Patent Application
20100198420
This disclosure relates to the operation of power systems using solar energy, such as solar farms using photovoltaic or solar-thermal technology, as well as other weather-dependent energy sources. In particular, it concerns applications using measurements to predict meteorological conditions in order to estimate power production and control power generation.
Utilities want and need predictable, stable power generation. End-use devices function best with a steady flow of electricity. The components of the grid system (wires, transformers, etc.) are most reliable when the flow of power is constant, or at least varies slowly and predictably.
Grid operation requires that the supply and demand of electricity be matched at all hours. During normal operation, utilities use power plants in ‘regulation’ mode to match moment-to-moment changes in load and intermittent power production and ‘load following’ to match changes in power as the power demand goes through normal daily load fluctuations. Under contingency operations (for example, when a power plant or transmission line is unexpectedly out-of-service), additional “spinning reserve” and “non-spinning reserve” resources are engaged to maintain grid reliability. Fossil fuel power plants, hydropower plants, power storage facilities, and customer load reductions all provide these services to the grid.
Different power plants have different operating ranges, time response periods and cost-to-respond profiles, and have different roles within grid operations. Some can be started very quickly, such as hydroelectric power plants. Others take longer to ramp up to full production, such as natural-gas-powered turbines. Still others take even longer to increase power production, such as coal-fired plants. If a major power production facility goes off-line or significantly reduces its electricity production, or if load demand increases more significantly than expected, the utilities must respond by starting up an alternative source quickly enough to prevent loss of power for end users.
Intermittent resources significantly impact the electrical grid because fluctuations in power from intermittent resources such as solar and wind occur during normal operation. The utility industry is beginning to deploy large-scale solar farms, producing 10 MW or more power from a single geographic location. Solar power systems produce electrical power as a function of the amount of light, referred to as insolation or irradiance, incident on the component solar panels. The irradiance affects various factors in the generation of electricity from a solar power system. If the solar power system provides a significant fraction of power to a grid operating area or section of the grid, changes in irradiance can have a significant impact on the stability of the power on the grid. The power production from solar farms is very predictable during fair weather because the ramp-up at sunrise and the ramp-down at sunset can be predicted from almanac data, and is suitably gradual that backup sources can be phased in and out at a reasonable rate. Also, on dense overcast days, the power production can drop down to 10% of the clear sky power production, although, the variation in solar production over short durations is not as significant.
Non-grid-connected systems that deliver power directly to end users also need to be stable because electrical equipment can malfunction or even be damaged by a fluctuating supply of power.
Technology exists for tracking major storms, and the corresponding effect on power output from a solar farm in the storm path should be easy to predict; however, the effects of other weather conditions, such as the passage of broken clouds, are presently very difficult to predict or compensate. On partly cloudy days, a solar photovoltaic farm will alternate between full production and 10% power production with ramp times down to seconds or minutes. These ramp times are too short for most common grid backup sources to be brought online from a “cold” start. Alternately, with a smaller control area, there may not be enough regulation or quick changing generation to compensate. Solar thermal systems have an inherent thermal inertia that causes them to react more slowly to irradiance changes than solar photovoltaic systems; however, intermittent shading by patchy clouds can cause unacceptable instabilities in power output for these systems too.
As indicated above, forecasting is used to increase predictability in power fluctuation. Given the forecasts of load demand and intermittent production, the grid operator may start additional power plants and operate the same or other power plants at less than full capacity such that there is sufficient system flexibility to respond to fluctuations. As a result, some grid operators are currently using centralized weather forecasting to predict power output of both wind and solar facilities. In addition, if forecasting information was readily available to operators of non-grid-connected solar and wind systems, the operators could also timely activate backup sources to prevent power interruptions, or reschedule their use of sensitive equipment.
Localized differences in wind speed due to different ground levels or obstructions will affect ambient and solar panel temperature. With changes in temperature, the output power from solar panels will change even if the irradiance does not change. Thus, local landscape features can cause different panels or arrays to produce differing power outputs at any given time.
Even if the terrain is perfectly featureless, as in some plains regions, broken or moving cloud patterns can affect the power outputs and operating factors, such as Maximum Power Point (MPP) of the PV panels below. The more area that a solar farm installation covers, the more opportunities for shifting cloud patterns or fog patches to decrease the power production in a part of a solar farm. Therefore, even with several sensors of sunlight intensity distributed across the area of the solar farm, it is difficult to accurately predict the total power that will be produced by the solar farm in the next few minutes.
A cloud passing over a part of a solar farm can quickly reduce the power generation of that part from maximum to less than 10% of maximum. A transmission grid may be limited in the amount of intermittent power generation that can be interconnected and operated using current technology while maintaining North American Electric Reliability Council (NERC) reliability requirements. Also, in accordance with NERC reliability requirements, each transmission grid operating area (balancing area) is required to identify any power exchange with other balancing areas in advance and then operate their system to strictly adhere to those schedules. Based on the power plants storage and load response available within a balancing area, the transmission grid has a limit to the amount of load fluctuation and intermittency the system can respond to and still meet reliability requirements. Depending on these factors, sometimes a new solar farm with its natural fluctuations can be accommodated with or without forecasting and advanced utility actions. In other cases, due to these factors and other intermittent generation effects, new solar farms' natural fluctuations cannot be accommodated by the existing grid.
Depicted in
Since there is movement of the clouds, it is possible to predict the future positions of these clouds based on their current movement. Thus, if the clouds are moving to the right in the image, corresponding changes in solar irradiation can be expected. Similarly, there are circumstances in which the density of clouds will change over a time period represented by the movement. These changes can be fairly predictable, based on current meteorological conditions and historical meteorological data. Examples of meteorological conditions include effects of wind and wind direction in areas near mountain ridges, stability of the air (a function of the environmental lapse rate), and time of day. Many of these meteorological conditions interact; for example, an upslope wind in warm unstable air in the afternoon is likely to result in rapid cloud formation. As another example, the towering cumulus cloud 149 has no present effect on the photovoltaic network in
Utility regulation significantly affects the production of energy from intermittent resources such as solar power. Based on utility and regulatory methods, there may also be some value assigned for how reliably it can provide power when power is needed (capacity value). A dispatchable (controllable) power plant can be under contract to provide both energy and power at the same time. For example, a 100 MW turbine might produce 80 MW of power delivered to the grid to supply energy and then have an additional 20 MW which can be used to provide ancillary services. For example, with frequency regulation, the grid operator sends signals on an ongoing basis, identifying what power level between 80 MW and 100 MW to operate the turbine until the next signal is received. In spinning reserve or non-spinning reserve, the 20 MW is held in reserve and can be provided to the grid on very short notice.
As part of the agreement to connect a power plant to the transmission grid, a power plant may be required to limit changes in the total electricity production at the point of interconnecting to the grid (production ramp). Therefore, a need exists for means of stabilizing the output from intermittent sources such as solar farms.
SENER Ingeniería y Sistemas S.A., of Getxo, Spain, provides a software package called SENSOL that uses historical weather data to predict overall solar farm performance. It does not, however, address the need for solar farms to anticipate and react to rapidly changing conditions in real time. A tool that would perform this function for solar farms and the utilities they serve would be a valuable contribution to the commercialization of large-scale solar power plants at a cost that could effectively compete with less-sustainable power sources.
Sandia National Laboratories has published a program concept on “Solar Energy Grid Integration Systems” (SEGIS). With SEGIS, a utility could connect and disconnect solar sources and storage sources from its grid and control customers' loads according to, among other things, weather “trends and forecasts” from the Internet. This approach has some drawbacks (even supposing that all customers want their power rationed by the utility, in proportions changing daily with the weather, and with less power available in colder weather). The trends and forecasts available on the Internet are typically data reductions geared toward human activities such as travel, recreation, or agriculture. Therefore, those specific weather-related factors that affect power-station output may not be separable from the mix of data. Furthermore, the trends and forecasts tend to be averaged over a large area and a fairly long time (8-24 hours), whereas a power station may be subject to a local microclimate and the fluctuations that most need to be addressed take place on the scale of several minutes.
Planes and boats have on-board weather radar to detect local variations in weather patterns. Also, local weather stations have been deployed for specialized applications such as detecting wind shear and microbursts near airports. These forms of local weather detection have not been applied to use in the prediction of distributed power systems, and have not been used to predict the effects of cloud cover on solar farms or wind farms. More generally, previous systems have not employed local weather detection and local cloud prediction to predict power production dynamics on a moment-by-moment basis.
Solar farms and wind farms have set up local monitoring stations for resource assessment and performance supervision, but this monitoring had not included local weather detection and local cloud prediction to predict power production in real time. This type of real-time data gathering and power prediction would be valuable, especially if linked to a system that enabled corrective action to stabilize farm power output or grid power in the event of an unacceptable degree of expected fluctuation.
Control is implemented in a power generation system and/or storage system associated with the power generation system, in which the power generation system generates power from least one weather-sensitive power source (that is, a source having a power output that may be affected by weather-related conditions). Information on weather-related factors that can affect the power output of the power generating system is obtained from measurements, from external sources, from stored data, or from a combination thereof. From this information, the control system calculates at least one of the amplitude, rate of change, onset, and duration of an expected change in power output from the power generating system. The control system compares the calculated expectations with predetermined thresholds representing acceptable changes in power output. Based on the results of the comparison, the control system selects and executes a response. For example, if the expected power-output change does not exceed any of the predetermined thresholds, the control system continues monitoring weather-related factors. It should be noted that, if the calculated expectations exceed a predetermined threshold, the control system may adjust the power output from the power generating system so that the actual output changes do not exceed the predetermined threshold.
As described herein, “solar farm” is intended to include a variety of configurations of a power generating station, including arrangements of photovoltaic, wind, or solar-thermal generators on open land, on diverse structures, such as buildings, other types of solar collectors, and combinations of these.
Atmospheric-condition tracking for predicting short term effects of clouds and other weather-related factors on a power station or other defined area includes a number of features relevant to observations:
Direct sensing of output power fluctuations in localized parts of a geographically extended power system (for instance, fluctuations in the output power of individual panels, sub-arrays or arrays) is another source of data that can be used to predict what will occur in other parts of the system. For instance, when clouds pass over an array, they cause output power fluctuations as they shade each panel. If the power delivered by individual panels or sub-arrays at known locations is tracked over time, the speed and direction of the transient shading fluctuations can be calculated. Once the speed and direction is known, further calculations can predict which other arrays will be similarly affected, and how soon.
Weather-Related Variations in Power Station Output
Calculations of Impact of Conditions
Calculations or impact of conditions used to predict localized irradiance changes due to cloud passage may also take into account almanac information relating to location, time of day and time of year. From the almanac information, the position and angle of the sun with respect to the solar panels during the passage of the clouds may be calculated. In addition, a prediction of cloud coverage and cloud movement augments the predicted irradiation by way of calculations of an expected amplitude, onset, rate of change, or duration of reduced power generation due to effects of atmospheric conditions, or any combination thereof. The effects of atmospheric conditions include shading by clouds and increased power generation due to dissipation of clouds. Calculation of the effects may include calculation of one or more of:
Clouding can also change the ambient temperature around solar panels, which can affect their efficiency and their maximum power operating point (MPP). In addition to the direct effects on photovoltaic solar cell performance, the effects of clouding can be used as a predictor for further clouding. By way of example, temperature change prediction can add another level of accuracy to the power prediction.
External Sources of Weather-Related Information
In one configuration, local cloud conditions are tracked by one or more weather tracking stations, such as weather tracking stations 171, 172, 173 depicted in
Atmospheric-condition tracking includes a combination of external report inputs, a modification of the reports in accordance with local conditions, and inputs from measured observations. Measured observations may be obtained directly from the output of a solar system or from external automated weather stations, typically provided by third party sources. Examples of external automated weather stations include Automated Weather Observing System (AWOS), Automated Surface Observing System (ASOS), and Automated Weather Sensor System (AWSS). Other external sources of weather information are widely available forecasts and measurements of meteorological data, including ultraviolet (UV) forecasts, external UV measurements, external insolation measurements, temperature/dew point spread, regional wind measurements, stability of the air (environmental lapse rate), nearby solar farm measurements and private weather station data. Externally obtained weather data which could come from a third party or from the solar farm operator controlled off-site monitoring station could also include regional wind measurements, visibility and obscurations, precipitation, cloud cover and ceiling, and temperature/dew point information. While most of these measurements can be used to predict cloud cover, the cloud cover and ceiling information is particularly useful because of the direct effect of clouds on irradiation.
Cloud cover and ceiling measurements can be made with a ceilometer. This instrument is one application of a LIDAR (Light Detection and Ranging) system, where the laser beam is pointed upward at the zenith and detects the amount and height of clouds. In addition, non-LIDAR ceilometer instruments such as optical drum ceilometers can be used to obtain this information. In a typical weather observation station, cloud density measurements are obtained by calculating an average or a weighted average of readings over time. One aspect of ceilometer measurements is that the ceilometer is able to (as the name suggests) take ceiling measurements, meaning the bottoms of cloud layers, but is less adept at determining densities of cloud layers. In its raw form, a ceilometer measurement provides no indication of cloud tops, thicknesses or densities, except in cases where thin layers create ambiguities in the actual measurements. The different types of responses obtained from LIDAR or ceilometer measurements and from radar measurements can be used to provide further information regarding cloud density and other cloud conditions.
In addition to vertical measurements, it is possible to obtain measurements at multiple angles. A more general scanning LIDAR can provide three-dimensional images within a cone of the sky, and a doppler LIDAR can measure cloud movements. The use of angled measurements, in combination with wind measurements, makes it possible to obtain information regarding the change of cloud coverage. It is possible to measure the movement of the edges of the clouds, which provides an indication of cloud movement, and this can be accomplished with vertical and/or angular measurements. In addition to observed changes in cloud movement, it is possible to use surface wind measurements, with an appropriate correction counterclockwise from the surface (northern hemisphere), to predict winds aloft.
Similar weather monitoring techniques can be used for other types of power systems that are weather-dependent. By operating (or connecting to) sources of data on selected relevant weather-related variables, those specific weather factors that affect power-station output can be isolated to streamline the analysis.
Measurement of Conditions
On-site forecasting may be achieved either from a single location or from multiple locations with distributed weather sensors, such as a network of monitoring systems throughout the solar farm site, around the periphery of the solar farm site or from off-site locations near the power generating system. It is also possible to provide weather sensors at off-site locations a significant distance away from the power generating system. Distributed sensors, including on-site sensors, can be more effective than central forecasting for predicting a solar farm's power production relative to the grid operator's forecasts because they can provide multiple checkpoints to take account of the effects of local terrain, local wind conditions, observed convective activity, etc. “Distributed sensors” can refer to any group of sensors at more than one location. The distributed sensors may be at various locations within the general perimeter of the solar farm, or can be off-site, provided that the sensor data is relevant to cloud prediction. This is advantageous in locations where multiple geographical features that interrupt wind currents or change the humidity content, such as mountains, rivers, bays, lakes, and volcanic-type features such as hot springs, create varied local microclimates. In such areas, cloud patterns can change significantly in the course of crossing from one microclimate to another.
Localized sensing can include sensing of local irradiance (e.g. using pyranometers), or sensing of the power output of reference photovoltaic cells as changes in ambient conditions affect them, as well as sensing atmospheric conditions with instruments such as radar, LIDAR, visual sensors such as cameras, and thermal sensors. The locally sensed data may be combined with data from other sources, such as ground and satellite-based weather stations.
Radar sensing allows detection of visible meteorological conditions. The solar farm may be provided with a small weather radar station, similar to those used on ships. With clouds typically at an elevation of 1500 meters, the weather station would be able to detect the presence of clouds for a 140 km radius around the local station associated with a solar farm. The radar may be capable of detecting the location of the clouds, their speed, and the direction of their movement. A simple computer program combining the radar data with the sun's calculated position (based on the date, the time, and the farm's location) could predict the upcoming shading of the solar farm and project the power production vs. time, and provide projections out into the future. The predictions would be directed to both the change in power production and the rate of change.
LIDAR sensing is best for identifying aerosols and visible clouds. Radar is better than LIDAR for identifying precipitation and, depending on the power and topography, may have a greater range. LIDAR is similar to radar, in that it computes the distance to the target by emitting electromagnetic pulses or continuous waveforms and measuring the time delay between the transmission and arrival of the reflected signal. In contrast with radar, the wavelength of the light emitted by the LIDAR system's laser is on the same order of the cloud particle diameters that scatter sunlight as described by the Mie theory (or Mie solution to Mie scattering). Thus clouds are very opaque to LIDAR, and provide a good back-scattered signal. This is in contrast to the much longer-wavelength radar which is good at detecting metallic objects and relatively large precipitating water droplets; however, clouds themselves are relatively transparent targets. Also, the laser's extremely narrow beam and low divergence affords superior spatial resolution not achievable with radar. Radar, on the other hand, can provide additional information relating to cloud density or thickness.
Cameras or other visual sensors can provide additional detail for forecasting changes in power output on the scale of minutes. For example, they can be used to distinguish thin from thick clouds and also to measure the spectrum of the light. Similar computer programs would be used to interpret the data received from the cameras and other visual sensors. This can be accomplished, for example, by densitometry measurements of digital images of clouds; the densitometry values are indicative of relative cloud opacity. Additionally, frame-to-frame motion and shape changes can be tracked to obtain speed, direction, and probability of convective action. Pattern recognition programs can be used to distinguish between cloud types, e.g., cirrus from cumulonimbus. A more complex program can combine locally gathered data with data from other sources to predict more than the change in brightness of the incoming sunlight. By way of non-limiting example, changes in the relative strength of the direct and diffuse components of sunlight, and changes in the spectral profile—both of which affect solar cell efficiency and hence solar farm output—can be tracked.
Clouds can also be detected and tracked by monitoring the actual power output of parts of the solar farm. When a cloud edge passes over a solar panel, the power output from that panel or string of panels will change. By measuring the output of the photovoltaic panels, a direct reading of irradiation is obtained. The movement of cloud cover is detected as the power output of sequential panels changes. When the power from one panel, sub-array, or array changes, that information can alert a human operator or automated control system that the power from other panels, sub-arrays, or arrays are about to experience similar changes. By consulting a look-up table of affected panels' geographical location, the speed and direction of the power changes can be mapped, and their future trajectory and arrival time at other locations predicted. Similarly, if two or more solar farms share the data, a farm that is presently undergoing power changes can alert another farm if the changes are moving in its direction so that the other farm can prepare to react to the changes.
Historical Information
Local regions, including the locations of power stations, sometimes have special characteristics that affect incoming changes in cloud cover and other weather-related effects. Because these special characteristics are largely related to geographic features such as land contours, bodies of water, and geothermal zones, in most cases their effects on incoming weather patterns are repeatable, or at least capable of being extrapolated from previous trends. Enhanced embodiments of the present subject matter take advantage of stored historical data related to special characteristics of the power station's location to enhance the accuracy of predictions. Almanacs provide historical data on an area's high, low, and average temperatures as well as other statistics. These statistics can be factored in to calculate the likelihood that a predicted change in atmospheric conditions is accurate.
Advanced embodiments of the present subject matter can “learn” from the data they have gathered in the past. Both predictions and actual results can be stored in archives, and correction factors to enhance the accuracy of future predictions can be calculated and updated. For example referring to
Predicting the Impact of Conditions on Power Station Output
With the various measurements of the local weather, the power output of a solar farm or a local collection of solar panels mounted on rooftops, in parking lots or on the ground can be predicted. For example, the weather radar can track the movements of distant clouds that may be approaching the local region and measure their speed and direction. Given the current speed, direction and distance, and knowing the position of the sun versus time of day, a computer program can calculate when the clouds will shade the solar panels. The weather radar can also estimate the density of the clouds and how much they will reduce the solar irradiance on the panels. As the leading edge of the cloud shadow moves onto one or more solar panels, it will decrease the power output from that solar panel; the decrease in power provides a direct measurement of the reduction in irradiance caused by the cloud's arrival.
Clouds can shade solar panels by moving across the sky to a location where their shadows fall upon the solar panels. However, clouds can also form directly above the solar panels from a previously clear sky. As they form, they might or might not also move. By adding data from external weather forecasts to the predictive calculations, the program(s) can predict whether the clouds are likely to increase or decrease in size, density or other properties that affect their shading properties as they form or move. For example, by combining a national weather-service prediction of summer afternoon thundershowers and local observation of new clouds forming, the program(s) could more accurately predict whether and when the cloud cover is likely to affect solar power production and by how much.
Typically, clouds move at a speed of less than 70 km/h. Therefore, with cloud tracking radar with a range of 140 km, in typical situations, the computer program would be able to predict power production for the next two hours. On a very, very windy day, with steady winds aloft of 140 km/h (equivalent to a Category 1 hurricane), it would still be possible to predict 1 hour into the future, which indicates that predictions are likely to exceed 1 hour in any weather conditions in which predictions would have value. Therefore, the solar farm or utility would have at least 1 hour, and often have 2 hours or more, of notice to prepare an alternative power generation source.
The output of the solar panels decreases with increasing temperature of the solar panel. The temperature of the solar panel is determined by the ambient temperature, the intensity of the sunlight hitting the solar panel and the amount of cooling of the solar panel by wind. More irradiance would generally result in more power output; however, if the increased irradiance heats the panel significantly, the power increase may be attenuated. For a given type of solar panel and a given mounting angle, its temperature can be determined experimentally as a function of ambient temperature, wind speed, wind incidence angle and sunlight intensity and then stored in a database in a computer system. Then, by measuring the local wind speed, the local wind direction, and the local intensity of sunlight affecting an installed group of solar panels, the computer can calculate the likely temperature of the solar panels from the stored information in its database and predict the likely level of the panels' power output. The accuracy of the prediction can be improved with the use of historical information and by comparing predicted to actual results from other solar farms.
Responding to Predictions of Power Output Changes
A control system for the power generating system is able to respond to predicted changes in power output that exceed threshold criteria imposed by the grid or load to which the solar farm supplies energy. Predetermined threshold criteria for acceptable amplitudes and rates of change (“ramping rates”) can be stored in the control system. These criteria can be based on customer specifications, limitations of the connected grid, production targets of scheduled power and intermittent energy, regulatory requirements, or power connectivity standards promulgated by organizations such as IEEE. They can also vary with measured or expected demand for the time of year and time of day of the expected onset and duration of the change; for example, a grid may be able to tolerate a larger amplitude of power decrease from the power generating station during non-peak demand hours than during peak demand hours.
Each calculation of an expected change can be compared to the relevant threshold criteria to produce a comparison result. If the comparison result does not exceed the threshold criteria, the power station can continue to monitor weather-related factors and take no further action. If the comparison result exceeds the threshold criteria, the power station can select and execute an appropriate response based on the characteristics of the comparison result.
The power generating system's operator (human or machine) may take various steps to mitigate the total fluctuation seen by the grid operator. This could include 1) proactively reducing power output at an acceptable rate before a downward trend begins, 2) controlling the upward ramp rate of power output, by limiting change of power when an upward trend begins, 3) using stored energy to limit the rate of increases or decreases in power output, 4) using backup generation resources which could produce energy to reduce the combined rate of change to target levels, 5) reducing the demand of a large energy consumer within the transmission and distribution area, to reduce the combined rate of change to target levels, and/or 6) communicating anticipated changes to a power utility or grid operator. In the case of energy storage facilities, in instances where wide fluctuations are expected during a particular time period, energy storage can be particularly useful because this allows the distributed power network to be operated at near maximum available power at any given time while providing a less dynamic or more stable output to the power grid. Actively managing the power could have various goals including: 1) reducing the change in power to within contracted or acceptable limits, 2) matching a contracted output profile, which could be in various increments including 10 minute, 15 minute, or 60 minute increments, or 3) matching a non-contracted pre-promised profile.
Control of Production from Solar Output
Proactively reducing the power output from the solar facility is implemented to prevent the facility from exceeding the grid's production ramp-rate requirements as clouds cover the facility. For example, in a photovoltaic farm, a control system that controls the inverters via a network can issue commands responsive to the gathered data. If a power plant has many inverters, the ramp rate of each inverter could be limited, or selected inverters could cease to deliver power (by being turned off, disconnected, or having their power diverted to storage or other backup sources) in stages while others continue to operate at maximum available power output. For the second case, the total power sent to the grid could be gradually ramped in either direction by sequentially switching inverters on or off. Those skilled in the art will recognize that “turning an inverter on or off” embraces other modes of activation and deactivation, such as exiting or entering a standby mode. It could also be the case that one inverter is dropping power while another inverter is increasing in power. In this case a central control system could monitor all inverters to maximize the total energy produced while still meeting maximum farm level ramp rates.
For a large farm with many inverters, a suitably gradual ramp may be achievable by selectively turning inverters on or off rather than operating them over a range of intermediate power levels to achieve a ramp in power level. Turning inverters on and off results in a ramp with “steps,” rather than a smoothly varying ramp, in the total output power from the farm. If the farm has many inverters, the step height from turning any single inverter on or off will be only a small fraction of the total output power, so the non-smoothness of the ramp may be insignificant to grid stability. On the other hand, a farm with fewer inverters operating at higher power may produce a ramp with unacceptably large steps by turning inverters on and off. In that situation, the control system could change the operating point of one or more selected inverters (toward or away from the MPP) to produce the desired power level and ramp rate. Other factors to be considered when choosing between the “on/off” and “intermediate power level” embodiments are the optimum operating ranges of the inverters and the capabilities of the inverter control system.
Control of the upward ramp rate of power output as clear sections of sky replace the clouds over the solar facility is implemented because power grids do not perform well when the output of a power source changes rapidly in either direction. Possible approaches for controlling the upward rate of power output whenever an unacceptably steep upward trend begins are:
The control of inverters in the departure of clouds and increase in power can be done in manners similar to the control of inverters in the anticipation of the arrival of clouds and reduction in power. In this case, a central control system, such as a Supervisory Control And Data Acquisition (SCADA) system, could achieve the ramp at the farm level either by turning inverters from off to on or from on to off or by operating inverters at less than their maximum available power output.
Alternatively, in a system with “smart” inverters, whether or not a control system is included in the inverter network, each inverter could be programmed to ensure that it does exceed certain power ramp requirements. If the DC input power entering the inverter begins increasing at too fast a rate, the inverter can operate off the maximum power point (MPP), for example by increasing voltage and decreasing current, thereby reducing its immediate AC power output to the grid. As the incoming DC power level stabilizes, the inverter can gradually return to MPP operation and optimum efficiency at a grid-compatible ramp rate.
Deploying Backup Power Sources to Compensate for Weather-Induced Power Fluctuations
Onsite or remote power storage may be used to mitigate fluctuations in power output to meet production ramp interconnection requirements. One version of this strategy uses centrally controlled, physically distributed storage. Each array routinely stores power in a battery, flywheel, or other energy storage system. The control system is able to monitor and optimally utilize the stored power in all the energy storage systems, as well as monitoring the solar farm output, from a single control point. Such a system may use sensors and algorithms in order to smooth out rapid fluctuations from cloud passage. The system may also be configured to connect the storage systems to the control system through the network used to control the inverters.
Alternate Generation
There may be entities within the transmission and distribution grid area that have backup generation capacity. If these entities produced energy as the output from the solar farm declined, the combined rate of change of the solar farm and the backup generation could be within target levels.
Control of Load Demand
There may be large energy consumers within the transmission and distribution grid area that could reduce their demand for power as the output from the solar farm declines. The combined rate of change of the solar farm and the large energy consumer could then be kept within target levels. In this manner, communication with users may include communicating with a non-utility partner who would then be able to adjust generation or load demand on the grid.
When the power ramps upward, as when clouds over a solar farm dissipate, some loads could be increased to slow the increase in supply to the grid. For instance, a greater fraction of produced power could be diverted to storage when the upward ramp is too steep.
Notifying Operators of Upcoming Power Fluctuations
Sending communications that notify a utility or grid system operator of upcoming output fluctuations from a solar farm or other power station allows the utility or grid operator to operate flexibly to mitigate the effects of the expected fluctuations on grid stability. This could be achieved by multiple methods including:
This communication function could also be used for sharing data between the solar farm and the utility, grid system operator, other solar or wind farms, or other local weather monitoring stations. A central control system could gather data for, and react on behalf of, multiple farms spread across a geographic area similarly affected by weather patterns.
Combined Controls
A combination of solar forecasting, inverter controls, storage or other responses, may be used. The computer program would be used to regulate the power to a certain level using a combination of the various techniques above.
Prediction-Enhanced Supply Flexibility from Weather-Affected Power Stations
Solar and other intermittent resources currently sell energy. As part of a power purchase agreement, any capacity or power value may be sold together with the energy. By combining storage and power control algorithms in a solar power plant, such a plant could function as both intermittent (non-dispatchable) and a dispatchable resource. Software is able to track the energy from the intermittent solar facility separately from the power or regulation services provided by the storage facility. Thus, the operator of a solar farm is able to conveniently sell both the intermittent energy (with its associated capacity value) and the dispatchable regulated power through the same interconnection point. The control system (with or without storage) could also provide ancillary services to the grid, including but not limited to voltage regulation, frequency regulation, power factor correction, load following, and spinning and non-spinning reserve.
Further Description of Diagrams
The input information is used to predict cloud effects on irradiation, as indicated at block 326.
The output of the prediction (block 326) may be used for several types of response, depending on the particular configuration of the system. Examples of responses include reducing solar farm output to smooth output fluctuations (block 331), using stored energy (block 333), alternate generation from backup sources (block 335), adjust load response to compensate for output fluctuations (block 337) and notification of the utility or grid of an anticipated fluctuation (block 339).
Operation
The prediction factors obtained at steps 511-514 are used to calculate a likelihood that clouds will shade portions of the farm (step 521). A determination (step 525) is made as to the likelihood of cloud cover being less than a predetermined threshold. In the case a negative determination (at 525), i.e., cloud cover exceeds a predetermined threshold, a calculation is made (step 529) as to when the clouds will probably shade the farm.
This calculation (step 529) is used to calculate (step 533) loss of irradiance vs. time from clouds, integrated over the farm area, using the measurement of cloud size, shape and opacity (step 516). The calculation is:
I/t integrated over Areapv
where
I=irradiance
t=time and
AreaPV=farm area.
Upon obtaining the loss of irradiance vs. time (step 533), a stored cumulative I/t curve for previously measured clouds is retrieved (step 537). The current I/t curve is added to the historical one by updating the cumulative I/t curve with calculations from the new measurements (step 539). The updating (step 539) is used to extrapolate the irradiance trend into the future. Stored spectral and temperature data and stored irradiance dependence of farm output power are provided (steps 541, 543) and the stored data is used to calculate expected power vs. time (step 545) from the cumulative I/t curve retrieved in steps 537 and 539. An evaluation of magnitudes & rates of power changes (step 551), such as down-ramps and up-ramps, is made. A determination (step 561) is then made of whether the rates of change exceed predetermined thresholds, using evaluation 551 and a value for ramp rate 565 tolerated by the grid.
Determination 561 is used to determine whether to ignore the anticipated change (step 571) or issue a “REACT” decision 580 which can be used for notification, ramp-rate-control, storage, back up generation or load response/reduction in demand.
Reacting to Predicted Changes
If the calculations based on newly-acquired data is not likely to be accurate (step 643), historical data is used to apply correction factors (step 645) and the historical data is used to update the cumulative I/t curve with calculations from the new measurements (step 539). In the case of a determination (step 643) that calculations based on newly-acquired data is likely to be accurate, the new data is provided for the purpose of updating the cumulative I/t curve with calculations from the new measurements (step 539).
The determination is continued (step 925) until an increase in local sensor data, in which power is diverted or suppressed (step 927) to slow down the up-ramping. The slowing (step 927) may be accomplished by a number of techniques, including intentionally operating off Maximum Power Point (MPP).
In the case of a down power ramp determination (at determination step 913), a calculation (step 941) is made of an earlier time to begin a gradual ramp-down. The calculation is sent (step 943) to the farm operator. The farm operator may be human, or alternatively, the farm operator may be a SCADA module 820 or other computerized system.
In the case of a down power ramp determination (at determination step 1013), a calculation (step 1041) is made of an earlier time to begin a gradual ramp-down. The calculation is sent (step 1043) to the farm operator. The farm operator may be human, or alternatively, the farm operator may be SCADA module 820 or other computerized system.
Sensing and Responding to Changed Power Output Conditions
In the event that the up-ramp is not detected (step 1113) or the up-ramp does not exceed the grid limit, the system continues to monitor power output (step 1121).
Using these techniques, power stations can dynamically respond to weather-related effects that change their output power, thus maintaining the desired stability of power to the grids they supply. With this capability, clean and renewable, but inherently intermittent and weather-sensitive, power sources such as solar and wind farms can mitigate the power output fluctuations that currently make them incompatible with smaller or less-flexible existing power grids. These techniques also enable such sources to sell scheduled power as well as the intermittent energy they normally provide. Further, with these predictive techniques, energy from intermittent and non-intermittent sources could be supplied to a grid through the same interconnection point.
The techniques and modules described herein may be implemented by various means. For example, these techniques may be implemented in hardware, software, or a combination thereof. For a hardware implementation, the processing units within an access point or an access terminal may be implemented within one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, micro-controllers, microprocessors, other electronic units designed to perform the functions described herein, or a combination thereof.
For a software implementation, the techniques described herein may be implemented with modules (e.g., procedures, functions, and so on) that perform the functions described herein. The software codes may be stored in digital storage media, memory units and executed by processors or demodulators. The memory unit may be implemented within the processor or external to the processor, in which case it can be communicatively coupled to the processor via various means.
It will be understood that many additional changes in the details, materials, steps and arrangement of parts, which have been herein described and illustrated to explain the nature of the subject matter, may be made by those skilled in the art within the principle and scope of the invention as expressed in the appended claims.
