The final array consisted of 16 TEPT4400 Ambient Light Sensors, and the outskirt sensor locations were modified to be at every 10° from 0° to 140°. From geometry, this array could detect the same number of cloud directions as the prototype array by assigning more sensor pairs among the outskirt ones.

Circuitry connections were modified as well. Instead of using a regular breadboard, a solder board was implemented to have firm/permanent connections. As shown in Figure 15, ribbon cable and associated pin connector heads were also used to make the circuitry clean and easy to follow.

            In addition, three 7805 +5V voltage regulators were added to the final design as shown in Figure 16.

When under direct sunlight, the sensors would draw approximately 2 amps of current total, which exceeded the current limit on Arduino board’s voltage regulator. Therefore, these three regulators were added to provide the required current consumption of the 16 sensors.  Overall, the system would consume approximately 2 – 3 amps to operate under direct sunlight, and 1 amp or less in an overcast time.

Final MCP Control Algorithm Design

Modified and developed from MATLAB simulation codes, the final MCP method was developed for the Arduino environment in order to provide the promising on-board processing power. The overall processing stages were summarized in the flow chart below:

Analysis of Performance

                The performance of the final design for the sensor array was judged mainly on the accuracy of cloud velocity measurements over a long period of time (at least one hour). Due to the inherent uncertainty of cloud motion, it was difficult to obtain a correct, quantified number for comparison. Therefore, the accuracy of the system was defined by comparing the velocity measurements of our device with the existing pyranometer device. 

Assumptions

Two important assumptions for the analysis:

1.      The cloud edge shape was simplified and assumed to be linear when reaching the sensor array.

2.      The cloud speed and direction were assumed to remain constant during the time taken to pass over the sensor array.

Other assumptions:

1.      The maximum cloud velocity was assumed to be 25 m/s.

2.      The direct sunlight was assumed to be in the viewing angle of all sensors at all times.

Analytical Methods Used

            The method used to analyze the data provided by the sensor array was the MCP method. Two versions of the MCP method were used to compare the results: one being the MATLAB code developed for the previous model, and the other one being the Arduino environment code.

Analytical Results

                When applying a known velocity of 0.25m/s in a westward direction and employing the MCP method to compute the correlation coefficients, the following resulted:

            As can be seen in the figure above, the most correlated pair relates to the westward direction of 270o, corresponding to an error of 0.00%. Next, using the times stored when each sensor was initially covered by the shadow, a velocity of 0.249m/s resulted, corresponding to a 0.85% error. Observing these low error values, it was concluded that the analytical method used was precise.

Description of Fabrication Process

                Due to the nature of this project, heavy machining/fabrication was not a main concern. Hence, Lasercamm and soldering were the two main processes applied in this project.

Sensor Array Fabrication

            The base plate that holds the ambient light sensors was made from 0.25 inch acrylic. The clear acrylic was intentionally chosen to minimize possible heat absorption, which could potentially cause expansion and cracking of the plate. In fact, the absorbed heat may cause the internal temperature of the enclosure to rise. The overall design was created in SolidWorks and later exported to AutoCAD. The printing of the array was done by the Lasercamm located in the UCSD design studio that belongs to the MAE department.

            Wiring and connections were finalized by soldering down to a solder board. Heat shrink was added to prevent the possibility of a short circuit and to reinforce the solder connections. In addition, referring back to Figure 15, ribbon cable and its associated connectors were used to provide a cleaner connection and enable a simpler interface to the Arduino.

Testing/Evaluation

Field Test

On a day that experienced intermittent cloud cover, both the pyranometer and phototransistor sensing systems were placed outside, facing west at 272o with clouds moving northwest at approximately 340o and were left to accumulate data for approximately 1.5 hours.

            Observing the data in the figures above, it can be seen that both react extremely similar to changes in light intensity in both magnitude and time of occurrence, proving the phototransistors to be an acceptable replacement for the pyranometers. In addition, something that the graphs don’t show is that over the same time recorded time span, the pyranometer sensing system completed a total of 941,376 samples, versus the 2,146,784 samples taken by the phototransistor sensing system. This proved that not only was the phototransistor system a good substitute as a light intensity sensing apparatus, but it also sampled 128% faster than the pyranometer system (Note: When not writing to the SD card it can sample up to 15 times faster).

Water Test

To ensure the waterproof capability of the NEMA 4X enclosure, a water test was performed. As shown in Figure , a shower was produced by a garden hose to simulate rain.

The test yielded positive results of the waterproof capability. Figure 20 was a photo taken after the simulated rain test. The interior of the NEMA enclosure remained dry and it was concluded that it would be waterproof based on the conditions it would encounter.

Cost Analysis/Projections

Data Logger/Processor Costs

            One of the more important parts of this project was finding a DAQ system that could provide both fast sampling and on-board processing. In general, a commercial data logger costs from several hundred to thousands of dollars depending on the number of channels and sampling rate. However, when examining the processing ability, Arduino Mega 2560 stood out as a cheap and qualified processor choice for the final design. The additional costs of the shields and modules that provided wireless transmission and SD card storage were also listed below in Table :

Table 7 - Circuitry Costs

The overall cost of the project came to $318.65, which was approximately 1% the cost of the existing design.

*Note: Battery and solar panels were not in the scope of current deliverables.

Impact on Society

If short term solar variability cannot be predicted or eliminated, it will ultimately hinder the integration of solar power. This is because variability affects the ramp rate (or the coming on and off line) of power plants. At low levels of solar energy generation fluctuations are negligible, and these variations are lost in the noise of the overall grid. However, at large levels of solar energy production, variability is very important. If the energy the utility is planning to receive from solar is not reliable its value drops tremendously, making its integration into the grid very difficult.  A system integrated with knowledge of cloud velocity and direction will have a huge impact on the capability of solar energy. Whether or not solar can successfully be tied into the current grid will ultimately shape how energy is used in the future.