Academia.eduAcademia.edu

Figure 3 – uploaded by Annamaria Signorini

See full PDFdownloadDownload figure
Fig. 2. Simplified CSPP Simulink model. focused on the turbines, the electro-hydraulic system and electric generator. a a a, ee a oe |

Figure 2 Simplified CSPP Simulink model. focused on the turbines, the electro-hydraulic system and electric generator. a a a, ee a oe |

Related Figures (10)

controller designed by means of a H,, loop-shaping robust method as in [15], or internal model control techniques for the design of a robust Proportional Integrative Derivative (PID) controller in a speed control application as in [16]. Predictive control approaches have been in- vestigated in [17], where the General Predictive Control and Con- strained Receding-Horizon Predictive Control have been exploited for the control of large steam turbines during load variations and in [18], where an Exponential Auto-Regressive eXogenous (ARX) model-based multistep predictive control algorithm is applied to control a thermal power plant. An interesting MPC application for the design of fault tolerant control systems is presented by Salahshoor et al. in [19], where a Generalized Predictive Control, combined with support vector ma- chine and adaptive neuro-fuzzy inference system, allows to recover the normal operation also in different fault scenarios. A complex control scheme, presented in [20], exploits a combined action of a feedforward controller based on nonlinear programming and a feedback controller designed with robust methods involving turbine governor valve and others control inputs, in order to control a fossil power plant in different load conditions. variations in the steam production, with the possibility to continuously operate by incorporating thermal energy storage or backup system in the plant, such as discussed by Zhang et al. in [8]. The continuity of energy production of a CSPP is also one of the main factors that may result effective in the thermal conversion of solar energy with respect to the photovoltaic, as highlighted by Desideri et al. in [9]. This approach can be coupled to non-standard control techniques, which allow to face the complexity of CSPP systems, by exploiting, for instance, a Model Predictive Control (MPC) approach, such as proposed by Vassallo and Bravo in [10], or adaptive control approaches and Artificial Intelligence (AD techniques, such as highlighted by Camacho et al. in [11]. The Steam Turbine (ST) has been originally designed for energy production from fossil fuels, thus both their mechanics and their control systems have been designed to face a quite stable steam production and very rare start-up and shut down cycles. As a consequence, in the context of CSPP, the standard PID control techniques currently applied to turbo-machinery are often not effective in the automatic adaptation to changing operating conditions of the machine [12]. Control system parameters are determined on the brand-new machine during the commissioning phase, with time consuming and effort-intensive pro- cedures. In addition, these parameters are only occasionally re-adjusted on the basis of semi-heuristic procedures. This forces the machine to work in non-optimal efficiency conditions during its lifetime. In the field of AI, Fuzzy Logic (FL) and Evolutionary Algorithm: (EA) have sometimes been exploited to control steam turbines. For in- stance, in [21] FL is used in order to adjust the parameters of a PID controller during speed control, while PID controllers whose para- meters are tuned through different EA methods are proposed in [22] and in [23]. However, none of the above-mentioned approaches defi- nitely proved to overcome the other ones in any operating scenarios. On the other hand, the control procedures need to allow correct and efficient machine operation in transient conditions without compro- mising its integrity, in particular focusing on variable operating con- ditions where steam turbines are subjected to continuous thermal stresses, premature aging and consequent lowering of efficiency. The turbine control problem with variable steam conditions can be ad- dressed by looking for applicability of innovative control strategy. Adaptive Control approaches, which allow to adapt gains in different loading conditions and uncertainties are known. An example of this kind of strategy is described in [13], where the turbine speed control is adapted as a function of the load, as well as Model Based Control schemes, such as the Feedback Linearization and linear quadratic reg- ulator described in [14], or Feedback Linearization combined with a PI
controller designed by means of a H,, loop-shaping robust method as in [15], or internal model control techniques for the design of a robust Proportional Integrative Derivative (PID) controller in a speed control application as in [16]. Predictive control approaches have been in- vestigated in [17], where the General Predictive Control and Con- strained Receding-Horizon Predictive Control have been exploited for the control of large steam turbines during load variations and in [18], where an Exponential Auto-Regressive eXogenous (ARX) model-based multistep predictive control algorithm is applied to control a thermal power plant. An interesting MPC application for the design of fault tolerant control systems is presented by Salahshoor et al. in [19], where a Generalized Predictive Control, combined with support vector ma- chine and adaptive neuro-fuzzy inference system, allows to recover the normal operation also in different fault scenarios. A complex control scheme, presented in [20], exploits a combined action of a feedforward controller based on nonlinear programming and a feedback controller designed with robust methods involving turbine governor valve and others control inputs, in order to control a fossil power plant in different load conditions. variations in the steam production, with the possibility to continuously operate by incorporating thermal energy storage or backup system in the plant, such as discussed by Zhang et al. in [8]. The continuity of energy production of a CSPP is also one of the main factors that may result effective in the thermal conversion of solar energy with respect to the photovoltaic, as highlighted by Desideri et al. in [9]. This approach can be coupled to non-standard control techniques, which allow to face the complexity of CSPP systems, by exploiting, for instance, a Model Predictive Control (MPC) approach, such as proposed by Vassallo and Bravo in [10], or adaptive control approaches and Artificial Intelligence (AD techniques, such as highlighted by Camacho et al. in [11]. The Steam Turbine (ST) has been originally designed for energy production from fossil fuels, thus both their mechanics and their control systems have been designed to face a quite stable steam production and very rare start-up and shut down cycles. As a consequence, in the context of CSPP, the standard PID control techniques currently applied to turbo-machinery are often not effective in the automatic adaptation to changing operating conditions of the machine [12]. Control system parameters are determined on the brand-new machine during the commissioning phase, with time consuming and effort-intensive pro- cedures. In addition, these parameters are only occasionally re-adjusted on the basis of semi-heuristic procedures. This forces the machine to work in non-optimal efficiency conditions during its lifetime. In the field of AI, Fuzzy Logic (FL) and Evolutionary Algorithm: (EA) have sometimes been exploited to control steam turbines. For in- stance, in [21] FL is used in order to adjust the parameters of a PID controller during speed control, while PID controllers whose para- meters are tuned through different EA methods are proposed in [22] and in [23]. However, none of the above-mentioned approaches defi- nitely proved to overcome the other ones in any operating scenarios. On the other hand, the control procedures need to allow correct and efficient machine operation in transient conditions without compro- mising its integrity, in particular focusing on variable operating con- ditions where steam turbines are subjected to continuous thermal stresses, premature aging and consequent lowering of efficiency. The turbine control problem with variable steam conditions can be ad- dressed by looking for applicability of innovative control strategy. Adaptive Control approaches, which allow to adapt gains in different loading conditions and uncertainties are known. An example of this kind of strategy is described in [13], where the turbine speed control is adapted as a function of the load, as well as Model Based Control schemes, such as the Feedback Linearization and linear quadratic reg- ulator described in [14], or Feedback Linearization combined with a PI
Fig. 1. CSPP schematic diagram of steam and turbine control systems. A real CSPP has been modeled focusing on the dynamic behaviors of the main component involved in the control loop of the generated electric power. In particular the CSPP is mainly composed by a steam generator and a steam re-heater, a non-condensing High Pressure (HP) ST, coupled with a gearbox to a condensing Low Pressure (LP) ST and an electric generator of rated power of 55 MW. The system is completed by two steam by-pass systems and a condensing system receiving the exhaust steam from the LP turbine outlet. An electro-hydraulic system controls the inlet valves of each turbine. The overall power control system is a cascaded PI control composed by an outer loop that follows a demand of power and gets a control signal demand from the inner loop. The inner loop controls the control oil pressure of the electro- The described system has been modeled through a quite complex model developed in the Matlab/Simulink environment shown in Fig. 4 and simplified in Fig. 2. In particular the modeling work has been
Fig. 1. CSPP schematic diagram of steam and turbine control systems. A real CSPP has been modeled focusing on the dynamic behaviors of the main component involved in the control loop of the generated electric power. In particular the CSPP is mainly composed by a steam generator and a steam re-heater, a non-condensing High Pressure (HP) ST, coupled with a gearbox to a condensing Low Pressure (LP) ST and an electric generator of rated power of 55 MW. The system is completed by two steam by-pass systems and a condensing system receiving the exhaust steam from the LP turbine outlet. An electro-hydraulic system controls the inlet valves of each turbine. The overall power control system is a cascaded PI control composed by an outer loop that follows a demand of power and gets a control signal demand from the inner loop. The inner loop controls the control oil pressure of the electro- The described system has been modeled through a quite complex model developed in the Matlab/Simulink environment shown in Fig. 4 and simplified in Fig. 2. In particular the modeling work has been
where the friction power losses Py,;. are computed as a function of Pratear the Power losses at rated synchronism speed w; and the rota- tional speed w as follows:
where the friction power losses Py,;. are computed as a function of Pratear the Power losses at rated synchronism speed w; and the rota- tional speed w as follows:
Fig. 4. CSPP Simulink model.
Fig. 4. CSPP Simulink model.
Fig. 5. Schematic of the FPID gains adapter. particularly in the case of complex nonlinear systems, starting from the studies described in [28,29], the foundations for the FL and the use of FL in the field of automatic controls. The term ”fuzzy” can lead to a misconception, but actually the FL is a structured framework allowing to formalise and represent imprecision and approximate reasoning, as pointed out in the article by Zadeh [30]. The main idea within FL is the possibility to describe complex systems with linguistic variables, that allow to make decision in a field similar to human reasoning and to convert a linguistic control strategy into an automatic controller, based on the knowledge acquired on the system, for example on its dynamic behavior. An example of the flexibility of FL is described in [31], where the author proposes a Fuzzy multi-criteria method (TOPSIS fuzzy) with the aim of investigating the use of a molten salt as heat transfer fluid in CSPP. One of the first turbo-machinery control approaches through the exploitation of FL is described in [32], where the pressure and speed error, the heat change and throttle change are described with linguistic qualitative variables and then used to control a steam engine and a boiler combination. The FL has been often juxtaposed with classical PID controllers, especially with the aim of achieving a trade-off between flexibility and simplicity, such as in [33], where the PID controller is obtained by means of a set of rules based on the error and its derivative. In [34] a fuzzy-based PID is designed by means of genetic algorithms which allow to optimally tune the FIS, whereas in [35] different fuzzy- based PID architectures are self-tuned by means of a set of rules within the FIS. In [36] a direct drive electro-hydraulic position servo system is presented, where a FL PID achieves automatic adjustment of PID
Fig. 5. Schematic of the FPID gains adapter. particularly in the case of complex nonlinear systems, starting from the studies described in [28,29], the foundations for the FL and the use of FL in the field of automatic controls. The term ”fuzzy” can lead to a misconception, but actually the FL is a structured framework allowing to formalise and represent imprecision and approximate reasoning, as pointed out in the article by Zadeh [30]. The main idea within FL is the possibility to describe complex systems with linguistic variables, that allow to make decision in a field similar to human reasoning and to convert a linguistic control strategy into an automatic controller, based on the knowledge acquired on the system, for example on its dynamic behavior. An example of the flexibility of FL is described in [31], where the author proposes a Fuzzy multi-criteria method (TOPSIS fuzzy) with the aim of investigating the use of a molten salt as heat transfer fluid in CSPP. One of the first turbo-machinery control approaches through the exploitation of FL is described in [32], where the pressure and speed error, the heat change and throttle change are described with linguistic qualitative variables and then used to control a steam engine and a boiler combination. The FL has been often juxtaposed with classical PID controllers, especially with the aim of achieving a trade-off between flexibility and simplicity, such as in [33], where the PID controller is obtained by means of a set of rules based on the error and its derivative. In [34] a fuzzy-based PID is designed by means of genetic algorithms which allow to optimally tune the FIS, whereas in [35] different fuzzy- based PID architectures are self-tuned by means of a set of rules within the FIS. In [36] a direct drive electro-hydraulic position servo system is presented, where a FL PID achieves automatic adjustment of PID
Input and output membership function parameters. Table 1 parameters.
Input and output membership function parameters. Table 1 parameters.
Fig. 6. Input (a) and output (b) membership functions. where SP(i) and PM(i) are, respectively, the Power Set Point and the Power Measure at time i. The outputs of FIS are the PID gains correction
Fig. 6. Input (a) and output (b) membership functions. where SP(i) and PM(i) are, respectively, the Power Set Point and the Power Measure at time i. The outputs of FIS are the PID gains correction
Performances of Relay PI, field PI and FPID applied in the trials. Table 3 6. Conclusions become adjustable parameters, and also to simplify some calculations with the aim of optimizing the computational time requested by the algorithm. The design phase of the control algorithm has been carried out in order to ensure that a site engineer could tune a set of para- meters:
Performances of Relay PI, field PI and FPID applied in the trials. Table 3 6. Conclusions become adjustable parameters, and also to simplify some calculations with the aim of optimizing the computational time requested by the algorithm. The design phase of the control algorithm has been carried out in order to ensure that a site engineer could tune a set of para- meters:
Fig. 9. Power Response with steam at 12% of nominal conditions. Comparison between PI Controller tuned by Relay method and Fuzzy PI controller performances (a). Compariso between PI Controller tuned with Field settings and Fuzzy PI controller performances (b).
Fig. 9. Power Response with steam at 12% of nominal conditions. Comparison between PI Controller tuned by Relay method and Fuzzy PI controller performances (a). Compariso between PI Controller tuned with Field settings and Fuzzy PI controller performances (b).
Related topics:
EngineeringEconomicsPid Controller
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved