Design, Control and Automation of MHPP - An Experimental Setup

In this paper the design, manufacturing and automation of a micro hydroelectric power plant (MHPP) prototype has been carried out. The experimental setup consists of three 1 kW synchronous generators (SGs) working in synchronization with each other and with the grid, three Pelton turbines with a single nozzle manufactured using a 3D printer, a water tank with a capacity of approximately one ton, a 5.5 kW centrifugal pump providing appropriate flow and head conditions and an 11 kW driver controlling the speed of this pump. The mechanical and electrical structure of the system and its working scenario are designed to be the closest to a real MHPP. S7-1200 PLC (Programmable Logic Controller) is used in order to control the voltage and frequency values of synchronous generators according to the load as well as for other control processes. It is possible to control and monitor the whole system through SCADA (Supervisory Control and Data Acquisition) screens. In this study, conventional PID control method and ANN (Artificial Neural Network) based PID are preferred for frequency and voltage control. The results have been evaluated by obtaining frequency-time, voltage-time, graphs of synchronous generators under different load conditions. Obtained results show that ANN-PID is faster than traditional PID and gives better results in terms of parameters such as settling time and overshoot.


INTRODUCTION
There are many equipment and intermediate stages in the production processes of HPPs. HPPs are complex and non-linear systems consisting of many mechanical and electrical-electronic equipment. There are many studies conducted on different aspects of HPPs. Since the aim of this study is to present the design and control of a prototype of a micro HPP, a summary of the work done especially on frequency and voltage control, has been given.
Micro Hydroelectric Power Plants with synchronous generators were stated to be advantageous in integration into the grid as compared to other renewable technologies. The PID-PI controllers were analyzed using the MATLAB / Simulink software package on the control of the speed regulator, which is one of the main components regulating the water flow of Micro Hydroelectric Power Plants. The synchronous generator voltage and rotor speed were compared using both control techniques under temporary and steady state conditions in their models, and they concluded that the PI control reached normal values faster than the PID control [1]. For a micro hydroelectric power plants, frequency control and voltage control were performed using Fuzzy Logic Controller and PI and PID controllers, respectively. The system was tested with active and reactive loads at different stages. As a result, the study also revealed the problems experienced in the correlation of experimental work with reality [2]. A model was created by predicting that using small powerful turbines with different powers instead of one large powerful turbine in MHPP would provide the highest benefit in variable water potential as a result of seasonal effects. In this model, the integration of the turbines into the grid was designed with a PLC using Fuzzy Logic algorithm. SCADA system was also used for remote monitoring and data recording of the system. In this model design, the system installed with three turbines of different powers were calculated to provide an increase in efficiency enough to compensate for the investment cost of the system with a single turbine that would occur in 5-5.5 years [3].
Although the parameter selection of PID controller remains challenging, it plays an important role in regulating quality and ensuring system stability. Using system decoupling in Jordan normal form with a state matrix, a state space model of hydraulic turbine regulating system (HTRS) with a surge tank was presented along with a formula derived for dimensionless rotating speed of turbine. The PID controller parameters were selected based on regulating quality parameters (RQP). Time-domain simulations as well as a comparative analysis of direct solving method (DSM) were performed based on a practical HPP. Same results were yielded by both methods; in providing a full RQP domain, DSM was found to be considerably efficient. The proposed DSM suggested the use of only state parameters in the selection of PID controller parameters. Considering varying conditions in the operation of hydro power plants, DSM was found to be effective in tuning purposes [4]. Another study presented a proportional integral derivative (PID) design for a hydro dominating energy system model, which was based on load frequency control (LFC) scheme. The designed controller was implemented in area-1 for at least 1% load disturbance and the values were compared with the values of classical PID. The performance of the controller was evaluated under various system conditions, thus, considering the non-linearity as well as the application results [5]. Based on fuzzy proportional derivative (PD) method, a new method was presented for LFC. The PD gains were tuned automatically using fuzzy rules to keep frequency at some nominal value under frequency variations. In order to control the frequency of mini hydropower system, the proposed controller was used. The performance of the system using the control scheme was evaluated by simulation performed under frequency variations. The results showed high performance frequency control when compared with conventional PD controller having fixed-gain [6]. In a paper, the design, modeling and experimental analysis of a hydropower plant with Automatic Generation Control (AGC) was presented. Adaptive Neuro Fuzzy Inference system (ANFIS) was used in the model. The aim was to reduce the frequency deviations occurring during power generation and to control the parameter selection for effective control of power in the HPP. Matlab was used in order to investigate the FLC, conventional PID and ANFIS controllers. The results showed that the ANFIS controller has a better performance rate as compared to conventional PID and FLC [7]. The study was conducted on non-linear model predictive controller (MPC) in order to control the variable speed hydropower (VSHP) plants. According to the simulation results, the proposed MPC was able to reduce water hammering experienced in the penstock pipes, damp the power oscillations found in the grid and improve turbine flow, power and head estimation. Therefore, VSHP may provide some considerable amount of FFR (fast frequency reserves) to the grid [8].
Many studies have been conducted on AVRs, which control the output voltage of synchronous generators. These are control methods performed together with various algorithms such as traditional PID implementations [9][10][11][12], cuckoo search algorithm [13], particle swarm optimization [14], coyote optimization algorithm [15], whale optimization algorithm [16,17], nonlinear robust coordinated [18], PSS and H∞ optimization [19] and kidney-inspired algorithm [20]. In addition to traditional PI and PID applications for the control of load frequencies of synchronous generators and power systems, control methods used with different algorithms, artificial intelligence and optimization techniques are also seen. Full order sliding mode [21], fuzzy proportional-integral [22,23], passivity-based control [24], stability boundary locus approach [25], ant colony optimization [26], bat algorithm [27] and fuzzy sliding mode controller [28] are also among the techniques used. There are such studies [29,30] in which traditional PID applications are performed and frequency control of other studies is reviewed [31,32].
In this study, micro HPP prototype experimental setup was designed and manufactured. Traditional PID and ANN based PID were used for voltage and frequency control. Voltage and frequency control was carried out simultaneously in the experimental environment.

GENERAL STRUCTURE OF THE MHPP PROTOTYPE
In this section, the general working principle of the closed-loop MHPP prototype, which can be operated in a laboratory environment with the working principle as close as possible to the working conditions and scenario of a real micro/mini/small HPP, and the equipment and their functions used for the manufacturing will be focused on.
In the prototype, the water absorbed from the water tank with a capacity of approximately one ton is brought to the proportional valves by using a horizontal shaft, four-stage pump driven by an asynchronous motor and a piping system.  The water pressure in the piping system is measured with an analog sensor. Centrifugal pump motor provides stabilization of water pressure in the system by means of frequency converter (driver) controlled by PLC. With proportional valves, the adjusted amount of water from the nozzle is hit to the Pelton turbines with an increased speed. The water loses its energy in the turbine and returns to the tank by free fall. The energy converted into rotational motion in the turbine generates electrical energy by being proportionally excited with the DC PWM clipper coupled to the turbine shaft with the SGs (Synchronous Generators). Frequency control of SGs is performed by controlling proportional valves with PLC. Integration of the electrical energy with the grid is performed with synchroscope devices. During all these processes, PLC monitors and controls the system with sensors and network analyzers. In addition, events can be monitored instantly with PLC-SCADA software. The block diagram of the system and general view of the experimental setup are given in Figures 1 and 2.

Frequency Control
The frequency control in the experimental setup is based on the control of the proportional valve connected to the analog outputs of the PLC, which processes the 0-10V signal received at the analog input comprising the analog frequency data from the network analyzers. Before synchronization, the frequency value of the generator is equalized to the network frequency value, i.e. 50 Hz. The flow chart of this cycle is given in Figure 3.
It is important to make the above-mentioned fine adjustment until the synchronization occurs and this loop is terminated once the synchronization process ends. Adjusting the proportional valve is essential for active power supply to the network.

Voltage Control
Adjustment of the voltage produced by the generator is performed by connecting a 0-5 V DC analog control input of the DC voltage clipper circuit placed for each generator at the power supply output, to the analog outputs of the PLC, and changing the excitation voltage. Before synchronization, the voltage value produced by the generator is equalized to the mains voltage.
In order to equalize the voltage generated before synchronization, processes shown in the flow diagram in Figure 4 are performed. It is important to make this fine adjustment until the synchronization occurs. This loop is terminated after the synchronization process.

PID Function
Adjustment of the frequency and voltage values of the SGs according to the load is provided by using the PID_Compact function in the PLC software. The PID algorithm works according to the equation given below: Where y is output value of the PID algorithm, Kp is proportional gain, s is Laplace operator, b is proportional action weighting, w is set point, x is process value, TI is integral action time, TD is derivative action time, a is derivative delay coefficient, c is derivative action weighting [33].
PID block diagrams related to frequency (rpm) and voltage set points in the system are shown in Figure 5 and Figure 6, respectively.

ANN-PID CONTROLLER
In this chapter, an ANN-based self-tuning (adaptive) PID controller is presented for frequency and voltage control. PID gains are tuned online using NN based algorithm. The robustness of control system is carried out by continuously tuning of PID parameters (Kp, Ki and Kd). PID controller tuning depends on the adjustment of its gains. As well known, PID controller equation in time domain can be expressed in equation 2.
e(t)= r(t)-y(t) (2) where e(t) is error value, r(t) is set point, y(t) is output value, u(t) is control variable.
A block diagram of self-tuned PID control with artificial neural network is shown in Fig. 9. Error, reference and output are fed into PID controller. ANN-based self-tuning algorithm adjust Kp, Ki and Kd values as system parameters change using ANN based PID, the PID gains are constantly updated. From the ANNs established with varying architectures, the artificial neural network architecture that provided the best results and that had two hidden layers with 8 neurons in the first layer and 4 neurons in the second layer was used. In total, 1325 data were used for training and 255 data for the test. In Figures 10  and 11, architecture of the designed ANN and view in Matlab, respectively, are given.  In Section 5, characteristic graphs of SGs drawn with data from the field have been given.

EXPERIMENT RESULTS
The data obtained in the experiments performed with the established experimental setup are graphed. As a result of the experiments, the highest active power produced by the generators is measured to be 400 W.
Although the line voltage value between phase-neutral is shown as 220 V in flow diagrams, ladder and SCADA screen displays, the set value of the voltage (L-N) is determined according to the network values at that instant since the network voltage value changes (by taking into account the synchronization with the grid) at the time of the experiments. Therefore, the test results and graphics were calculated and drawn accordingly.
The results of the transition from no load to 180 W load (0-45%) are given in Figures 12 and 13. In the experiment using conventional PID, the lowest voltage value was 228 V (overshoot is 7 V) and the lowest frequency value at the rate of 48.3 Hz (overshoot is 1.7 Hz). Voltage and frequency values were made to reach the set values in 18 seconds by the PID controller. In the experiment with ANN-PID, the lowest voltage value was 232 V (overshoot is 3 V), and the lowest frequency value at the rate of 49.6 Hz (overshoot is 0.4 Hz). Voltage value was made to reach the set values in 9 seconds, while frequency value in 6 seconds with the ANN-PID controller.
The test loads are switched on at t=0 point and in the pulse transition form (not linear increase).     The results obtained from traditional PID and ANN-PID in the experiments are given in Tables 1 and 2 for voltage and frequency control respectively.

CONCLUSIONS
In this article, a micro-level prototype of a HPP with three SG and Pelton turbines was established. The design, installation, control and automation of a non-linear system consisting of many mechanical and electrical equipment and subsystems were performed. Since three turbines could be operated at the same time with the experimental setup, different control modes were created for SGs such as synchronization of more than one SG among themselves and with the interconnected grid, as in a real MHPP. Automation of the system was also conducted with PLC and SCADA software. The voltage and frequency control of the system was performed under different loads. Conventional PID and ANN-PID control mode were selected for control. In normal operation, while the voltage and frequency are made to reach the set value simultaneously by the PIDs, the indirect or direct effect of these two parameters causes the voltage to reach the set value later than in the individual operation. Generally, only the frequency or voltage of the SGs is controlled in the papers. Theoretically, it is assumed that frequency and voltage do not affect each other, but in practice, when evaluated together with other parameters, they do. In this study, the voltage and frequency of a real system (MHPP) built as a prototype were controlled simultaneously under changing loads. It has been observed that ANN-PID gives much better results than traditional PID in terms of parameters such as settling time and overshoot. The results of the present study showed that ANN is an effective and reliable method for voltage and frequency control of SGs, which is a nonlinear and complex system. Although the settling time seems to be long in the experiments, it can be understood that these times are considerably shorter when the acceptable tolerances for voltage and frequency are taken into account.
In future studies, the results can be compared using different artificial intelligence techniques and algorithms for voltage and frequency control. The ANN algorithm can be further improved.