Analytical Optimisation of Industrial Systems and Applications to Refineries, Petrochemicals

A new method for optimising process networks is presented in this paper. The method uses economic analysis of existing systems based on the value analysis method derived by Sadhukhan (2002) as the basis to derive the optimum network design. Optimising a large scale industrial system (e.g. refineries, petrochemicals) where multiple processes, many material streams and a number of supporting systems (e.g. energy) are involved, is a difficult task to achieve. For such a case, a fundamental, practical and systematic methodology for detailed differential economic analysis (Sadhukhan et al, 2003) of an industrial system at any market and environmental condition can be very useful for achieving its optimal operation. Application of the non-linear discrete / continuous mathematical programming techniques (Grossmann and Daichendt, 1996; Grossmann et al, 1999) for process network optimisation does not provide a clear and transparent economic value structure of individual components (streams and processes) in a network. Overcoming this drawback is the sole driving force of this work that aims to develop a novel optimisation technique called analytical optimisation, for process industries (Sadhukhan et al, 2004). The essence of integration and optimisation of process networks is the evaluation and differential analysis of the economic performance of individual streams and processes. It builds from graph theory by representing any production network as a graph that consists of arcs (streams) and nodes (processes), interconnected to form paths and trees (Mah, 1983). Every stream in a process network can be characterized by a value on processing and a cost of production. Once they are evaluated for a stream the difference between them provides the specific economic margin achieved from the stream. The profit margins of individual paths and trees and finally the entire network are predicted from the economic margins of streams. Using this new value analysis method an overall integration strategy is developed by Sadhukhan et al (2003) so as to capture the impacts of real plant operations (no fixed operating conditions) and the effects of network interactions in the detailed economic analysis of a complex system. The analytical optimisation procedure by Sadhukhan et al (2004) is designed based on comprehensive economic analysis of process networks discussed above for maximising the overall system economics. The economic analysis of systems helps to identify the weakest links and their integration opportunities existing in a network configuration. Thus, an optimum network flowsheet can be derived using the existing processes and also including new process technologies as necessary. The analytical optimisation of a process network comprises of three steps. Market integration is the first step that fully exploits the available market opportunities for selling and purchasing streams based on individual marginal contributions from productions and processing of streams. This activity does not incur any capital investment and is an easy and straight – forward way for achieving quick benefits. The second step deals with optimisation of network flowsheet / connections. The economic margins of various paths of network are used to determine the weaker paths and their corresponding stronger paths where the loads of weaker paths can be shifted. This load shifting among paths leads up to the overall benefits of a system. In general, this stage involves minor capital investment for piping, rerouting of streams etc. At the core of the analysis there is the opportunity to perform a detailed process level optimisation for improvement of non-profitable / less profitable process units. Thus it deals with optimisation of individual processes in order to improve their marginal contributions in the overall economics. Optimisation of processes can incur significant capital investment if integration of new process technologies is considered. For all the three considerations the methodology has been further extended to determine the achievable marginal benefit from every modification suggested. Thus, using such a technique the more economically profitable projects can be readily screened for further evaluation. All the above three aspects of analytical optimisation have been illustrated with the help of refinery case studies (Sadhukhan et al, 2003 and 2004). The effectiveness of the methodology has also been demonstrated in design and scheduling of petrochemical complexes in changing economic scenarios. 1. Market integration of process networks Sadhukhan et al (2003) has developed marginal correlations for elements (trees and paths) of production and processing of a stream in terms of its market price, cost of production (COP) (Eq. 1) and value on processing (VOP), market price (Eq. 2) respectively. u US UNIT i ∈ ∑ _ ( ) ∆u(Fe-i) = {(Fe-i)MP − (Fe-i)COP} × e i F m − ∀i∈UNIT, e∈EU(i) (Eq. 1) d DS UNIT i ∈ ∑ _ ( ) ∆d(Fe-i) + ∆i(Fe-i) = {(Fe-i)VOP − (Fe-i)MP} × e i F m − ∀i∈UNIT, e∈EU(i) (Eq. 2) Summation of the two expressions provides the economic margin of the entire element where the stream belongs to (Eq. 3). u US UNIT i ∈ ∑ _ ( ) ∆u(Fe-i) + d DS UNIT i ∈ ∑ _ ( ) ∆d(Fe-i) + ∆i(Fe-i) = {(Fe-i)VOP − (Fe-i)COP} × e i F m − ∀i∈UNIT, e∈EU(i) (Eq. 3) These marginal correlations (Eqs. 1-3) are used to demonstrate how the various market integration opportunities can be fully exploited in a process network. There are 3! arrangements possible among the three values of a stream, VOP, COP and market price (VOP > COP > market price; VOP > market price > COP; market price > VOP > COP; and so on) based on which various market integration strategies are developed. In an ideal situation, the market price of a stream is in between its VOP and COP {(Fe-i)VOP ≥ (Fe-i)MP ≥ (Fe-i)COP}. The margins incurred from both the elements of processing and production of the stream are positive (Eqs. 1-2) and so is the overall economic margin of the stream. No market strategy is required to improve the economics. However, if production of a stream is non-profitable (market price of the stream is less than its COP), an improvement in the stream’s economic margin can be achieved by reducing production of the stream and instead increasing purchasing of the stream. The difference between the COP and the market price is the scope for value improvement of the stream per unit flowrate. The marginal improvement achievable from the stream is determined by multiplying the scope for value improvement with the purchasing potential of the stream given by the difference between the current flowrate and the minimum production requirement of the stream. 2. Optimisation of network connections The steps for optimising network connections at site level are as follows. (1) The streams in an existing network system are evaluated for VOP and COP (Sadhukhan et al, 2003). Using Eqs. 1-3 economic margins of elements (paths) are determined. (2) The weaker or less efficient paths (with negative / lower economic margins) are screened based on individual economic margins (Eqs. 1-3) and organised in the ascending order of economic margins at any design stage k. (3) For the current weakest path with the least economic margin the stronger optional paths with better economic margins in the current system at any design stage k are identified where the loads of the current weakest path can be shifted. If no optional path is identified the next weakest path and its set of stronger optional paths are taken into consideration at the design stage k. (4) The load from the current weakest path is shifted to an identified stronger optional path at any design stage k. Thus a new link is set up between the current weakest path and the stronger optional path. (5) The maximum marginal benefit achievable from every such new link set up is determined. If the marginal benefit achieved is justified the network is updated with the new link and a new flowsheet is obtained for optimisation in the next design stage k + 1. (6) The procedure from step 1 is followed until no integration opportunity is identified. k = k + 1. The various steps of analytical optimisation of network connections are presented in Figure 1. Due to load shifting from the weaker paths to the stronger paths, the weaker paths receive a share of the stronger economics. The net result is optimisation of overall network connections resulting from integration between weaker and stronger paths. 2.1 Correlations of marginal benefits from new links set up The marginal benefit achieved from a load shifting is calculated from improvement in economic margin of the shifted load (Sadhukhan et al, 2004). In case the COP remains the same for the shifted load, the marginal benefit is equal to the improvement in VOP multiplied by the flowrate of the shifted load. Figure 1. Analytical optimisation procedure Figure 2. Illustration of system optimisation for network connections. through improvement of existing processes. 2.2 Determination of optimal load shifting in new links set up For every new link set up the optimum amount of load shifting is determined in order to predict the maximum marginal benefit achievable from a new link set up. The marginal benefit achievable from a new link set up is a variable depending on the amount and VOP, COP of the load to be shifted. The amount of load to be shifted is a decision variable whereas VOP and COP of the load are dependent on process operating conditions. There are two levels of models to be used to determine the optimal load shifting and thus the maximum marginal benefit from a new link set up, site level and process level models (Figure 1). The site level model predicts VOP, COP of streams (Sadhukhan et al, 2003) and marginal benefit (Sadhukhan et al, 2004) for a given amount of load shifting through a link. The process level model (Sadhukhan et al, 2004; Sadhukhan and Zhu, 2002) provides the process operating conditions (operating costs, yields and properties of streams) as the input to the site level model. Thus, the site level and the process level models co-ordinate with each other to generate a relationship between the marginal benefit and the amount of load shifting through a new link set up. This relationship is finally used to determine the optimal load shifting for which the marginal benefit is the maximum. Organise non-profitable paths in the descending order of individual marginal losses (stage k=k+1) Identify the current weakest path at stage k Any non-profitable path left? stage k