based on CvaR, therefore, while reducing the operation cost of the data center microgrid, this embodiment can effectively reduce the economic risks brought by load uncertainty.

[0176] Embodiment 2:

[0177] A computer-readable storage medium includes a stored computer program; when the computer program is executed by a processor, it controls the device where the computer-readable storage medium is located to execute the method for optimizing the operation of the data center microgrid based on demand response provided in the above Embodiment 1.

[0178] The following further explains and illustrates the beneficial effects that can be achieved by the present invention in combination with a specific application example.

[0179] Suppose the microgrid of the data center consists of a generating unit, a wind farm, and a photovoltaic power station respectively, and is connected to the main grid at the same time. Table 1 lists the technical parameters of each device in the data center power system, and Table 2 lists the economic parameters of each device in the data center power system. The IT load related to the data center power load is shown in Table 3. The wind speed data and the photovoltaic power generation data are obtained from the meteorological data of Zhangjiakou area in Hebei Province on the Renewables.ninja website. Suppose the wind farm consists of 15 1.6MW wind turbines, and the installed capacity of the photovoltaic power station is 40MW. The predicted power generation of the wind farm and the photovoltaic power station is as Figure 2 shown. The participants in the power market include 10 power generation companies on the generation side, 3 load providers on the load side, and 1 data center operator. The quotation information of the power generation companies is shown in Table 4. λ1 = 1.5 and λ2 = 0.6 specified by PSO, the confidence level α is 0.9, and the initial value of the risk aversion factor ρ is 0.1. The total system load generates system load scenarios according to the load forecast published on the official website of EMA.

[0180] Table 1 List of technical parameters of each device in DCMG

[0181]

[0182] Table 2 List of economic parameters of each device in DCMG

[0183]

[0184] Table 3 Data settings of IT load

[0185]

[0186] Table 4 Bidding information of the generation side in the power market

[0187]

[0188]

[0189] Table 5 Load curtailment volume of the load provider's quotation information in the power market

[0190]

[0191] Table 6 Load curtailment price of the load provider (and DCMG) in the power market

[0192]

[0193] Based on the quotation mechanism in the above-mentioned Embodiment 1, the day-ahead quotation information provided by the load provider and the DCMG operator is shown in Table 5 and Table 6.

[0194] After the PSO obtains the day-ahead offers on the power generation side and the load side, it clears the market based on the risk-averse two-stage DR power market clearing model with CVaR. The electricity market clearing price is as follows Figure 3 shown. The clearing electricity prices of electricity, reserve, and frequency regulation, the two ancillary services, determined by the market clearing model proposed in the present invention are higher during the "shortage of supply" stage at the peak load periods from 9 to 15 hours and from 18 to 21 hours, and lower during the "oversupply" stage at the low load periods such as from 2 to 7 hours. The time-of-use prices are consistent with the chronological load characteristics.

[0195] The planned load curtailment volume of DCMG after market clearing is as follows Figure 4 shown. The columns above the abscissa in the figure represent the load curtailment volume of DCMG in the day-ahead, and the columns below the abscissa represent the planned load curtailment volume of DCMG after market clearing. It can be seen that during the time periods with lower clearing prices, such as from 1 to 7 hours and 24 hours, the planned load curtailment volume of DCMG after clearing is less than the load curtailment electricity that can be cut in the day-ahead offer, that is, DCMG tends to purchase electricity in the power market. While during the time periods with higher clearing prices, the planned load curtailment volume of DCMG is larger, that is, DCMG tries to avoid purchasing electricity from the power market, which reflects the effectiveness of the demand response mechanism.

[0196] Table 7 Comparison of objective functions under different models

[0197]

[0198] Table 7 shows the comparison of the objective functions under different models (the model not based on CVaR and the model based on CVaR). Compared with the optimization model not based on CVaR, the conditional value at risk (CVaR) in the optimization result of the two-stage model based on CVaR is $428.8 higher. That is, the social welfare in some probability scenarios with lower objective function values has been optimized, while the risk optimization has led to a slight reduction in the overall expected social welfare, that is, the expected social welfare has decreased by $8.0. Therefore, introducing CVaR into the two-stage market clearing model of the present invention can effectively suppress risks, make a more robust decision by sacrificing a small amount of expected social welfare, and improve the social welfare in the worse probability scenarios.

[0199] As follows Figure 5 shown, as the risk-averse factor ρ increases, the CVaR, that is, the social welfare in the worse probability scenarios gradually increases, and at the same time, the overall expected social welfare value on the ordinate decreases accordingly. It shows that the risk-averse market clearing model based on CVaR proposed in the present invention can effectively reduce the economic risks brought by system uncertainties by using the CVaR factor regarding social welfare, and the degree of risk aversion can be controlled by the magnitude of the risk-averse factor ρ.

[0200] After clearing the DR-based power market to obtain time-of-use prices, the day-ahead operation of DCMG is optimized and scheduled based on the forecast information. A fixed electricity price (the average of the time-of-use electricity prices) is used as a comparison case for simulation experiments. The scheduling results of ESS under the background of fixed electricity price and time-of-use electricity price are shown in (a) and (b) of Figure 6 respectively. The scheduling results of each unit in DCMG under the background of fixed electricity price and time-of-use electricity price are shown in (a) and (b) of Figure 7 respectively; as shown in Figure 6 and Figure 7 , the operating cost of DCMG under the background of fixed electricity price is 24.71 k$, while the operating cost under the background of DR power market is 24.15 k$. This shows that for DCMG, participating in the power market is beneficial to the optimization of its own operating cost.

[0201] It is easy for those skilled in the art to understand that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for optimizing the operation of a data center microgrid based on demand response, characterized in that, Including: Obtaining the day-ahead bids of the load side and the generation side respectively; The day-ahead bids of the load side include the total load demands of each load provider and the load curtailment electricity and corresponding prices for each time period, and the day-ahead bids of the generation side include the generation plans and corresponding prices for each unit for each time period; The load providers on the load side include DCMG; Establishing a two-stage electricity market clearing model based on CVaR and solving for the decision variables; In the two-stage electricity market clearing model, the optimization objective in the first stage is to maximize social welfare based on the day-ahead bids of the load side and the generation side. The decision variables include the generation plans for each unit of the generation side for each time period after clearing, and the load curtailment plans for each time period of the load side after clearing. After scenario sampling in the second stage, the adjustment amounts of the decision variables in the first stage for each scenario are solved, and the decision variables in the first stage are adjusted to obtain the decision variables corresponding to each scenario. After multiplying by the corresponding scenario probabilities and accumulating, the decision variables in the second stage are obtained. The load curtailment plans for each time period of DCMG after clearing are extracted therefrom, and the clearing electricity price corresponding to the decision variables in the second stage is calculated; Aiming at minimizing the operating cost of DCMG, performing day-ahead scheduling optimization on the data center microgrid according to the clearing electricity price and the load curtailment plans for each time period of DCMG after clearing to obtain the output plans of each device in DCMG; In the two-stage electricity market clearing model, the objective function in the first stage is: Among them, represents the set of decision variables ; T represents the total number of scheduling periods; i represents the index of the generating units on the generation side, i = 1, 2, …, I , I represents the number of generating units; j represents the index of the price - quantity pairs of bids segmented by the generation side; represents the generation plan of each generating unit in each period after clearing on the generation side, , , respectively represent the planned output results of electricity, reserve, and frequency regulation in the t period for the i th j segment of the bid segmentation of the generating unit, , , respectively represent the prices of electricity, reserve, and frequency regulation for the i th j segment of the bid segmentation of the generating unit, , , respectively represent the total number of bid segments of electricity, reserve, and frequency regulation for the i generating unit; is a 0 / 1 variable. 0 means that the generating unit t does not provide frequency regulation in the i period, and 1 means that the generating unit t provides frequency regulation in the i period; n represents the index of the load provider, k represents the index of the bid segments on the load side, K n represents the total number of bid segments of the load provider n ; represents the load shedding plan of each period on the load side after clearing, represents the non - shedable load planned volume in the t period on the load side after clearing, represents the price of the non - shedable load on the load side, represents the shedable load planned volume in the t period for the n th k bid segment of the load provider on the load side after clearing, represents the shedable load price in the t period for the n th k bid segment of the load provider on the load side after clearing.

2. The data center microgrid operation optimization method based on demand-side response according to claim 1, wherein In the two-stage electricity market clearing model, the constraint conditions in the first stage include: R1-1: The total amount of electricity, reserve, and frequency regulation plans of each generating unit does not exceed its generating capacity; R1-2: The electricity, reserve, and frequency regulation plan outputs of each unit for each time period do not exceed the upper and lower limits of its bid electricity; R1-3: The power generation on the generation side is balanced with the load on the load side; R1-4: The load on the load side does not exceed the upper and lower limits of its bid electricity; R1-5: Frequency regulation constraint, the expression is as follows: R1-6: Reserve constraint, the expression is as follows: Among them, represents a predefined lower limit value, represents a predefined upper limit value; represents a preset risk adjustment coefficient, represents a preset proportionality coefficient.

3. The data center microgrid operation optimization method based on demand side response according to claim 2, characterized in that, In the two-stage electricity market clearing model, the objective function in the second stage is: Among them, represents the total number of scenarios, s represents the scenario index; represents the risk aversion factor, ; represents the threshold of the expected value of social welfare, represents the confidence level; represents the s th auxiliary variable in the p s scenario; the set of decision variables represents the second-stage adjustment amount of the first-stage decision variable in the s th scenario, , , , , respectively represent the adjustment amounts of s in the first-stage decision variable in the , , , , in the th scenario.

4. The data center microgrid operation optimization method based on demand response according to claim 3, wherein In the two-stage electricity market clearing model, the constraint conditions in the second stage further include: R2-1: After adjustment, the total amount of electricity, reserve, and frequency regulation plans of each generating unit does not exceed its generating capacity; R2-2: After adjustment, the electricity, reserve, and frequency regulation plan outputs of each unit for each time period do not exceed the upper and lower limits of its bid electricity; R2-3: After adjustment, the power generation on the generation side is balanced with the load on the load side; R2-4: After adjustment, the load on the load side does not exceed the upper and lower limits of its bid electricity; R2-5: Frequency regulation constraint, the expression is as follows: R2-6: Reserve constraint, the expression is as follows: 。 5. The data center microgrid operation optimization method based on demand-side response according to claim 4, characterized in that Before solving the two-stage electricity market clearing model, it further includes: Linearizing the frequency regulation constraints in the two stages.

6. The method for optimizing the operation of a data center microgrid based on demand-side response according to any one of claims 1 to 5, characterized in that For any load provider on the load side, in its day-ahead offer, the curtailable load power t and the price for any time period are respectively as follows: wherein, u represents the index of the traditional unit, , U represents the total number of traditional units managed by the load provider; is the conservative predicted value of the net load for time period t , represents the output of the traditional unit u in time period t -1, represents the upper limit of the power output of the traditional unit u , RU u represents the upward ramp rate of the traditional unit u , represents the fuel cost parameter of the traditional unit u .

7. The method for optimizing the operation of a data center microgrid based on demand-side response according to claim 6, characterized in that The objective function for the day-ahead scheduling optimization of the data center microgrid is: Among them, represents the no-load cost of the traditional unit u ; represents the electric power output of the traditional unit u during the time period t ; SU u represents the start-up cost of the traditional unit u , SD u represents the shutdown cost of the traditional unit u ; and are both 0 / 1 variables. Being 1 means the traditional unit u is started during the corresponding time period, and being 0 means the traditional unit u is not started during the corresponding time period; represents the clearing price of the time period t ; represents the electricity purchase quantity during the time period t .

8. The method for optimizing the operation of a data center microgrid based on demand-side response according to claim 7, wherein The constraint conditions for the day-ahead scheduling optimization of the data center microgrid include: Power balance constraint: Output upper and lower limits and ramp rate constraint: Operating constraint of the electrical energy storage system ESS: When the data center purchases electricity from the power grid by accessing the local regional power grid, it is restricted by the capacity of the transmission line: Among them, represents the discharge power within the time period t ; is a 0 / 1 variable. Being 1 means discharging within the time period t , and being 0 means not discharging within the time period t ; ER t represents the new energy power generation within the time period t ; represents the power load of the data center within the time period t ; represents the charging power within the time period t ; is a 0 / 1 variable. Being 1 means charging within the time period t , and being 0 means not charging within the time period t ; and respectively represent the lower limit and upper limit of the electric power output of the traditional unit u ; and represent the downward ramp power and upward ramp power of the traditional unit u , represents the electric power output of the traditional unit u within the time period t +1; and respectively represent the charging states of the electrical energy storage system within the time periods t and t +1, and respectively represent the minimum and maximum charging states of the electrical energy storage system, and respectively represent the charging efficiency and discharging efficiency of the electrical energy storage system; and respectively represent the maximum charging power and maximum discharging power of the electrical energy storage system; represents the line transmission capacity.

9. A computer-readable storage medium, characterized in that, including a stored computer program; when the computer program is executed by a processor, it controls the device where the computer-readable storage medium is located to execute the method for optimizing the operation of the data center microgrid based on demand response according to any one of claims 1 to 8.

Images