ПОИСК ТРАНСПОРТНОГО СТОХАСТИЧЕСКОГО РАВНОВЕСИЯ С ПРИЛОЖЕНИЯМИ В ГАЗОТРАНСПОРТНОЙ МОДЕЛИ РОССИИ
- Авторы: Колосницын А.В.1
-
Учреждения:
- Институт систем энергетики им. Л. А. Мелентьева СО РАН
- Выпуск: Том 17, № 1 (2016)
- Страницы: 56-61
- Раздел: Статьи
- URL: https://journals.eco-vector.com/2712-8970/article/view/504705
- ID: 504705
Цитировать
Полный текст
Аннотация
Рассматривается поиск равновесия в условиях неопределенности на примере сетевой газотранспортной модели России. Модель включает в себя поставщиков и потребителей природного газа, объединенных сетью транспортных линий. Подробно описывается двухэтапный подход к поиску сетевого равновесия. Для первого этапа разбирается способ формирования функций спроса и предложения для потребителей и поставщиков газа, а также поиск равновесной цены, объемов производства и потребления газа. На втором этапе ставится задача определения оптимального плана транспортировок газа с учетом сетевых ограничений. Рассматривается случай неопределенности спроса, и ставится задача поиска стохастического равновесия в газотранспортной модели. Определяются объемы производства и цена на газ, которые обеспечивают удовлетворение спроса с заданной вероятностью. Описанная методика поиска стохастического равновесия применяется к газотранспортной модели России. Приводятся результаты численных расчетов.
Полный текст
Introduction. We consider a gas network system in conditions of deregulated operation. This means that we have to take into account different (not necessarily antagonistic) aims of the participants of the system. The below suggested model is based on game-theoretical approach (see also [1]). The key point is consideration of stochastic gas demand uncertainty. In general, uncertainty is one of the main problems in energy system modelling [2]. We assume that uncertainty has stochastic nature, i. e. all uncertain parameters are random variables with known distributions. Different approaches are used for studying games with stochastic uncertainty. In [3] authors base their investigations on Markov chains and stochastic dynamic programming. Random variational inequalities are used for finding the Wardrop traffic equilibria in [4]. Among other interesting and relevant topics the following can be mentioned: stochastic differential equations [5], stochastic Stackelberg games [6], generalization of the maxmin strategy [7; 8], stochastic Cournot models [9], two-stage stochastic programming model [10]. In our approach we use standard stochastic programming technique [11; 12] and basic knowledge of probability theory [13]. Recent investigation concerning the paper topic can be found in [14] and [15]. The standard transportation model that joins gas producers and consumers and takes into account network constraints is considered. Minimizing total network costs, we obtain optimal gas transportation plan as a solution of the following mathematical programming problem: (1) Here A is an m×n incidence matrix, which defines relations between m nodes and n transportation lines; xj - volume of the gas, transporting on the line j; cj(xj) - transportation costs; bi - volume of the gas production or consumption in the node i; dj - capacity of the line j. Two stage approach to finding the network equilibrium. We assume that the considering model reflects operation in the market conditions. It means that supply and demand functions are given in every production and consumption node respectively. Having that data, we can build total demand and supply functions that let us to find an equilibrium price, production and consumption volumes of gas in corresponding nodes. Let Ip be node set of producers, Ic - node set of consumers, I0 - set of the branching nodes. Production costs have nonlinear structure in general and for our purposes we represent it as the quadratic functions that practically appropriate: (2) where , - specified constants. Consumption linearly depends on market price p: where , - specified constants. The following explanation can be given here to make justification to introduced functions representations. Production cost form is our assumption as the model designer, whereas consumption function form is standard assumption that reflects inverse dependency of the consumption volume from the price level. In the branch nodes we have One can naturally suppose that producers maximize their profit: (3) where p - market price on the product. Then using (2), (3) we can build supply function of every producer: (4) Total supply equals to (5) Total demand function equals to Intersection point of total supply and demand functions defines an equilibrium price p*, which is set on the market of gas production and consumption: (6) We consider (6) as the key condition in defining the state of the network equilibrium in the represented model. Situation on the market of gas production and consumption in which total volume of production equals to the total volume of consumption taking into account network constraints will be understood to be the network equilibrium. Note that considering market conditions it is quite reasonable to divide searching the network equilibrium into two stages. On the first stage we find equilibrium market price and also equilibrium volumes of gas production and consumption. To do this we need to implement next sequence of steps: 1. Considering gas production costs build supply function of each gas producer in the form of (4). 2. Build total supply function S(p). 3. Build total demand function D(p). 4. Using the relation (6) find the equilibrium price p*. 5. Determine equilibrium volumes of gas production for each provider: 6. Determine equilibrium volumes of gas demand for each consumer: On the second stage we optimize gas transportation volumes considering throughput of each transport line and demand satisfaction in every consumption node. We can represent mathematical model of this problem in the next form: (7) where - equilibrium volumes of gas production and consumption that were obtained on the first stage. Solution of the problem (7) is optimal plan of gas transportation in the market conditions with network constraints. Network equilibrium with demand uncertainty. Sometimes we can come across the difficulty in determination of consumers demand functions. In this case we can consider its uncertainty. In our model demand functions have the form . Now let the parameters and be random variables with normal distribution and known expectations and variances. Hence, and become random variables as well with normal distribution. This representation of demand function includes a particular case, when parameters . Then demand value of each consumer will be determined by the random variable , with normal distribution. In the uncertainty conditions we can set the problem of searching such price and corresponding gas production volumes that provide satisfaction gas consumption with the probability not less than given positive constant δ. It means that we move to finding stochastic equilibrium in our model. We can write this condition formally in the next form: (8) We can rewrite (8) as , where - is the distribution function of the random variable with fixed price. It is well known that the probability of random variable realization from the interval can be found using Laplace function [13]: To use Laplace function we need to normalize the random variable , that is count expression where and - are expectation and variance of the random variable accordingly. Using Laplace function we have (9) transform the expression (9): or where Total demand will be entirely satisfied with the probability not less then δ with the price . Following the two stage approach of finding the stochastic equilibrium in describing model, now we need to solve gas transportation problem, using the obtained price and gas production volume. For this purpose it is quite convenient to use statement of the convex programming problem with chance constraints [11; 12]: (10) Here on the one hand, expressions - are chance constraints, reflecting the requirement of demand satisfaction in each consumption node with the determine probabilities , and with fixed price . On the other hand, this constraints may be considered as distribution functions of the random variables . It is important to note that for each consumption node the probability of demand satisfaction differs from the probability which is set on the first stage problem. It is related with weaker condition of demand satisfaction for whole model in general in comparison with the condition of demand satisfaction for each consumption node with the same probability. In order to find probabilities of demand satisfaction in every consumption node, we consider , to be the variables, that we try to approximate to the probability of demand satisfaction of the first stage problem as close as possible. Note that throughput constraints may prevent us from obtaining the solution of the problem (10), that is why we add variables , , which give us the possibility to broaden throughput of transport lines if it is necessary. After determination of the values , , we can use the Laplace function for moving to the determine equivalent in the next form: (11) where , . Constants , - are capital costs for broaden transport line j. Solution of the problem (11) is the optimal plan of gas transportation in the market conditions with the demand uncertainty. Stochastic network equilibrium in the gas transportation model of Russia. Described approach to finding the stochastic equilibrium was applied to gas transportation model of Russia, which includes 70 transport lines and 51 nodes. There are 9 of these nodes that contain gas producers, 41 nodes with consumers and 4 branch nodes (fig. 1). Fig. 1. Gas transportation system of Russia Basic data of the model let us to set the production cost in the form (2) for every gas producer and build supply functions and total supply function in the form (5). The demand functions were set for each consumer in two variants. In the first variant we set the deterministic parameters of the demand functions and then we obtained and solved deterministic problem of the finding the network equilibrium. In this case demand functions were formed automatically with using special technique described in [16] by means of solving next mathematical programming problem: (12) We can define dual variables for equalities , from the solution of the problem (12). These variables or shadow prices reflect the resource value for producers and consumers and equal for them out of the problem properties (matrix A is the incidence matrix). Denote this price as . We can assume that the demand elasticity is known (almost inelastic demand). Moreover we know the desirable consumption level from basic data of the model, denote it as , . To set the consumption functions in the form , we need to find parameters and , from the following equations system: (13) For instance we need to deliver 27 units of gas to Tyumen, elasticity of demand equals and shadow price obtained from problem (12) equals 44.3 conventional units. Use the system (13): Finally we get and . Consumption function in Tyumen equals to b1(p) = 28.89 - 0.043p. Second variant of our gas system modeling included stochastic uncertainty of the demand functions that led to another solution of the stochastic network equilibrium problem. Both variants of problems were solved using program complex GAMS [17], the results are demonstrated in the fig. 2, where it is showed how demand uncertainty influence the main parameters of our model. We obtained the following results for deterministic model: billions m3, price for billion m3 of gas equals to 33.9 conventional units. For stochastic model we set the level of demand satisfaction with the probability equals to 90 % and we got billions m3 and price equals to 36.7 conventional units. As we can see, considering demand uncertainty increase the price, supply of the gas and the network load in the transport model. We received 4.5 % of the price growth, 8.3 % of the supply volume growth and 37.5 % of the network load growth under uncertainty conditions with the 90 % of demand satisfaction probability (fig. 2). Fig. 2. How demand uncertainty influence the main parameters of the model Conclusion. We represented the methodology of the finding of stochastic network equilibrium with specified probability level for natural gas demand volumes. Our approach consists of two stages. On the first stage we defined the equilibrium volumes of natural gas consumption and supply and equilibrium price. On the second stage we obtained the optimal plan of gas transportation. To consider the gas consumption uncertainty we modified our two-stage approach to finding the network equilibrium and obtained problem statement for finding the stochastic network equilibrium that included convex programming problem with chance constraints. This problem was solved by program complex GAMS. Our method of finding the stochastic network equilibrium was successfully applied to the gas transportation model of Russia.×
Об авторах
А. В. Колосницын
Институт систем энергетики им. Л. А. Мелентьева СО РАН
Email: ankolos25@mail.ru
Российская Федерация, 664033, г. Иркутск, ул. Лермонтова, 130
Список литературы
- Klassen G., Kryazhimskii A. V., Tarasyev A. M. Multiequilibrium game of timing and competition of gas pipeline projects. Journal of Optimization Theory and Applications, 2004. Vol. 120, No. 1, P. 147-179.
- Weber C. Uncertainty in the electric power industry: methods and models for decision support. Springer, 2005, 291 p.
- Bonnans J. F., Cen Z., Christel T. Energy contract management by stochastic programming technique. Annals of Operations Research, 2012, Vol. 200, P. 199-222.
- Gwinner J., Racity F. Some equilibrium problems under uncertainty and random variational inequalities. Annals of Operations Research, 2012, Vol. 200, P. 299-319.
- Bensoussan A., Frehse J. Stochastic games for N players. Journal of Optimization Theory and Applications, 2000, Vol. 105, No. 3, P. 534-565.
- Evgrafov A., Patriksson M. On the existence of solutions to stochastic mathematical programs with equilibrium constraints. Journal of Optimization Theory and Applications, 2004, Vol. 121, No. 1, P. 65-76.
- Zhukovskiy V., Kudryavtsev K. [Equilibrating conflicts under uncertainty. I. Analogue of a saddle-point]. Matematicheskaya teoriya igr i eye prilozheniya. 2013, Vol. 5(1), P. 27-44 (In Russ.).
- Zhukovskiy V., Kudryavtsev K. [Equilibrating conflicts under uncertainty. II. Analogue of a maximin]. Matematicheskaya teoriya igr i eye prilozheniya. 2013, Vol. 5(2), P. 3-45 (In Russ).
- Zhang D., Xu Y., Wu Y. A two stage stochastic equilibrium model for electricity markets with two way contracts. Mathematical Methods of Operations Research, 2010, Vol. 71, P. 1-45.
- Shapiro A. Stochastic programming with equilibrium constraints. Journal of Optimization Theory and Applications. 2006, Vol. 128, No. 1, P. 223-243.
- Kall P., Wallace S.W. Stochastic programming. John Wiley & Sons, 1994, 307 p.
- Judin D. Matematicheskie metody upravleniya v usloviyakh nepolnoj informatsii [Mathematical methods of controlling under incomlete information conditions]. Moscow, Sovietskoe radio Publ., 1974, 400 p.
- Gmurman V. Rukovodstvo k resheniyu zadach po teorii veroyatnostej i matematicheskoj statistike [Guidance to solving the probability theory and mathematical statistics problems]. Moscow, Vyshaya shkola publ., 1998, 400 p.
- Dolgov S., Il’kevich N., Rabchuk V. Upravlenie rezhimami raboty potrebitelej gaza pri narusheniyakh gazosnabzheniya [Mode control operation of gas consumers in violation of gas supply]. Novosibirsk, “Geo” Academic Publ., 2013, 90 p.
- Il’kevich N., Dzyubina T., Kalinina Z. H. Mnogourovnevoe modelirovanie razvitiya sistem gazosnabzheniya [Multilevel modeling of gas-supply systems]. Novosibirsk, Nauka Publ., 2014, 217 p.
- Podkoval’nikov S. V., Khamisov O. V. [Imperfect Electric Power Markets: Modelling and Investigation of Development of Capacities]. Izvestiya Akademii Nauk. Energetika, 2011, No. 2, P. 57-76 (In Russ.).
- Rosenthal R. GAMS - A user’s guide. GAMS development corporation. Available at: www.gams.com (accessed 26.09.2015).
- Klassen G., Kryazhimskii A. V., Tarasyev A. M. Multiequilibrium game of timing and competition of gas pipeline projects // Journal of Optimization Theory and Applications. 2004, Vol. 120, № 1. P. 147-179.
- Weber C. Uncertainty in the electric power industry: methods and models for decision support. Springer, 2005. 291 p.
- Bonnans J. F., Cen Z., Christel T. Energy contract management by stochastic programming technique // Annals of Operations Research. 2012. Vol. 200. P. 199-222.
- Gwinner J., Racity F. Some equilibrium problems under uncertainty and random variational inequalities // Annals of Operations Research. 2012. Vol. 200. P. 299-319.
- Bensoussan A., Frehse J. Stochastic games for N players // Journal of Optimization Theory and Applications. 2000. Vol. 105, № 3. P. 534-565.
- Evgrafov A., Patriksson M. On the existence of solutions to stochastic mathematical programs with equilibrium constraints // Journal of Optimization Theory and Applications. 2004. Vol. 121, № 1. P. 65-76.
- Жуковский В. И., Кудрявцев К. Н. Уравновешивание конфликтов при неопределённости. I. Аналог седловой точки // Математическая теория игр и её приложения. 2013. Т. 5, № 1. C. 27-44.
- Жуковский В. И., Кудрявцев К. Н. Уравновешивание конфликтов при неопределённости. II. Аналог максимина // Математическая теория игр и её приложения. 2013. Т. 5, № 2. С. 3-45.
- Zhang D., Xu Y., Wu Y. A two stage stochastic equilibrium model for electricity markets with two way contracts // Mathematical Methods of Operations Research. 2010. Vol. 71. P. 1-45.
- Shapiro A. Stochastic programming with equilibrium constraints // Journal of Optimization Theory and Applications. 2006. Vol. 128, № 1. P. 223-243.
- Kall P., Wallace S. W. Stochastic programming. John Wiley & Sons, 1994. 307 p.
- Юдин Д. Б. Математические методы управления в условиях неполной информации. М. : Советское радио, 1974. 400 с.
- Гмурман В. E. Теория вероятностей и математическая статистика. М. : Высш. шк., 2002. 479 c.
- Долгов С. И., Илькевич Н. И., Рабчук В. И. Управление режимами работы потребителей газа при нарушениях газоснабжения / отв. ред. С. М. Сендеров. Новосибирск : Гео, 2013. 90 с.
- Илькевич Н. И., Дзюбина Т. В., Калинина Ж. В. Многоуровневое моделирование развития систем газоснабжения. Новосибирск : Наука, 2014. 217 с.
- Подковальников С. В., Хамисов О. В. Несовершенные электроэнергетические рынки: моделирование и исследование развития генерирующих мощностей // Известия РАН. Энергетика. 2011. № 2. C. 57-76.
- Rosenthal R. GAMS. A user’s guide. GAMS development corporation [Электронный ресурс]. URL: www.gams.com (дата обращения: 26.09.2015).
Дополнительные файлы
![](/img/style/loading.gif)