技术领域Technical field
本发明涉及综合能源系统优化调度领域，特别是一种考虑需求响应与碳价不确定性的综合能源系统优化调度方法。The invention relates to the field of integrated energy system optimization and dispatching, in particular to an integrated energy system optimization and dispatching method that considers demand response and carbon price uncertainty.
背景技术Background technique
需求响应作为可调节需求侧资源对降低负荷波动，缓解系统供能压力具有积极作用，国内外学者已进行了大量研究。相关研究表明，冷热电联产系统供需侧的负荷匹配与动态平衡对系统节能率和利用率有一定的影响。因此考虑需求响应提高系统供需侧的负荷匹配与动态平衡进一步促进IES节能减排。As an adjustable demand-side resource, demand response plays a positive role in reducing load fluctuations and alleviating system energy supply pressure. Domestic and foreign scholars have conducted a large number of studies. Relevant research shows that the load matching and dynamic balance on the supply and demand side of the combined cooling, heating and power system have a certain impact on the system energy saving rate and utilization rate. Therefore, considering demand response to improve load matching and dynamic balance on the supply and demand side of the system further promotes IES energy conservation and emission reduction.
随着我国碳交易市场的发展，碳排放权交易价格的不确定性给系统运行调度带来了新的问题，由于碳交易政策、碳排放权供需总量以及天气环境等多重因素的相互作用，碳排放权交易价格往往在一定范围内波动，而碳价的波动会影响系统的经济性和低碳性，因此，需考虑碳价不确定性对系统调度的影响。With the development of my country's carbon trading market, the uncertainty of carbon emission rights trading prices has brought new problems to system operation and scheduling. Due to the interaction of multiple factors such as carbon trading policies, the total supply and demand of carbon emission rights, and the weather environment, Carbon emissions trading prices often fluctuate within a certain range, and fluctuations in carbon prices will affect the economics and low-carbon nature of the system. Therefore, the impact of carbon price uncertainty on system scheduling needs to be considered.
发明内容Contents of the invention
为充分调动综合能源系统需求侧响应的灵活性，并克服碳排放权交易价格的不确定性对调度计划的影响。本发明提供一种考虑需求响应与碳价不确定性的综合能源系统优化调度方法，该方法综合考虑需求响应和碳价不确定性，使系统的热电匹配度提高，进一步提高能源利用率，有效促进风电消纳和IES低碳经济运行。In order to fully mobilize the flexibility of the demand side response of the integrated energy system and overcome the impact of the uncertainty of carbon emissions trading prices on dispatch plans. The present invention provides a comprehensive energy system optimization dispatching method that considers demand response and carbon price uncertainty. This method comprehensively considers demand response and carbon price uncertainty to improve the thermoelectric matching degree of the system, further improve energy utilization, and effectively Promote wind power consumption and IES low-carbon economic operation.
本发明采取的技术方案为：The technical solutions adopted by the present invention are:
考虑需求响应与碳价不确定性的综合能源系统优化调度方法，包括以下步骤：An integrated energy system optimization dispatch method considering demand response and carbon price uncertainty includes the following steps:
步骤一：获取系统初始电、热负荷及风电预测功率，以及系统中各设备的参数及取值上、下限；Step 1: Obtain the initial electricity, heat load and wind power forecast power of the system, as well as the parameters and upper and lower limits of each equipment in the system;
步骤二：建立负荷需求响应模型，对电、热负荷需求响应目标函数进行求解，得到响应后的电、热负荷数据；Step 2: Establish a load demand response model, solve the electric and heating load demand response objective functions, and obtain the response electric and heating load data;
步骤三：基于随机场景法建立碳价不确定性模型，利用生成对抗网络生成碳价场景；Step 3: Establish a carbon price uncertainty model based on the random scenario method, and use a generative adversarial network to generate carbon price scenarios;
步骤四：通过后向削减法将步骤三生成的碳价场景削减为具有代表性的若干个典型场景；Step 4: Use the backward reduction method to reduce the carbon price scenario generated in step 3 to several representative typical scenarios;
步骤五：建立电转气-碳捕集协调运行模型，分析电-碳-气间的能量耦合关系；Step 5: Establish an electricity-to-gas-carbon capture coordinated operation model and analyze the energy coupling relationship between electricity, carbon and gas;
步骤六：以各时段系统运行成本最小为目标，建立综合能源系统优化调度模型；Step 6: Establish an integrated energy system optimization dispatch model with the goal of minimizing system operating costs in each period;
步骤七：在满足系统约束的条件下，采用YALMIP调用CPLEX 12.6进行优化求解得到系统内各设备的最优出力计划。Step 7: Under the condition that the system constraints are met, use YALMIP to call CPLEX 12.6 for optimization and solution to obtain the optimal output plan of each equipment in the system.
所述步骤二中，建立负荷需求响应模型的目标函数具体如下：In the second step, the objective function of establishing the load demand response model is as follows:
电负荷需求响应目标函数为：The electric load demand response objective function is:
式(1)中：F1为电负荷需求响应的优化目标；T为调度周期，取T＝24；PE,t、PDR,t分别为需求响应前、后t时段的电负荷；Pavg为电负荷平均值。取一个调度周期T＝24，t表示24小时中的第t个时段；In formula (1): F1 is the optimization target of the electric load demand response; T is the dispatch period, taking T = 24; PE,t and PDR,t are the electric load in the t period before and after the demand response respectively; Pavg is the average electrical load. Take a scheduling period T=24, where t represents the tth period in 24 hours;
热负荷需求响应目标函数为：The heat load demand response objective function is:
式(2)中：F2为热负荷需求响应的优化目标；Kbest为最优热电比；HDR,t表示需求响应后的热负荷；Hbest,t表示最优热电比下的热负荷。In formula (2): F2 is the optimization target of heat load demand response; Kbest is the optimal heat to power ratio; HDR,t represents the heat load after demand response; Hbest, t represents the heat load under the optimal heat to power ratio. .
所述步骤二中，建立负荷需求响应模型的约束条件具体如下：In the second step, the constraints for establishing the load demand response model are as follows:
1)电负荷需求响应弹性系数矩阵等式约束：1) Electric load demand response elastic coefficient matrix equality constraints:
通过电价弹性系数矩阵建立电价变化引起负荷变化的响应关系，电价弹性系数可定义为：The response relationship between load changes caused by changes in electricity prices is established through the electricity price elasticity coefficient matrix. The electricity price elasticity coefficient can be defined as:
式(3)中：εi,j为i时段电负荷对j时段电价的弹性系数；ΔPDR,i、ΔcDR,j分别为响应后i时段的电负荷变化量和j时段的电价变化量；PE,i、cE,j分别响应前i时段的电负荷和j时段的电价；δE为弹性电负荷的比例。In formula (3): εi,j is the elastic coefficient of the electric load in period i to the electricity price in period j; ΔPDR,i and ΔcDR,j are respectively the change amount of electric load in period i and the change amount of electricity price in period j after the response. ; PE,i and cE,j respectively respond to the electric load in the previous i period and the electricity price in j period; δE is the proportion of elastic electric load.
则需求响应后t时段的电负荷PDR,t可表示为：Then the electrical load PDR,t in period t after demand response can be expressed as:
式(4)中：ΔPDR,t表示响应后t时段的电负荷变化量；PE,t表示响应前t时段的电负荷；εt,j表示t时段电负荷对j时段电价的弹性系数；cE,t表示响应前t时段的电价；ΔcDR,t表示响应后t时段的电价变化量。In formula (4): ΔPDR,t represents the change in electric load in period t after the response; PE,t represents the electric load in period t before response; εt,j represents the elastic coefficient of the electric load in period t to the electricity price in period j. ; cE,t represents the electricity price in period t before the response; ΔcDR,t represents the change in electricity price in period t after the response.
2)热负荷与温度关系约束：2) Constraints on the relationship between heat load and temperature:
热负荷需求与室内温度满足一阶常微分方程，可表示为：The heat load demand and indoor temperature satisfy the first-order ordinary differential equation, which can be expressed as:
式(5)中：Tin,t、Tout,t分别为t时段建筑物室内温度和环境温度；Cair为室内空气热容；Hload,t为t时段热负荷需求，也即t时段系统向建筑物提供的热功率；R为建筑物的等效热阻。In formula (5): Tin,t and Tout,t are the indoor temperature and ambient temperature of the building in period t respectively; Cair is the heat capacity of indoor air; Hload,t is the heat load demand in period t, that is, period t Thermal power provided by the system to the building; R is the equivalent thermal resistance of the building.
将式(5)离散化处理后可得用户室内温度变化与供暖功率、建筑物环境温度的关系如下：After discretizing equation (5), the relationship between user indoor temperature changes, heating power, and building ambient temperature can be obtained as follows:
Tin,t+1＝Tin,te-△t/τ+(RHload,t+Tout,t)(1-e-△t/τ) (6)；Tin,t+1 =Tin,te -Δt/τ +(RHload,t +Tout,t )(1-e-Δt/τ ) (6);
式(6)中：Δt为单位调度时间，设为1h；τ＝R·Cair，对于给定的建筑物，R和Cair可记为常数。In formula (6): Δt is the unit dispatch time, which is set to 1h; τ = R·Cair . For a given building, R and Cair can be recorded as constants.
3)用户温度舒适度约束：3) User temperature comfort constraints:
为了使室内温度维持在用户舒适范围内，对Tin,t作如下约束：In order to maintain the indoor temperature within the user's comfort range, the following constraints are imposed on Tin,t :
Tmin≤Tin,t≤Tmax (7)；Tmin ≤Tin,t ≤Tmax (7);
式(7)中：Tmax、Tmin分别为用户舒适温度上、下限。In formula (7): Tmax and Tmin are the upper and lower limits of user comfort temperature respectively.
4)需求响应前后电价约束：4) Electricity price constraints before and after demand response:
电负荷需求响应后用户支付的电费应低于响应前的电费才能吸引用户积极参与需求响应：The electricity bill paid by users after the electric load demand response should be lower than the electricity bill before the response to attract users to actively participate in demand response:
式(8)中：ΔcDR,t为需求响应后t时段的电价变化量。In formula (8): ΔcDR,t is the change in electricity price in period t after demand response.
5)负荷平移约束：5) Load translation constraints:
在调度周期内响应前后负荷总量不变，同时单位时间内负荷平移量在系统允许的范围限制内：During the scheduling period, the total load before and after the response remains unchanged, and the load shift amount per unit time is within the limits allowed by the system:
式(9)中：ΔHDR,t为响应后t时段的热负荷变化量；θ1、θ2分别为单时段电、热负荷平移量限值。In formula (9): ΔHDR,t is the thermal load change amount in the t period after the response; θ1 and θ2 are the single-period electrical and thermal load translation limits respectively.
所述步骤三中，基于随机场景法建立碳价不确定性模型，具体如下：In the third step, a carbon price uncertainty model is established based on the random scenario method, as follows:
利用生成对抗网络(generative adversarial network，GAN)生成大量碳价场景。定义历史碳价数据为真实数据x，将这些历史数据之间存在的某种复杂且难以建模的分布关系设为pdata(x)，假设随机噪声数据z是从某个已知的简单分布pz(z)(如正态分布、均匀分布等)随机抽样获得。生成器输入为随机噪声数据z～pz(z)，输出为生成的数据样本G(z)，其概率分布为pG(z)；判别器接收G(z)和真实数据样本x作为输入，输出为D(x)和D(G(z))，分别表示x和G(z)在判别器中判别为真的概率。Generative adversarial network (GAN) is used to generate a large number of carbon price scenarios. Define historical carbon price data as real data x, set some complex and difficult-to-model distribution relationship between these historical data as pdata (x), and assume that random noise data z comes from a known simple distribution pz (z) (such as normal distribution, uniform distribution, etc.) is obtained by random sampling. The input of the generator is random noise data z～pz (z), and the output is the generated data sample G(z), whose probability distribution is pG (z); the discriminator receives G(z) and the real data sample x as input , the output is D(x) and D(G(z)), which respectively represent the probability that x and G(z) are judged to be true in the discriminator.
生成器和判别器的损失函数LG和LD如下：The loss functionsLG andLD of the generator and discriminator are as follows:
式中：x～pdata(x)表示x服从真实数据分布pdata(x)；logD(x)表示输出D(x)的对数。In the formula: x～pdata (x) means that x obeys the real data distribution pdata (x); logD(x) means the logarithm of the output D(x).
E表示分布的期望；LG的值越小，则D(G(z))就越大，也即生成器生成的数据越真实；LD的值越大，则D(G(z))就越小，判别器的判别能力越好。因此，G的优化目标是使得LG最小，D的优化目标是使得LD最大，对LG和LD进行组合，建立GAN的极小极大化博弈模型如下：E represents the expectation of distribution; the smaller the value of LG , the larger D(G(z)), that is, the more realistic the data generated by the generator is; the larger the value of LD , then D(G(z)) The smaller it is, the better the discriminant ability of the discriminator is. Therefore, the optimization goal of G is to minimize LG, and the optimization goal of D is to maximize LD. Combining LG and LD,theminimax game model of GAN is established as follows:
碳价不确定模型是利用生成对抗网络GAN生成大量碳价场景，进一步通过场景削减技术获取有代表性的若干个典型场景，再针对典型场景进行分析；The carbon price uncertainty model uses a generative adversarial network (GAN) to generate a large number of carbon price scenarios, and further uses scenario reduction technology to obtain several representative typical scenarios, and then analyzes the typical scenarios;
上述判别器的输入参数x即为真实碳价数据样本，当迭代达到一定次数，输出G(z)即为生成的碳价数据样本。The input parameter x of the above discriminator is the real carbon price data sample. When the iteration reaches a certain number of times, the output G(z) is the generated carbon price data sample.
通过建立GAN的极小极大化博弈模型，使得GAN的生成器和判别器交替对抗训练不断提高自身能力，最终达到纳什均衡，有效生成大量碳价场景数据。By establishing the minimax game model of GAN, the generator and discriminator of GAN can be trained alternately to continuously improve their capabilities, eventually reaching Nash equilibrium and effectively generating a large amount of carbon price scenario data.
所述步骤四中，通过后向削减法将步骤三生成的碳价场景削减为具有代表性的若干个典型场景；得到尽可能准确表达随机变量不确定性的典型场景集，具体如下：In the fourth step, the carbon price scenario generated in step three is reduced to several representative typical scenarios through the backward reduction method; a set of typical scenarios that express the uncertainty of random variables as accurately as possible is obtained, as follows:
1)初始化各个场景的概率，即：1) Initialize the probability of each scenario, that is:
式(13)中：pi为场景s的概率；H为初始场景数。In formula (13): pi is the probability of scenario s; H is the initial number of scenarios.
2)设h*为削减过程中的场景个数，对h*个场景计算任意两个场景之间的Kantorovich距离：2) Let h* be the number of scenes in the reduction process, and calculate the Kantorovich distance between any two scenes for h* scenes:
d(Xi,Xj)＝|Xi-Xj| (14)；d(Xi ,Xj )=|Xi -Xj | (14);
Xi和Xj分别表示h*个场景中的第i个和第j个场景。Xi and Xj respectively represent the i-th and j-th scenes in h* scenes.
3)对于任意场景Xi，寻找一个场景Xj使得其与Xi的距离最小，也即min{d(Xi,Xj),i≠j}，同时计算该最小场景距离与Xi的场景概率的乘积PKDi：3) For any scene Xi , find a scene Xj such that the distance betweenit and Xi is the smallest, that is, min{d(Xi , The product of scenario probabilities PKDi :
PKDi＝min{d(Xi,Xj),i≠j}×pi (15)；PKDi =min{d(Xi ,Xj ),i≠j}×pi (15);
4)在h*个场景中，找到最小的PKD，记为PKDs，并得到符合PKDi＝PKDs的场景Xi及其对应的场景Xj，可能不唯一，设其一共有hi个。4) Among the h* scenes, find the smallest PKD , recorded as PKDs , and obtain the scene Xi and its corresponding scene Xj that conform to PKDi= PKDs , which may not be unique. Let there be a total of hi indivual.
PKDs＝min{PKDi|1≤i≤h*} (16)；PKDs =min{PKDi |1≤i≤h* } (16);
式(16)中：h*表示削减过程中的场景个数；i为PKDi的下标，在1～h*内搜索最小的PKDs，个数为hi个。In formula (16): h* represents the number of scenes in the reduction process; i is the subscript of PKDi . The smallest PKDs are searched within 1 to h* , and the number is hi .
5)更新场景概率，同时将场景Xi从初始场景集中削减，从h*个场景中削减掉hi个场景，即：5) Update the scene probability, and at the same time cut the sceneXi from the initial scene set, and cut hi scenes from h* scenes, that is:
式(17)中：pj＝pj+pi表示更新场景概率，将pj+pi赋给pj；X＝X-Xi表示将场景Xi从初始场景集中削减；h*＝h*-hi表示更新削减后的场景数，从h*个场景中削减掉hi个场景。In formula (17): pj = pj + pi represents updating the scene probability, and pj + pi is assigned to pj ; X = XXi represents cutting the scene Xi from the initial scene set; h* = h* -hi indicates updating the number of scenes after reduction, and cutting hi scenes from h* scenes.
6)如果h*<H*，H*为所需的目标场景数，则完成场景削减，否则转到第2)步再次削减直到场景个数满足要求。6) If h* < H* and H* is the required target number of scenes, complete scene reduction, otherwise go to step 2) and reduce again until the number of scenes meets the requirements.
所述步骤五中，建立电转气-碳捕集协调运行模型，具体如下：In the fifth step, a power-to-gas-carbon capture coordinated operation model is established, as follows:
碳捕集设备的能耗由固定能耗和运行能耗组成，其中，固定能耗占比较小且与碳捕集设备运行状态无关，可视为恒定值，运行能耗与捕集的CO2量基本成正比。The energy consumption of carbon capture equipment consists of fixed energy consumption and operating energy consumption. The fixed energy consumption accounts for a small proportion and has nothing to do with the operating status of the carbon capture equipment. It can be regarded as a constant value. The operating energy consumption is related to the captured CO2 The quantity is basically proportional.
式(18)中：PGPPCC,t、PCO2,t、MGPPCC,CO2,t、ηc,t分别为t时段碳捕集设备的总能耗、运行能耗、碳捕集量、碳捕集率；PA为碳捕集设备的固定能耗；λ为处理单位CO2的运行能耗；eCHP,e和eCHP,h分别为CHP单位发电和发热功率的碳排放强度；eGB为GB单位发热功率的碳排放强度；PCHP,t和HCHP,t分别为t时段CHP的电、热出力；HGB,t为t时段GB的热出力。In formula (18): PGPPCC,t , PCO2,t , MGPPCC,CO2,t , and ηc,t are respectively the total energy consumption, operating energy consumption, carbon capture volume, and carbon capture equipment during t period. Capture rate; PA is the fixed energy consumption of the carbon capture equipment; λ is the operating energy consumption per unit of CO2 treated; eCHP, e and eCHP, h are the carbon emission intensity of CHP unit power generation and heating power respectively; eGB is the carbon emission intensity per unit heating power of GB; PCHP,t and HCHP,t are the electrical and thermal output of CHP in period t respectively; HGB,t is the thermal output of GB in period t.
P2G产生的CH4体积与其耗电功率的关系为：The relationship between the volume ofCH4 generated by P2G and its power consumption is:
式(19)中：VP2G,CH4,t为P2G在t时段产生的CH4体积；PP2G,t为t时段P2G的能耗；ηP2G为P2G的运行效率；HL为天然气热值；MP2G,CO2,t为t时段P2G的CO2吸收量；ρCO2为CO2密度。In formula (19): VP2G,CH4,t is the volume of CH4 produced by P2G in period t; PP2G,t is the energy consumption of P2G in period t; ηP2G is the operating efficiency of P2G; HL is the calorific value of natural gas; MP2G,CO2,t is the CO2 absorption amount of P2G during t period; ρCO2 is the CO2 density.
所述步骤六中，建立综合能源系统优化调度模型，目标函数和约束条件具体如下：In the sixth step, an integrated energy system optimization dispatch model is established. The objective function and constraints are as follows:
目标函数：Objective function:
式(20)中：NS为场景个数；P(s)为场景s发生概率；Fom,t,s为各个设备在场景s下t时段的运维成本；Ffuel,t,s、FCO2,t,s、Fseal,t,s、Floss,t,s、FEex,t,s、FHex,t,s分别为场景s下t时段的燃料成本、碳交易成本、碳封存成本、弃风惩罚成本、电网交互成本、热网交互成本。In formula (20): NS is the number of scenarios; P(s) is the occurrence probability of scenario s; Fom,t,s is the operation and maintenance cost of each device in scenario s during t period; Ffuel,t,s , FCO2,t,s , Fseal,t,s , Floss,t,s , FEex,t,s , FHex,t,s are respectively the fuel cost, carbon transaction cost, carbon Storage costs, wind curtailment penalty costs, grid interaction costs, and heat grid interaction costs.
(1)运维成本：(1) Operation and maintenance costs:
式(21)中：kom,CHP、kom,WT、kom,P2G、kom,EB、kom,GB、kom,ES、kom,HS、kom,GPPCC分别为CHP、WT、P2G、EB、GB、ES、HS、GPPCC的单位功率运维成本；PWT,t,s、PEB,t,s分别为场景s下t时段风电机组的实际出力、EB消耗的电功率；PdisES,t,s、/>HdisHS,t,s分别为场景s下t时段ES的充放电功率、HS的储放热功率。In formula (21): kom,CHP , kom,WT , kom,P2G , kom,EB , kom,GB , kom,ES , kom,HS , kom,GPPCC are CHP and WT respectively , unit power operation and maintenance costs of P2G, EB, GB, ES, HS and GPPCC; PWT,t,s and PEB,t,s are respectively the actual output of the wind turbine unit and the electric power consumed by EB in the t period under scenario s; PdisES,t,s ,/> HdisHS,t,s are respectively the charge and discharge power of ES and the heat storage and discharge power of HS in period t under scenario s.
(2)燃料成本：(2) Fuel cost:
式(22)中：kCH4为天然气单价；ηCHP,e为CHP的发电效率；ηGB为GB的转换效率。In formula (22): kCH4 is the unit price of natural gas; etaCHP, e is the power generation efficiency of CHP; etaGB is the conversion efficiency of GB.
(3)碳交易成本：(3)Carbon trading costs:
式(23)中：CCO2,s为场景s下的碳交易价格；MCO2,dis,t,s、MCO2,quota,t,s分别为场景s下t时段系统的碳排放量和碳排放配额；egrid、αgrid分别为外购单位电量火电的碳排放强度和碳排放配额；αCHP,e、αCHP,h分别为CHP单位发电和发热功率的碳排放配额；αGB为GB单位发热功率的碳排放配额；为场景s下t时段从电网购电功率；δgrid为其中火电的占比系数。In formula (23): CCO2,s is the carbon trading price under scenario s; MCO2,dis,t,s and MCO2,quota,t,s are respectively the carbon emissions and carbon emissions of the system during t period under scenario s. Emission quota; egrid and αgrid are respectively the carbon emission intensity and carbon emission quota of purchased unit thermal power; αCHP,e and αCHP,h are respectively the carbon emission quota of CHP unit power generation and heating power; αGB is GB Carbon emission quota per unit of heating power; represents the power purchased from the grid during t period under scenario s; δgrid is the proportion coefficient of thermal power.
(4)碳封存成本：(4) Carbon storage cost:
式(24)中：kseal为单位质量CO2的封存成本。In formula (24): kseal is the storage cost of CO2 per unit mass.
(5)弃风惩罚成本：(5) Wind abandonment penalty cost:
式(25)中：kloss为单位弃风功率惩罚成本；为场景s下t时段WT的预测出力。In formula (25): kloss is the unit wind power penalty cost; Contribute to the prediction of WT in period t under scenario s.
(6)电、热交互成本：(6) Electricity and heat interaction costs:
式(26)中：为t时段IES的购、售电单价；/>为t时段IES的购、售热单价；/>为场景s下t时段IES向电网售电功率；/>为场景s下t时段IES向热网购、售热功率。In formula (26): is the unit price of electricity purchased and sold by IES during period t;/> is the purchase and sale unit price of IES during period t;/> IES sells electricity power to the grid for period t under scenario s;/> For scenario s and period t, IES purchases and sells thermal power online.
约束条件：Restrictions:
(1)功率平衡约束(1) Power balance constraints
(2)P2G-GPPCC系统运行约束(2)P2G-GPPCC system operation constraints
式中：PP2G,max和PP2G,min为场景s下t时段P2G的耗电功率上下限；PGPPCC,max和PGPPCC,min为GPPCC能耗上下限；ηc,max和ηc,min为碳捕集率上下限。In the formula: PP2G,max and PP2G,min are the upper and lower limits of P2G power consumption in scenario s during t period; PGPPCC,max and PGPPCC,min are the upper and lower limits of GPPCC energy consumption; ηc,max and ηc, min is the upper and lower limits of carbon capture rate.
(3)机组运行上下限约束(3) Unit operating upper and lower limit constraints
式中：PCHP,max和PCHP,min为CHP的出力上下限；PEB,max和PEB,min为EB运行功率上下限；HGB,max和HGB,min为GB的热出力上下限。In the formula: PCHP,max and PCHP,min are the upper and lower limits of CHP output; PEB,max and PEB,min are the upper and lower limits of EB operating power; HGB,max and HGB,min are the upper thermal output limits of GB. lower limit.
(4)电、热网络功率交互约束(4) Electrical and thermal network power interaction constraints
式中：和/>分别为IES向上级电网购、售电功率上限；/>和/>分别为IES向上级热网购、售热功率上限；/>为0-1变量，分别表示场景s下t时段IES与上级电、热网交互状态。In the formula: and/> They are the upper limit of the power purchased and sold by IES from the superior power grid;/> and/> They are the upper limit of thermal power for IES to purchase and sell thermal power to superiors;/> are 0-1 variables, respectively representing the interaction status between IES and the superior electricity and heating network during t period under scenario s.
(5)储能约束(5) Energy storage constraints
式中：EES,max和EES,min为ES的储能容量上下限；和/>为ES的充放功率上限；和/>为0-1变量，表示场景s下t时段ES的充、放能状态；EES,1,s和EES,T,s为ES在场景s下初始时刻和末尾时刻的容量，考虑储能设备运行的周期性，其在始末时刻的容量相等。HS与ES约束条件相同，不再赘述。In the formula: EES,max and EES,min are the upper and lower limits of the energy storage capacity of ES; and/> is the upper limit of the charging and discharging power of ES; and/> is a 0-1 variable, indicating the charging and discharging state of ES in period t under scenario s; EES,1,s and EES,T,s are the capacity of ES at the initial moment and the end moment in scenario s, taking into account energy storage The periodicity of equipment operation, its capacity at the beginning and end moments is equal. The constraints of HS and ES are the same and will not be described again.
所述步骤七中，在满足系统约束的条件下，采用YALMIP调用CPLEX 12.6进行优化求解。其中，负荷需求响应模型决策变量为各时段的电价变化量及负荷变化量，综合能源系统优化调度模型以系统内各设备在各时段的出力为决策变量。In step seven, under the condition that system constraints are met, YALMIP is used to call CPLEX 12.6 for optimization and solution. Among them, the decision variables of the load demand response model are the changes in electricity prices and load changes in each period, and the integrated energy system optimization dispatch model uses the output of each equipment in the system in each period as the decision variables.
本发明一种考虑需求响应与碳价不确定性的综合能源系统优化调度方法，技术效果如下：This invention is an integrated energy system optimization dispatching method that considers demand response and carbon price uncertainty. The technical effects are as follows:
1)本发明负荷需求响应模型实现了对负荷的削峰填谷，同时使系统的热电匹配度提高，可有效促进风电消纳、提高能源利用率，实现IES与用户间的双赢。1) The load demand response model of the present invention realizes load peak shaving and valley filling, and at the same time improves the thermoelectric matching degree of the system, which can effectively promote wind power consumption, improve energy utilization, and achieve a win-win situation between IES and users.
2)本发明采用GAN生成碳价场景，随着对抗训练迭代次数的增加，GAN的生成器可以有效模拟真实碳价数据。2) The present invention uses GAN to generate carbon price scenarios. As the number of iterations of adversarial training increases, the GAN generator can effectively simulate real carbon price data.
3)本发明对碳价的不确定性进行建模，将不确定性优化问题转化为多个确定性场景问题；考虑碳价不确定性使得系统总成本和净碳排放量进一步降低，经济性和低碳性提高。3) This invention models the uncertainty of carbon prices and transforms the uncertainty optimization problem into multiple deterministic scenario problems; considering the uncertainty of carbon prices, the total system cost and net carbon emissions are further reduced, and the economy is improved. and improved low-carbon properties.
附图说明Description of the drawings
图1为本发明所述的GAN基本结构图。Figure 1 is a basic structural diagram of the GAN according to the present invention.
图2为本发明的IES结构图。Figure 2 is a structural diagram of the IES of the present invention.
图3为本发明的模型求解流程图。Figure 3 is a flow chart of the model solution of the present invention.
图4为本发明的判别器损失函数值变化曲线图。Figure 4 is a graph of changes in the discriminator loss function value of the present invention.
图5为本发明的1000个碳排放权交易价格样本图。Figure 5 is a sample diagram of 1,000 carbon emission rights trading prices according to the present invention.
图6为本发明的综合需求响应前后电、热负荷曲线图。Figure 6 is a graph of electricity and heat loads before and after the comprehensive demand response of the present invention.
图7为本发明的供电调度优化结果图。Figure 7 is a diagram showing the power supply dispatch optimization results of the present invention.
图8为本发明的供热调度优化结果图。Figure 8 is a diagram of the heating dispatch optimization results of the present invention.
具体实施方式Detailed ways
一种考虑需求响应与碳价不确定性的综合能源系统优化调度方法，首先为缓解综合能源系统供能压力，建立需求响应模型以减少负荷峰谷差并提高供需侧的热电匹配度；其次基于随机场景法建立碳价不确定性模型，利用生成对抗网络生成碳价场景，通过后向削减法将生成的碳价场景削减为具有代表性的若干个典型场景；然后建立电转气-碳捕集协调运行模型，分析电-碳-气间的能量耦合关系，在此基础上，以各时段系统运行成本最小为目标，建立综合能源系统优化调度模型；最后在满足系统约束的条件下，采用YALMIP调用CPLEX进行优化求解得到系统内各设备的最优出力计划。An integrated energy system optimization dispatching method that considers demand response and carbon price uncertainty. First, in order to alleviate the energy supply pressure of the integrated energy system, a demand response model is established to reduce load peak and valley differences and improve the heat and power matching on the supply and demand side; secondly, based on The random scenario method establishes a carbon price uncertainty model, uses a generative adversarial network to generate carbon price scenarios, and uses the backward reduction method to reduce the generated carbon price scenarios into several representative typical scenarios; then establish a power-to-gas-carbon capture The coordinated operation model analyzes the energy coupling relationship between electricity, carbon, and gas. On this basis, with the goal of minimizing system operation costs in each period, an integrated energy system optimization dispatch model is established. Finally, YALMIP is adopted under the condition that system constraints are met. Call CPLEX for optimization and solution to obtain the optimal output plan of each equipment in the system.
具体包括以下步骤：Specifically, it includes the following steps:
步骤1:输入初始电热负荷及其他负荷需求响应模型参数；Step 1: Enter the initial electric heating load and other load demand response model parameters;
24h初始电热负荷数据。其他负荷需求响应模型参数包括典型日室外环境温度，系统中弹性负荷的比例，电价弹性系数，建筑物等效热阻，室内空气热容，用户舒适温度，最优热电比。24h initial electrothermal load data. Other load demand response model parameters include typical daily outdoor ambient temperature, the proportion of elastic loads in the system, electricity price elasticity coefficient, building equivalent thermal resistance, indoor air heat capacity, user comfort temperature, and optimal heat to power ratio.
步骤2:以负荷波动最小为目标平移电负荷，在供需两侧的热电均衡点处系统节能性最好，热电均衡点下的热电比即为系统的最优热电比，因此热负荷需求响应目标为响应后的热负荷HDR,t与最优热电比下的热负荷Hbest,t的偏差最小，负荷需求响应模型的目标函数具体如下：Step 2: Shift the electric load with the goal of minimizing load fluctuations. The system has the best energy saving at the thermal and electric balance point on both sides of supply and demand. The thermal and electric ratio at the thermal and electric balance point is the optimal thermal and electric ratio of the system. Therefore, the thermal load demand response target In order to minimize the deviation between the heat load HDR,t after the response and the heat load Hbest,t under the optimal heat/power ratio, the objective function of the load demand response model is as follows:
式中：F1为电负荷需求响应的优化目标；T为调度周期，取T＝24；PE,t、PDR,t分别为需求响应前、后t时段的电负荷；Pavg为电负荷平均值。In the formula: F1 is the optimization target of the electric load demand response; T is the dispatch period, taking T = 24; PE,t and PDR,t are the electric load in the t period before and after the demand response respectively; Pavg is the electric load. load average.
式中：F2为热负荷需求响应的优化目标；Kbest为最优热电比。In the formula: F2 is the optimization target of heat load demand response; Kbest is the optimal heat to electricity ratio.
建立负荷需求响应模型的约束条件，包括：电负荷需求响应弹性系数矩阵等式约束，热负荷与温度关系约束，用户温度舒适度约束，需求响应前后电价约束，负荷平移约束，具体如下：The constraints for establishing the load demand response model include: electric load demand response elastic coefficient matrix equality constraints, thermal load and temperature relationship constraints, user temperature comfort constraints, electricity price constraints before and after demand response, and load translation constraints, as follows:
1)电负荷需求响应弹性系数矩阵等式约束1) Electric load demand response elasticity coefficient matrix equality constraint
通过电价弹性系数矩阵建立电价变化引起负荷变化的响应关系，电价弹性系数可定义为：The response relationship between load changes caused by changes in electricity prices is established through the electricity price elasticity coefficient matrix. The electricity price elasticity coefficient can be defined as:
式中：εi,j为i时段电负荷对j时段电价的弹性系数；ΔPDR,i和ΔcDR,j为响应后i时段的电负荷变化量和j时段的电价变化量；PE,i和cE,j为响应前i时段的电负荷和j时段的电价；δE为弹性电负荷的比例。In the formula: εi,j is the elastic coefficient of the electric load in period i to the electricity price in period j; ΔPDR,i and ΔcDR,j are the change in electric load in period i and the change in electricity price in period j after the response; PE, i and cE,j are the electric load in period i and the electricity price in period j before the response; δE is the proportion of elastic electric load.
则需求响应后t时段的电负荷PDR,t可表示为：Then the electrical load PDR,t in period t after demand response can be expressed as:
2)热负荷与温度关系约束2) Constraints on the relationship between heat load and temperature
在保证用户舒适的范围内，通过调节室内供暖温度来实现热负荷的转移，为热负荷需求响应提供了前提。热负荷需求与室内温度满足一阶常微分方程，可表示为：Within the scope of ensuring user comfort, the heat load is transferred by adjusting the indoor heating temperature, which provides a prerequisite for heat load demand response. The heat load demand and indoor temperature satisfy the first-order ordinary differential equation, which can be expressed as:
式中：Tin,t、Tout,t分别为t时段建筑物室内温度和环境温度；Cair为室内空气热容；Hload,t为t时段热负荷需求，也即t时段系统向建筑物提供的热功率；R为建筑物的等效热阻。将式(5)离散化处理后可得用户室内温度变化与供暖功率、建筑物环境温度的关系如下：In the formula: Tin,t and Tout,t are the indoor temperature and ambient temperature of the building in period t respectively;C airis the heat capacity of indoor air; The thermal power provided by the object; R is the equivalent thermal resistance of the building. After discretizing equation (5), the relationship between user indoor temperature changes, heating power, and building ambient temperature can be obtained as follows:
Tin,t+1＝Tin,te-△t/τ+(RHload,t+Tout,t)(1-e-△t/τ) (6)Tin,t+1 =Tin,t e-△t/τ +(RHload,t +Tout,t )(1-e-△t/τ ) (6)
式中：Δt为单位调度时间，设为1h；τ＝R·Cair，对于给定的建筑物，R和Cair可记为常数。In the formula: Δt is the unit dispatch time, set to 1h; τ = R·Cair . For a given building, R and Cair can be recorded as constants.
3)用户温度舒适度约束3) User temperature comfort constraints
此外，为了使室内温度维持在用户舒适范围内，对Tin,t作如下约束：In addition, in order to maintain the indoor temperature within the user's comfort range, the following constraints are imposed on Tin,t :
Tmin≤Tin,t≤Tmax (7)Tmin ≤Tin,t ≤Tmax (7)
式中：Tmax和Tmin为用户舒适温度上下限。In the formula: Tmax and Tmin are the upper and lower limits of user comfort temperature.
4)需求响应前后电价约束4) Electricity price constraints before and after demand response
电负荷需求响应后用户支付的电费应低于响应前的电费才能吸引用户积极参与需求响应：The electricity bill paid by users after the electric load demand response should be lower than the electricity bill before the response to attract users to actively participate in demand response:
式中：ΔcDR,t为需求响应后t时段的电价变化量。In the formula: ΔcDR,t is the change in electricity price in period t after demand response.
5)负荷平移约束5) Load translation constraints
在调度周期内响应前后负荷总量不变，同时单位时间内负荷平移量在系统允许的范围限制内：During the scheduling period, the total load before and after the response remains unchanged, and the load shift amount per unit time is within the limits allowed by the system:
式中：ΔHDR,t为响应后t时段的热负荷变化量；θ1、θ2分别为单时段电、热负荷平移量限值。In the formula: ΔHDR,t is the thermal load change in period t after the response; θ1 and θ2 are the single-period electrical and thermal load translation limits respectively.
步骤3:将响应后的电、热负荷数据传递至第二阶段综合能源系统优化调度模型；Step 3: Transfer the responded electricity and heating load data to the second-stage integrated energy system optimization dispatch model;
步骤4:输入风电预测功率数据，以及系统中各设备的参数及取值上、下限。Step 4: Enter the wind power forecast power data, as well as the parameters and upper and lower limits of each equipment in the system.
步骤5:基于随机场景法建立碳价不确定性模型，利用生成对抗网络生成碳价场景，GAN基本结构见图1。Step 5: Establish a carbon price uncertainty model based on the random scenario method, and use a generative adversarial network to generate carbon price scenarios. The basic structure of GAN is shown in Figure 1.
GAN主要由生成器(generator，G)和判别器(discriminator，D)两部分组成。G旨在学习真实数据潜在的概率分布，生成近似于真实的样本；D旨在尽可能准确的判别输入数据来源于G生成的“假”数据还是真实数据，输出判别结果并反向传递给自身和生成器，二者通过交替对抗训练不断提高自身能力，最终达到纳什均衡：G能产生准确反映真实数据概率分布的样本，D无法判别样本来源。GAN mainly consists of two parts: generator (generator, G) and discriminator (discriminator, D). G aims to learn the underlying probability distribution of real data and generate samples that are close to real; D aims to distinguish as accurately as possible whether the input data comes from the "fake" data generated by G or real data, output the judgment results and pass them back to itself and generator, both of which continuously improve their capabilities through alternating confrontation training, and finally reach Nash equilibrium: G can generate samples that accurately reflect the probability distribution of real data, and D cannot identify the source of the samples.
定义历史碳价数据为真实数据x，将这些历史数据之间存在的某种复杂且难以建模的分布关系设为pdata(x)，假设随机噪声数据z是从某个已知的简单分布pz(z)(如正态分布、均匀分布等)随机抽样获得。生成器输入为随机噪声数据z～pz(z)，输出为生成的数据样本G(z)，其概率分布为pG(z)；判别器接收G(z)和真实数据样本x作为输入，输出为D(x)和D(G(z))，分别表示x和G(z)在判别器中判别为真的概率。Define historical carbon price data as real data x, set some complex and difficult-to-model distribution relationship between these historical data as pdata (x), and assume that random noise data z comes from a known simple distribution pz (z) (such as normal distribution, uniform distribution, etc.) is obtained by random sampling. The input of the generator is random noise data z～pz (z), and the output is the generated data sample G(z), whose probability distribution is pG (z); the discriminator receives G(z) and the real data sample x as input , the output is D(x) and D(G(z)), which respectively represent the probability that x and G(z) are judged to be true in the discriminator.
生成器和判别器的损失函数LG和LD如下：The loss functionsLG andLD of the generator and discriminator are as follows:
式中：E表示分布的期望。LG的值越小，则D(G(z))就越大，也即生成器生成的数据越真实；LD的值越大，则D(G(z))就越小，判别器的判别能力越好。因此，G的优化目标是使得LG最小，D的优化目标是使得LD最大，对LG和LD进行组合，建立GAN的极小极大化博弈模型如下：In the formula: E represents the expectation of distribution. The smaller the value of LG , the larger D(G(z)), that is, the more realistic the data generated by the generator is; the larger the value of LD , the smaller D(G(z)), and the discriminator The better the discriminating ability. Therefore, the optimization goal of G is to minimize LG, and the optimization goal of D is to maximize LD. Combining LG and LD,theminimax game model of GAN is established as follows:
生成器和判别器的目标是相矛盾的，即先从判别器角度固定生成器使得V(G,D)最大化，再从生成器角度固定判别器使V(G,D)最小化，从而实现生成器与判别器在训练过程中的相互对抗。The goals of the generator and the discriminator are contradictory, that is, first fix the generator from the perspective of the discriminator to maximize V(G,D), and then fix the discriminator from the perspective of the generator to minimize V(G,D), so that Realize the mutual confrontation between the generator and the discriminator during the training process.
步骤6:通过后向削减法将步骤三生成的碳价场景削减为具有代表性的若干个典型场景，后向削减法详细步骤如下：Step 6: Use the backward reduction method to reduce the carbon price scenario generated in step 3 to several representative typical scenarios. The detailed steps of the backward reduction method are as follows:
1)初始化各个场景的概率，即：1) Initialize the probability of each scenario, that is:
式中：pi为场景s的概率；H为初始场景数。In the formula: pi is the probability of scenario s; H is the initial number of scenarios.
2)设h*为削减过程中的场景个数，对h*个场景计算任意两个场景之间的Kantorovich距离：2) Let h* be the number of scenes in the reduction process, and calculate the Kantorovich distance between any two scenes for h* scenes:
d(Xi,Xj)＝|Xi-Xj| (14)d(Xi ,Xj )=|Xi -Xj | (14)
3)对于任意场景Xi，寻找一个场景Xj使得其与Xi的距离最小，也即min{d(Xi,Xj),i≠j}，同时计算该最小场景距离与Xi的场景概率的乘积PKDi：3) For any scene Xi , find a scene Xj such that the distance betweenit and Xi is the smallest, that is, min{d(Xi , The product of scenario probabilities PKDi :
PKDi＝min{d(Xi,Xj),i≠j}×pi (15)PKDi =min{d(Xi ,Xj ),i≠j}×pi (15)
4)在h*个场景中，找到最小的PKD，记为PKDs，并得到符合PKDi＝PKDs的场景Xi及其对应的场景Xj，可能不唯一，设其一共有hi个。4) Among the h* scenes, find the smallest PKD , recorded as PKDs , and obtain the scene Xi and its corresponding scene Xj that conform to PKDi =PKDs , which may not be unique. Assume that they havea total of hi indivual.
PKDs＝min{PKDi|1≤i≤h*} (16)PKDs ＝min{PKDi |1≤i≤h* } (16)
5)更新场景概率，同时将场景Xi从初始场景集中削减，从h*个场景中削减掉hi个场景，即5) Update the scene probability, and at the same time cut the sceneXi from the initial scene set, and cut out hi scenes from h* scenes, that is,
6)如果h*<H*(H*为所需的目标场景数)，则完成场景削减，否则转到第二步再次削减直到场景个数满足要求。6) If h* < H* (H* is the required target number of scenes), complete scene reduction, otherwise go to the second step and reduce again until the number of scenes meets the requirements.
步骤7:建立电转气-碳捕集协调运行模型，分析电-碳-气间的能量耦合关系。电转气-碳捕集协调运行模型具体如下：Step 7: Establish an electricity-to-gas-carbon capture coordinated operation model and analyze the energy coupling relationship between electricity, carbon and gas. The details of the power-to-gas-carbon capture coordinated operation model are as follows:
碳捕集设备的能耗由固定能耗和运行能耗组成，其中固定能耗占比较小且与碳捕集设备运行状态无关，可视为恒定值，运行能耗与捕集的CO2量基本成正比。The energy consumption of carbon capture equipment is composed of fixed energy consumption and operating energy consumption. The fixed energy consumption accounts for a small proportion and has nothing to do with the operating status of the carbon capture equipment. It can be regarded as a constant value. The operating energy consumption is related to the amount ofCO2 captured. Basically proportional.
式中：PGPPCC,t、PCO2,t、MGPPCC,CO2,t、ηc,t分别为t时段碳捕集设备的总能耗、运行能耗、碳捕集量、碳捕集率；PA为碳捕集设备的固定能耗；λ为处理单位CO2的运行能耗；eCHP,e和eCHP,h分别为CHP单位发电和发热功率的碳排放强度；eGB为GB单位发热功率的碳排放强度；PCHP,t和HCHP,t分别为t时段CHP的电、热出力；HGB,t为t时段GB的热出力。P2G产生的CH4体积与其耗电功率的关系为：In the formula: PGPPCC,t , PCO2,t , MGPPCC,CO2,t , and ηc,t are respectively the total energy consumption, operating energy consumption, carbon capture volume, and carbon capture rate of the carbon capture equipment during t period. ; PA is the fixed energy consumption of the carbon capture equipment; λ is the operating energy consumption per unit CO2 processed; eCHP, e and eCHP, h are the carbon emission intensity of CHP unit power generation and heating power respectively; eGB is GB Carbon emission intensity per unit heating power; PCHP,t and HCHP,t are the electricity and heat output of CHP in period t respectively; HGB,t is the heat output of GB in period t. The relationship between the volume ofCH4 generated by P2G and its power consumption is:
式中：VP2G,CH4,t为P2G在t时段产生的CH4体积；PP2G,t为t时段P2G的能耗；ηP2G为P2G的运行效率；HL为天然气热值；MP2G,CO2,t为t时段P2G的CO2吸收量；ρCO2为CO2密度。In the formula: VP2G,CH4,t is the volume of CH4 produced by P2G in period t; PP2G,t is the energy consumption of P2G in period t; ηP2G is the operating efficiency of P2G; HL is the calorific value of natural gas; MP2G, CO2,t is the CO2 absorption amount of P2G during t period; ρCO2 is the CO2 density.
步骤8:以各时段系统运行成本最小为目标，建立综合能源系统优化调度模型，目标函数和约束条件具体如下：Step 8: With the goal of minimizing system operating costs in each period, establish an integrated energy system optimization dispatch model. The objective function and constraints are as follows:
目标函数：Objective function:
式中：NS为场景个数；P(s)为场景s发生概率；Fom,t,s为各个设备在场景s下t时段的运维成本；Ffuel,t,s、FCO2,t,s、Fseal,t,s、Floss,t,s、FEex,t,s、FHex,t,s分别为场景s下t时段的燃料成本、碳交易成本、碳封存成本、弃风惩罚成本、电网交互成本、热网交互成本。In the formula: NS is the number of scenarios; P(s) is the probability of occurrence of scenario s; Fom,t,s is the operation and maintenance cost of each equipment in scenario s during t period; Ffuel,t,s , FCO2, t,s , Fseal,t,s , Floss,t,s , FEex,t,s , and FHex,t,s are respectively the fuel cost, carbon transaction cost, and carbon storage cost in the t period under scenario s. Penalty costs for wind curtailment, grid interaction costs, and heat grid interaction costs.
(1)运维成本(1)Operation and maintenance costs
式中：kom,CHP、kom,WT、kom,P2G、kom,EB、kom,GB、kom,ES、kom,HS、kom,GPPCC分别为CHP、WT、P2G、EB、GB、ES、HS、GPPCC的单位功率运维成本；PWT,t,s、PEB,t,s分别为场景s下t时段风电机组的实际出力、EB消耗的电功率；PdisES,t,s、/>HdisHS,t,s分别为场景s下t时段ES的充放电功率、HS的储放热功率。In the formula: kom,CHP , kom,WT , kom,P2G , kom,EB , kom,GB , kom,ES , kom,HS , kom,GPPCC are CHP, WT, P2G, The unit power operation and maintenance costs of EB, GB, ES, HS and GPPCC; PWT,t,s and PEB,t,s are respectively the actual output of the wind turbine unit and the electric power consumed by EB in the t period under scenario s; PdisES,t,s ,/> HdisHS,t,s are respectively the charge and discharge power of ES and the heat storage and discharge power of HS in period t under scenario s.
(2)燃料成本(2) Fuel cost
式中：kCH4为天然气单价；ηCHP,e为CHP的发电效率；ηGB为GB的转换效率。In the formula: kCH4 is the unit price of natural gas; etaCHP, e is the power generation efficiency of CHP; etaGB is the conversion efficiency of GB.
(3)碳交易成本(3)Carbon trading costs
式中：CCO2,s为场景s下的碳交易价格；MCO2,dis,t,s、MCO2,quota,t,s分别为场景s下t时段系统的碳排放量和碳排放配额；egrid、αgrid分别为外购单位电量火电的碳排放强度和碳排放配额；αCHP,e、αCHP,h分别为CHP单位发电和发热功率的碳排放配额；αGB为GB单位发热功率的碳排放配额；为场景s下t时段从电网购电功率；δgrid为其中火电的占比系数。In the formula: CCO2,s is the carbon trading price under scenario s; MCO2,dis,t,s and MCO2,quota,t,s are the carbon emissions and carbon emission quota of the system in period t under scenario s respectively; egrid and αgrid are respectively the carbon emission intensity and carbon emission quota of purchased unit thermal power; αCHP,e and αCHP,h are respectively the carbon emission quota of CHP unit power generation and heating power; αGB is the unit heating power of GB carbon emission quota; represents the power purchased from the grid during t period under scenario s; δgrid is the proportion coefficient of thermal power.
(4)碳封存成本(4)Carbon storage cost
式中：kseal为单位质量CO2的封存成本。In the formula: kseal is the storage cost of CO2 per unit mass.
(5)弃风惩罚成本(5) Wind abandonment penalty cost
式中：kloss为单位弃风功率惩罚成本；为场景s下t时段WT的预测出力。In the formula: kloss is the unit wind power penalty cost; Contribute to the prediction of WT in period t under scenario s.
(6)电、热交互成本(6) Electricity and heat interaction costs
式中：为t时段IES的购、售电单价；/>为t时段IES的购、售热单价；为场景s下t时段IES向电网售电功率；/>为场景s下t时段IES向热网购、售热功率。In the formula: is the unit price of electricity purchased and sold by IES during period t;/> is the purchase and sale unit price of IES during period t; IES sells electricity power to the grid for period t under scenario s;/> For scenario s and period t, IES purchases and sells thermal power online.
约束条件：Restrictions:
(1)功率平衡约束(1) Power balance constraints
(2)P2G-GPPCC系统运行约束(2)P2G-GPPCC system operation constraints
式中：PP2G,max和PP2G,min为场景s下t时段P2G的耗电功率上下限；PGPPCC,max和PGPPCC,min为GPPCC能耗上下限；ηc,max和ηc,min为碳捕集率上下限。In the formula: PP2G,max and PP2G,min are the upper and lower limits of P2G power consumption in scenario s during t period; PGPPCC,max and PGPPCC,min are the upper and lower limits of GPPCC energy consumption; ηc,max and ηc, min is the upper and lower limits of carbon capture rate.
(3)机组运行上下限约束(3) Unit operating upper and lower limit constraints
式中：PCHP,max和PCHP,min为CHP的出力上下限；PEB,max和PEB,min为EB运行功率上下限；HGB,max和HGB,min为GB的热出力上下限。In the formula: PCHP,max and PCHP,min are the upper and lower limits of CHP output; PEB,max and PEB,min are the upper and lower limits of EB operating power; HGB,max and HGB,min are the upper thermal output limits of GB. lower limit.
(4)电、热网络功率交互约束(4) Electrical and thermal network power interaction constraints
式中：和/>分别为IES向上级电网购、售电功率上限；/>和/>分别为IES向上级热网购、售热功率上限；/>为0-1变量，分别表示场景s下t时段IES与上级电、热网交互状态。In the formula: and/> They are the upper limit of the power purchased and sold by IES from the superior power grid;/> and/> They are the upper limit of thermal power for IES to purchase and sell thermal power to superiors;/> are 0-1 variables, respectively representing the interaction status between IES and the superior electricity and heating network during t period under scenario s.
(5)储能约束(5) Energy storage constraints
式中：EES,max和EES,min为ES的储能容量上下限；和/>为ES的充放功率上限；和/>为0-1变量，表示场景s下t时段ES的充、放能状态；EES,1,s和EES,T,s为ES在场景s下初始时刻和末尾时刻的容量，考虑储能设备运行的周期性，其在始末时刻的容量相等。HS与ES约束条件相同，不再赘述。In the formula: EES,max and EES,min are the upper and lower limits of the energy storage capacity of ES; and/> is the upper limit of charging and discharging power of ES; and/> is a 0-1 variable, indicating the charging and discharging state of ES in period t under scenario s; EES,1,s and EES,T,s are the capacity of ES at the initial moment and the end moment in scenario s, taking into account energy storage The periodicity of equipment operation, its capacity at the beginning and end moments is equal. The constraints of HS and ES are the same and will not be described again.
步骤9:在满足系统约束的条件下，采用YALMIP调用CPLEX 12.6进行优化求解得到系统内各设备的最优出力计划。其中负荷需求响应模型决策变量为各时段的电价变化量及负荷变化量，综合能源系统优化调度模型以系统内各设备在各时段的出力为决策变量。Step 9: Under the condition that the system constraints are met, YALMIP is used to call CPLEX 12.6 for optimization and solution to obtain the optimal output plan of each equipment in the system. Among them, the decision variables of the load demand response model are the changes in electricity prices and load changes in each period, and the integrated energy system optimization dispatch model uses the output of each equipment in the system in each period as the decision variables.
IES结构如图2所示，模型的具体求解流程见图3。The structure of IES is shown in Figure 2, and the specific solution process of the model is shown in Figure 3.
本发明使用碳价历史数据作为真实数据集来训练GAN，迭代训练次数最大设置为2000次，随着对抗训练的进行，判别器损失函数值LD的变化如图4所示。在刚开始迭代训练时，由于此时生成器还未能学习真实数据的分布特征，生成数据与真实数据样本的差异较大，因此判别器可以轻松判别输入的数据来源，D(x)接近1而D(G(z))接近0，故LD较大；随着交替训练的进行，生成器生成的数据逐渐真实，LD逐渐减小，最终训练达到稳态，判别器无法判别数据来源，生成器可以模拟真实碳价数据。生成的1000个碳价样本如图5所示。This invention uses carbon price historical data as a real data set to train GAN, and the maximum number of iterative trainings is set to 2000. As the adversarial training proceeds, the change of the discriminator loss function valueLD is shown in Figure 4. At the beginning of iterative training, because the generator has not yet learned the distribution characteristics of real data, the difference between the generated data and real data samples is large, so the discriminator can easily identify the source of the input data, and D(x) is close to 1 And D(G(z)) is close to 0, soLD is larger; as the alternating training proceeds, the data generated by the generator gradually becomes real,LD gradually decreases, and finally the training reaches a steady state, and the discriminator cannot distinguish the source of the data. , the generator can simulate real carbon price data. The 1,000 carbon price samples generated are shown in Figure 5.
经后向削减法削减后的5个典型场景的碳价及其概率如表1所示。The carbon prices and their probabilities of five typical scenarios after reduction using the backward reduction method are shown in Table 1.
表1各个场景对应的碳价及其发生概率Table 1 Carbon prices corresponding to each scenario and their probability of occurrence
对本发明所建的负荷需求响应模型进行求解，可得到需求响应前后电、热负荷曲线如图6所示。By solving the load demand response model built in the present invention, the electrical and thermal load curves before and after demand response can be obtained, as shown in Figure 6.
分析图6可知，在分时电价的引导下，电负荷峰谷差较响应前下降了33.79％、标准差较响应前下降了37.99％。对于热负荷，热负荷峰谷差较响应前下降了14.90％、标准差较响应前下降了17.74％，另外，响应后的热负荷与最优热电比下目标热负荷的欧氏距离减小了3.54％。由此可见，考虑电热综合需求响应能够有效平抑电、热负荷波动，同时使得整个调度周期内负荷的热电比更接近系统供能侧的热电比，实现负荷需求与能源供给之间的协调运行。Analysis of Figure 6 shows that under the guidance of time-of-use electricity prices, the peak-to-trough difference of electrical load dropped by 33.79% compared with before the response, and the standard deviation dropped by 37.99% compared with before the response. For the heat load, the peak-to-trough difference of the heat load dropped by 14.90% compared with before the response, and the standard deviation dropped by 17.74% than before the response. In addition, the Euclidean distance between the heat load after the response and the target heat load under the optimal heat-to-power ratio decreased. 3.54%. It can be seen that considering the comprehensive demand response of electricity and heat can effectively smooth the fluctuations of electricity and heat loads, and at the same time make the heat-to-power ratio of the load during the entire dispatch cycle closer to the heat-to-power ratio on the energy supply side of the system, achieving coordinated operation between load demand and energy supply.
为探究需求响应和碳价不确定性对IES调度结果的影响，构建以下4种情景进行对比分析：In order to explore the impact of demand response and carbon price uncertainty on IES dispatch results, the following four scenarios were constructed for comparative analysis:
情景1：不考虑综合需求响应，碳价取确定值进行优化调度；Scenario 1: Comprehensive demand response is not considered, and the carbon price takes a certain value for optimal dispatch;
情景2：考虑综合需求响应，碳价取确定值进行优化调度；Scenario 2: Considering comprehensive demand response, the carbon price takes a certain value for optimal dispatch;
情景3：不考虑综合需求响应，考虑碳价不确定性进行优化调度；Scenario 3: Do not consider comprehensive demand response, but consider carbon price uncertainty for optimal dispatch;
情景4：同时考虑综合需求响应和碳价不确定性进行优化调度。Scenario 4: Optimize dispatch while considering comprehensive demand response and carbon price uncertainty.
不同情景的调度结果如表2所示。The scheduling results of different scenarios are shown in Table 2.
表2不同情景调度结果对比Table 2 Comparison of scheduling results in different scenarios
与情景1相比，情景2由于考虑了需求响应的影响，总成本减少了1.2万元，同时弃风成本减少了59.52％，能源利用率由80.79％提高到了85.43％，一方面是因为IDR的削峰填谷作用，引导峰时电负荷转移至风电资源较丰富的谷时段，减少因谷时负荷需求量较低而造成的弃风，使得新能源弃用率降低，同时降低CHP的供能压力，间接减少系统净碳排放量；另一方面，响应后的负荷波动性降低，系统供需侧的热电匹配度提高，使得系统内各设备的出力更平稳，提高了CHP的运行效率，进一步提高系统的能源利用效率。Compared with Scenario 1, Scenario 2 takes into account the impact of demand response, and the total cost is reduced by 12,000 yuan. At the same time, the wind curtailment cost is reduced by 59.52%, and the energy utilization rate is increased from 80.79% to 85.43%. On the one hand, it is because of the IDR The function of peak shaving and valley filling guides the transfer of peak electricity load to valley periods when wind power resources are abundant, reducing wind abandonment caused by low load demand in valley hours, reducing the abandonment rate of new energy sources and reducing the energy supply of CHP. pressure, indirectly reducing the net carbon emissions of the system; on the other hand, the load fluctuation after response is reduced, and the heat and power matching on the supply and demand side of the system is improved, making the output of each equipment in the system more stable, improving the operating efficiency of CHP, and further improving System energy efficiency.
与情景1相比，情景3的总成本明显减小，碳交易成本减少了63.16％，弃风成本减少了30.95％，净碳排放量减少了209.81吨，但碳封存成本增加了0.57万元，这是由于考虑了碳价不确定性后，系统在碳价较高的场景下通过碳捕集系统捕集大部分CO2，减少碳排放从而可以出售多余的碳排放配额降低碳交易成本，同时碳捕集系统运行所需能耗可有效消纳弃风，减少弃风成本；情景4综合考虑需求响应和碳价不确定性，总成本和净碳排放量进一步降低，使得系统经济性和低碳性提高。Compared with scenario 1, the total cost of scenario 3 is significantly reduced, with carbon trading costs reduced by 63.16%, wind curtailment costs reduced by 30.95%, net carbon emissions reduced by 209.81 tons, but carbon storage costs increased by 5,700 yuan. This is because after taking into account the uncertainty of carbon prices, the system captures most of the CO2 through the carbon capture system in scenarios with high carbon prices, reducing carbon emissions and selling excess carbon emission quotas to reduce carbon trading costs. The energy consumption required for the operation of the carbon capture system can effectively absorb wind curtailment and reduce the cost of wind curtailment; Scenario 4 comprehensively considers demand response and carbon price uncertainty, and the total cost and net carbon emissions are further reduced, making the system economical and low Carbonity is improved.
情景4的优化调度结果如图7、图8所示，其中，图7为IES的供电调度优化结果，ES出力大于0时为放电，小于0时为充电，电网交互功率大于0时为购电，小于0时为售电；图8为供热调度优化结果，HS出力大于0时为放热，小于0时为储热，热网交互功率大于0时为购热，小于0时为售热。The optimized dispatch results of Scenario 4 are shown in Figures 7 and 8. Figure 7 shows the power supply dispatch optimization results of IES. When the ES output is greater than 0, it is discharge, when it is less than 0, it is charging, and when the grid interactive power is greater than 0, it is electricity purchase. , when it is less than 0, it is electricity sales; Figure 8 shows the heating dispatch optimization result. When the HS output is greater than 0, it is heat release, and when it is less than 0, it is heat storage. When the heat network interactive power is greater than 0, it is heat purchase, and when it is less than 0, it is heat sale. .
联系峰谷分时电价比较优化结果分析：在01:00-04:00时段风电资源较为丰富，电负荷需求较低而热负荷需求较高，由于GB、EB的出力限制，其满负荷工作仍无法满足热负荷需求，因此不足的热负荷需由CHP补充，电能由风电和CHP满足，盈余的电能由ES存储以及向电网出售以降低运行成本；而在05:00-08:00时段电价为低谷期且可用风电资源减少，此时系统向电网低价购电满足运行需求；在09:00-18:00时段购电价格以及电负荷需求升高且可用风电资源较少，购电成本大于CHP发电成本，故系统增加CHP出力，同时释放ES在负荷低谷期和电价较低时段储存的电能来满足电负荷需求，由于该时段电价较高，系统出售部分电能降低运行成本；19:00-08:00为风电出力高峰时段，富余的风电通过P2G进一步消纳；此外，热负荷需求在11:00-18:00等日间时段处于低谷，系统盈余的热能一部分通过HS储存用于夜间热负荷高峰时段，一部分向热网出售提高经济效益。Analysis of the peak and valley time-of-use electricity price comparison optimization results: Wind power resources are relatively abundant during the 01:00-04:00 period, the electricity load demand is low and the heat load demand is high. Due to the output limitations of GB and EB, its full load operation is still The heat load demand cannot be met, so the insufficient heat load needs to be supplemented by CHP, the electric energy is met by wind power and CHP, and the surplus electric energy is stored by ES and sold to the grid to reduce operating costs; while the electricity price during 05:00-08:00 is During the trough period and the available wind power resources are reduced, the system purchases electricity from the power grid at a low price to meet operating needs; during the 09:00-18:00 period, the electricity purchase price and power load demand increase and the available wind power resources are less, and the power purchase cost is greater than CHP power generation costs, so the system increases CHP output, and at the same time releases the electric energy stored by ES during the load low period and low electricity price period to meet the electric load demand. Due to the high electricity price during this period, the system sells part of the electric energy to reduce operating costs; 19:00- 08:00 is the peak period of wind power output, and the excess wind power is further consumed through P2G; in addition, the heat load demand is at a low during the daytime periods such as 11:00-18:00, and part of the system's surplus thermal energy is stored through HS for night heating. During peak load hours, part of it is sold to the heating network to improve economic efficiency.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311505996.7ACN117709511A (en) | 2023-11-13 | 2023-11-13 | Comprehensive energy system optimal scheduling method considering demand response and carbon price uncertainty |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311505996.7ACN117709511A (en) | 2023-11-13 | 2023-11-13 | Comprehensive energy system optimal scheduling method considering demand response and carbon price uncertainty |
| Publication Number | Publication Date |
|---|---|
| CN117709511Atrue CN117709511A (en) | 2024-03-15 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311505996.7APendingCN117709511A (en) | 2023-11-13 | 2023-11-13 | Comprehensive energy system optimal scheduling method considering demand response and carbon price uncertainty |
| Country | Link |
|---|---|
| CN (1) | CN117709511A (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118607769A (en)* | 2024-06-05 | 2024-09-06 | 辽宁石油化工大学 | A dispatching method for integrated energy system |
| CN119151271A (en)* | 2024-11-21 | 2024-12-17 | 山东大学 | Electric and thermal comprehensive energy system construction method introduced into carbon trade market |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118607769A (en)* | 2024-06-05 | 2024-09-06 | 辽宁石油化工大学 | A dispatching method for integrated energy system |
| CN119151271A (en)* | 2024-11-21 | 2024-12-17 | 山东大学 | Electric and thermal comprehensive energy system construction method introduced into carbon trade market |
| Publication | Publication Date | Title |
|---|---|---|
| Tan et al. | Combined electricity-heat-cooling-gas load forecasting model for integrated energy system based on multi-task learning and least square support vector machine | |
| Jin et al. | Game theoretical analysis on capacity configuration for microgrid based on multi-agent system | |
| CN106228258B (en) | It is a kind of meter and demand side management home energy source local area network energy optimal control method | |
| CN117709511A (en) | Comprehensive energy system optimal scheduling method considering demand response and carbon price uncertainty | |
| CN108154309A (en) | The energy internet economy dispatching method of meter and the more load dynamic responses of cool and thermal power | |
| CN112131712B (en) | Multi-objective optimization method and system for multi-energy system on client side | |
| CN114266468B (en) | Planning method of integrated energy system for industrial parks considering demand response under carbon constraints | |
| CN111681130A (en) | An Optimal Scheduling Method of Integrated Energy System Considering Conditional Value-at-Risk | |
| CN112580897B (en) | Method for optimal scheduling of multi-energy power system based on parrot algorithm | |
| CN104065072A (en) | A microgrid operation optimization method based on dynamic electricity price | |
| CN116542370A (en) | Park low-carbon economic operation method considering carbon capture and carbon transaction | |
| CN114997715A (en) | Configuration method of combined cooling, heating and power system based on improved fuzzy C-means clustering | |
| CN117726143A (en) | Environment-friendly microgrid optimization scheduling method and system based on deep reinforcement learning | |
| CN116011697A (en) | A method and system for optimal scheduling considering P2X multi-energy coupling operation | |
| CN112068436A (en) | Layered and distributed control method and system for comprehensive energy system of industrial park | |
| CN115293485A (en) | Low-carbon scheduling method of comprehensive energy system considering electric automobile and demand response | |
| Xue et al. | Optimal capacity allocation method of integrated energy system considering renewable energy uncertainty | |
| CN117996766A (en) | Charging station-containing micro-grid multi-target energy optimal scheduling method considering carbon emission | |
| CN110489915A (en) | Meter and the electric heating combined scheduling method and system of integration requirement response | |
| CN116865233A (en) | A low-carbon economic operation optimization method for microgrids containing electricity-to-gas equipment | |
| Luo et al. | A multi-objective dual dynamic genetic algorithm-based approach for thermoelectric optimization of integrated urban energy systems | |
| CN116307437A (en) | Virtual power plant optimal scheduling method and system based on carbon emission right exchange mechanism | |
| CN114611793A (en) | Comprehensive energy system optimization method and device considering short-term load prediction | |
| CN117350410B (en) | Micro-grid group cooperative operation optimization method based on multi-agent federation reinforcement learning | |
| CN118171884A (en) | Model predictive control-based park energy system optimal scheduling method and related device thereof |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |