JP2018007347A - Performance evaluation method of photovoltaic power generation - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an evaluation method of photovoltaic power generation through the grasp of power generation performance by the slide of a learning period.SOLUTION: The performance of photovoltaic power generation is evaluated based on a difference between a first estimated power generation amount at maintained initial power generation performance, which is obtained through parameter learning from data obtained by measuring a meteorological condition and a power generation amount related to power generation per a content time in a constant period after the start of operation, and a second estimated power generation amount, which is obtained through parameter learning from similar data to the above, under the same condition of the meteorological condition related to the first estimated power generation amount data, in a most recent constant period after the lapse of the period after the start of operation.SELECTED DRAWING: Figure 1

Description

The present invention relates to a method for quantitatively grasping a long-term change in power generation performance in a solar power plant and evaluating the performance of solar power generation.

As with thermal power generation and nuclear power generation, it is easy to measure the initial power generation amount and the current power generation amount with a constant input of raw materials, and compare these to evaluate the performance of the current power generation equipment. However, with regard to solar power generation, the amount of power generation changes due to changes in the irradiation intensity of the sun, which is the raw material supplier, and the outside air temperature, so the performance is evaluated simply by comparing the initial power generation amount with the current power generation amount. It ’s difficult.

On the other hand, in order to ensure safety and improve business profits in the operation of large-scale solar power plants, it is necessary to grasp the operating status through power plant monitoring. Therefore, it is difficult to properly grasp the operating status by simply performing monitoring because it is greatly affected by the surrounding environmental conditions such as the ambient temperature and the specific conditions such as the size and configuration of the power plant.

Thus, many methods have been devised for estimating the amount of power generation from ambient environment data such as the amount of solar radiation and outside temperature, and facility information of the power plant, but sufficient accuracy cannot be obtained. Or, in order to obtain accuracy, there are problems such as providing as many measuring facilities as possible and performing detailed measurement and analysis by experts.

In the “power generation amount prediction device and method” of Patent Document 1, estimation is performed by estimating the weather condition at the site where the solar power generation system is installed from the system coefficient and the power generation experiment data for each of the solar power generation systems. A weather condition estimation unit for obtaining a weather condition value, a weather condition space correction unit for obtaining a corrected weather condition value by correcting the estimated weather condition value according to an installation position of each of the solar power generation systems, and the sun For each photovoltaic power generation system, based on the corrected weather condition value and the actual power generation data, a parameter learning unit that updates the system coefficient, and for each photovoltaic power generation system, based on the reference weather data and the system coefficient, A power generation prediction unit that performs power generation amount prediction.

JP 2014-63372 A

In Patent Document 1, the parameters of a predetermined power generation amount estimation model are learned from past performance data (sunlight, outside temperature, power generation amount, etc.) of a power plant to be estimated, and subsequent power generation is performed using the parameters. The estimation accuracy is autonomously improved by estimating the quantity. In addition, in the support system for power generation estimation learning in photovoltaic power generation, it is possible to compress the data necessary for learning and to change the learning conditions flexibly by maintaining the matrix generated during the learning process. Has been.

In such a background, it is generally said that a photovoltaic power generation system has a deterioration of about 0.5 to 1% / year. Therefore, by using the measurement data at the initial stage of the operation of the solar power plant, learning the parameters of the power generation amount estimation model as described in Patent Document 1, and using the parameters, as shown in FIG. Estimate the power generation amount when the initial power generation performance is maintained.

In FIG. 10, the array capacity of the photovoltaic power plant to be evaluated and the capacity of the power conditioner (PCS) are input in advance, the solar radiation intensity and the outside air temperature are measured, the PCS power generation amount and the interconnection point power are measured, The parameters of the power generation amount estimation model are learned, and the subsequent power generation amounts are estimated using the parameters. In this way, it is possible to estimate “the amount of power generated when the power plant maintains the initial power generation performance”, but that alone estimates the amount of power generation including degradation when time has elapsed since the start of operation. It is not possible.

As a countermeasure to this problem, as shown in FIG. 11, a deterioration rate over time is assumed in advance, and the estimated power generation amount including the deterioration is calculated by multiplying the estimated power generation amount in the initial performance by the deterioration rate. However, in this case, since the set deterioration rate does not always coincide with the actual deterioration rate, the estimation accuracy may be reduced.

In addition, since the amount of power generated by the solar power generation system depends on weather conditions such as solar radiation and temperature, for example, when trying to compare “total power generation for one year” in different years, the weather conditions in each year are naturally Since they are not identical, they cannot be compared as they are.

Although it is possible to eliminate the influence of the difference in weather conditions by estimating the power generation amount in the initial performance by the method shown in FIG. 10 and looking at the ratio of the actual power generation amount to each estimated power generation amount, It is not possible to eliminate errors in case of different operating conditions, such as a power plant shutdown. In order to eliminate the influence of this operating situation, it is necessary to make a comparison after extracting data only when the power plant is fully operating based on the operating history of the power plant. This work is very complicated, and there is a problem that the number of target data is reduced.

In view of this, an object of the present invention is to provide a solar power generation evaluation method by grasping a change in power generation performance due to a slide in a learning period.

The invention of claim 1 is a case where initial power generation performance obtained by learning parameters from data obtained by measuring meteorological condition values and power generation amount for a certain period of time for a certain period after the start of operation is maintained. Second estimation obtained by learning parameters from the first estimated power generation amount and data similar to the above under the same weather conditions as the data of the first estimated power generation amount for the most recent fixed period after the passage of the period It was set as the performance evaluation method of the solar power generation which evaluates the performance of the said solar power generation by the difference with the electric power generation amount.

According to a second aspect of the present invention, the first estimated power generation amount and the second estimated power generation amount are obtained by measuring a weather condition value and a power generation amount related to power generation in the solar power generation system over a certain period of time. Are stored as data, and inappropriate power generation data determined in advance are excluded from these, and the remaining power generation data is used to update this sequentially to learn the estimation parameters. The photovoltaic power generation performance evaluation method according to claim 1, which is determined and calculated from the estimated parameters in consideration of weather condition values related to the power generation.

Further, in the invention of claim 3, the power generation amount data determined in advance as the inappropriate power generation amount data includes power generation amount data at a time when the solar altitude is low, power generation amount data below a solar radiation intensity threshold value, Using the estimation parameters calculated from the remaining data excluding the power generation data at the time when the solar altitude is low and the power generation data below the solar radiation intensity threshold for the measurement data of the power generation during the period, Estimate the power generation amount using the estimated model parameters calculated from the power generation amount data at the time when the estimated power generation amount and the error of the measured data deviate more than a certain value, and the actual data of other structurally similar power plants 3. The method for evaluating the performance of photovoltaic power generation according to claim 2, wherein the estimation method is one or a plurality of power generation data at a time when an error between the estimated power generation and the measured data deviates by a predetermined value or more. And the.

According to the first to third aspects of the invention, the weather condition value and the power generation amount related to the power generation for a certain period at the start of the operation of the photovoltaic power generation are measured over a certain period of time and accumulated as data, Using multiple regression analysis that sequentially updates these data and learns the estimated parameters, the first estimated power generation amount based on the estimated parameters determined thereby, and the second estimated power generation amount obtained in the same manner after the predetermined period of time, By comparing these, the power generation performance deterioration status of the entire power plant can be quantitatively grasped. Therefore, it is possible to grasp a change in pure power generation performance that is not affected by weather conditions and operating conditions, and more accurate power generation performance evaluation is possible. Moreover, it is possible to diagnose deterioration (including failure) of the entire solar power plant by applying long-term operation data. If the measurement range can be subdivided as part of the power plant performance, partial degradation can be grasped by the same method.

Further, according to the second and third aspects of the invention, the weather condition value and the power generation amount related to the power generation in the solar power generation system are measured over a certain period of time and accumulated as data. From this, the inappropriate weather condition value data and power generation amount data determined in advance are excluded, and the remaining weather condition value data and power generation amount data are sequentially updated using the remaining weather condition value data and power generation amount data, thereby learning the estimation parameter. Therefore, it is possible to perform parameter learning using only measurement data when the power plant exhibits its original performance. In addition, by analyzing the exclusion status of inappropriate data and the learning status of estimated parameters, it is possible to optimize the data exclusion condition itself in parameter learning. As a result, a more accurate estimated power generation amount can be obtained. Therefore, the power generation performance deterioration status of the entire power plant can be grasped more accurately.

It is a schematic block diagram of the power generation performance evaluation method of Embodiment 1 of this invention. It is a block block diagram of the power generation performance evaluation method of Embodiment 1 of this invention. It is a graph which shows the power generation performance change by the power generation performance evaluation method of Embodiment 1 of this invention. It is a block diagram which shows the measurement outline | summary of the estimation electric power generation amount in the electric power generation performance evaluation method of Embodiment 1 of this invention. It is a block diagram which shows the computer system used for measurement of the estimated electric power generation amount in the electric power generation performance evaluation method of Example 1 of this invention. It is a graph which shows the exclusion range of the inappropriate data by time used for the measurement of the estimated electric power generation amount in the electric power generation performance evaluation method of Embodiment 1 of this invention. It is a graph which shows the exclusion range of the inappropriate data by solar radiation intensity used for the measurement of the estimation electric power generation amount in the electric power generation performance evaluation method of Embodiment 1 of this invention. It is a graph which shows the exclusion range of the improper data by the learning on the day used for the measurement of the estimated electric power generation amount in the electric power generation performance evaluation method of Embodiment 1 of this invention. It is a graph which shows the example of exclusion of the improper data by other power plant parameters used for measurement of the estimated power generation amount in the power generation performance evaluation method of Embodiment 1 of this invention. It is the schematic which shows the example of learning of an electric power generation amount estimation parameter. It is the schematic which shows the calculation method of the estimated electric power generation amount also including degradation.

(Embodiment 1)
Hereinafter, a method for evaluating the performance of photovoltaic power generation according to Embodiment 1 of the present invention will be described with reference to the drawings.

As shown in FIG. 1 and FIG. 2, the present invention measures weather condition values relating to power generation at regular time intervals for one year after the start of operation, for example, solar radiation intensity, outside air temperature, and power generation amount, and learns estimated parameters. The unit 3 learns an estimation parameter from these data, and obtains a power generation amount estimation parameter 4. Then, the first estimated power generation amount 1 when the initial power generation performance is maintained is obtained through the power generation amount estimation unit 5.

In addition, after the above period has elapsed, weather condition values relating to power generation every fixed time in the most recent year, for example, solar radiation intensity, outside air temperature, and power generation amount are measured, and these data are estimated by the estimation parameter learning unit 3 ′. From this, the estimation parameter is learned, and the power generation amount estimation parameter 4 ′ is obtained. Then, the latest second estimated power generation amount 2 is obtained through the power generation amount estimation unit 5 ′.

In these power generation amount estimation units 5 and 5 ', data of the same weather conditions such as solar radiation intensity and outside temperature are extracted, and "estimated power generation amount when the power plant has maintained the initial power generation performance" and "most recent 1 "Estimated power generation based on annual power generation performance" is calculated. For this reason, the difference between the two power generation amounts indicates the difference in power generation performance between the initial year and the most recent one year. By applying this over a long period of time, It is possible to quantitatively grasp the power generation performance deterioration status.

FIG. 3 shows changes in power generation performance using the evaluation method of the present invention in an actual solar power plant. This estimates the power generation amount for each day using the parameters for the initial one year and the parameters for the most recent one year, and compares the accumulated movement for one year for each estimated power generation amount. This power plant has seen a decline in power generation performance of less than 2% in two and a half years from the start of operation. This is consistent with the generally-known deterioration rate “0.5 to 1%”, and it was confirmed that the power generation performance degradation (deterioration) of the power plant can be accurately grasped by the evaluation method of the present invention.

In this graph, the power generation performance does not decrease during the initial year, but decreases after the lapse of one year, but actually decreases even during the initial year. Also, within two years from the start of operation, the effect of the initial one year parameters remains, and it is difficult to say that the actual deterioration rate is correctly determined. More accurate degradation can be determined after two years have elapsed since the start of operation.

In the first embodiment, the data is based on data for one year from the start of operation of the power plant and the most recent one year. However, the present invention is not limited to one year, such as three months at the same time in one year. Data for a certain period of time may be used.

Next, acquisition of the estimated power generation amount used in the solar power generation performance evaluation method will be described.

FIG. 4 is a block diagram showing an outline of instrumentation of the estimated power generation amount. As shown in FIG. 4, as shown in FIG. 4, a solar radiation meter and a thermometer provided at the power plant for the first year of the start of operation of the solar power plant. Then, a weather condition value related to power generation such as solar radiation intensity and outside temperature and a power generation amount are measured every minute for a certain period of time by using an ammeter or the like (step S-1). These measured values are stored as data in the storage unit of the computer. At that time, the array capacity and PCS capacity of the photovoltaic cells of the power plant are also input to the storage unit as the facility information.

Next, data inappropriate for parameter learning to be described later is excluded from the power generation amount data in the accumulated data (step S-2). This is selected by comparison with a predetermined time, threshold value or the like. Parameter learning is performed based on the data remaining by this processing (step S-3). In this, the above-described measurement, data accumulation by the measurement, data exclusion, etc. are performed for one year, and during that time, data etc. are sequentially updated to perform parameter learning.

Thereafter, an estimated parameter is determined based on parameter learning (step S-4), and an estimated power generation amount is calculated in consideration of the estimated parameter and data on weather condition values related to power generation such as solar radiation intensity and outside temperature. (Step S-5).

Each of the above steps is a computer system comprising measuring means, storage means for storing each measured value as data, data excluding means for excluding inappropriate data, parameter learning means, estimated parameter determining means, and estimated power generation amount calculating means Can be realized using. And the said step effect | action by each of these structure means can be implement | achieved as a computer program module, for example, and each function can be implement | achieved in a computer system by the program containing each program module.

As shown in FIG. 5, the computer system includes a CPU 11 for executing program instructions, a main storage device 12 such as a memory, an external storage device 13 such as a hard disk, a magnetic disk device or a magneto-optical disk device, a data input device 14, A display device 15 and a bus 16 for connecting them to each other are provided. The program is stored in the external storage device 13, and the CPU 11 expands the program in the main storage device 12, and sequentially reads and executes the expanded program.

Next, in order to exclude the inappropriate data without excess or deficiency, the following processing is performed. This will be described with reference to FIGS.

(1) Excluded by solar altitude or time The amount of generated power at the time when the solar altitude is low and the power generation is reduced due to shadows created by surrounding structures and adjacent solar cell arrays is not suitable for parameter learning. Excludes data for times when the solar altitude is below a certain level. The solar altitude can be calculated from the date and time information, but here, as a simple method, the power generation amount data in the early morning and evening hours when the solar altitude is low is excluded.

An example of exclusion by time is shown in FIG. The vertical axis on the left of the graph in FIG. 6 represents solar radiation intensity (kW / m ² ), the vertical axis on the right represents the amount of power generation (kW), and the horizontal axis represents the time of the day. Here, the data from 0 o'clock to less than 9 o'clock and from 15 o'clock to 24 o'clock is excluded, and only the data from 9 o'clock to 15 o'clock is left and used for parameter learning.

(2) Exclusion by the solar radiation intensity threshold Even if the above case (1) does not apply, if the solar radiation intensity is extremely low, the conversion efficiency of the PCS (power conditioner) is greatly reduced. Because the load on the air conditioning system of the PCS is reduced, the operation is performed under conditions different from the case where the solar radiation intensity exceeds a certain level, which is inappropriate for parameter learning. Therefore, a solar radiation intensity threshold value (for example, 0.2 kW / m ² ) is provided, and power generation amount data at a time when the measured solar radiation intensity is equal to or less than the threshold value is excluded.

An example of exclusion by solar radiation intensity is shown in FIG. The vertical axis on the left of the graph of FIG. 7 represents the solar radiation intensity (kW / m ² ), the right vertical axis represents the amount of power generation (kW), and the horizontal axis represents the time of the day. Here, only the data when the solar radiation intensity is 0.2 kW / m ² or more is left and used for parameter learning. The solar radiation intensity threshold is not limited to 0.2 kW / ^{m 2,} it may be a 0.1 kW / ^{m 2} or 0.05 kW / ^{m 2,} such as arbitrary threshold.

(3) Exclusion by comparison with estimated amount after learning only for the relevant day Even if the conditions of (1) and (2) above are not met, power generation stops due to power system accident or work stoppage The data at that time is inappropriate for parameter learning.

Therefore, for the measurement data for a certain period (example: 1 day), multiple regression analysis is performed after removing inappropriate data under the conditions (1) and (2), and the calculated power generation amount estimation parameter is used. If the error between the estimated power generation amount and the measured data is more than a certain value (eg 10% of the estimated value), the operating condition of the power plant is abnormal. It was decided to exclude the data at the corresponding time.

An example of exclusion by learning on the day is shown in FIG. The vertical axis on the left of the graph in FIG. 8 represents the solar radiation intensity (kW / m ² ), the vertical axis on the right represents the amount of power generation (kW), and the horizontal axis represents the time of the day (however, in the right graph, the amount of power generation) Is omitted). Here, as shown in the graph on the right, since the error between the estimated power generation amount and the actually measured power generation amount is more than a certain value, it is assumed that there was an abnormality in the operating state of the power plant. The data at 16:00 was excluded.

(4) Exclusion by comparison with estimated amount after learning based on the results of similar power plants If the operation of the power plant is temporarily abnormal within a certain period due to the condition of (3) above, It is possible to exclude the data of the time zone, but if the power generation amount decreases uniformly over the entire period due to the initial failure of the solar cell panel, forgetting to turn on the DC circuit, PCS stop, etc., ( The method 3) cannot determine suitability.

Therefore, the power generation amount is estimated using estimated model parameters calculated from the actual data of other power plants that are structurally similar, and the error between the estimated power generation amount and the measured data is a constant value (eg, the estimated value 10%) or more, it was decided to exclude the relevant data because there was an abnormality in the initial state or operation state of the power plant.

An example of exclusion by using parameters of other power plants is shown in FIG. The graph on the left in FIG. 9 shows the power generation performance of the power plant, and the power generation amount decreased all day due to the PCS stoppage.
The right vertical axis represents the solar radiation intensity (kW / m ² ), the right vertical axis represents the power generation amount (kW), and the horizontal axis represents the time of the day. The amount of solar radiation is represented by a solid line, and the curve with the apex at the location above the arrow indicates the amount of power that should be generated according to the amount of solar radiation. The curve with the apex at the location below the arrow indicates the actual power generation Indicates the amount. The graph on the right side of FIG. 9 shows that the largest mountain curve is the estimated power generation using the parameters of other power plants, and the lower mountain curve is the actual power generation and learning on the day of the power plant. Indicates the estimated power generation amount.

With the above measures, it has become possible to perform parameter learning using only the measurement data when the power plant exhibits its original performance. The exclusion conditions (3) and (4) above can also be used as a function to perform preliminary power generation status diagnosis by adjusting the pass / fail judgment threshold and to output an alarm when there is an abnormality. is there.

In addition, in a certain power plant (or part of it) on the system regarding “what time data was excluded as inappropriate due to which conditions”, “change in power generation estimation accuracy when the exclusion conditions were changed”, etc. By performing the analysis, it is possible to autonomously optimize the data exclusion method itself at the corresponding location.

This is, for example, totaling “which time data was excluded by the above condition (3) or (4)”, and based on the result, the above condition (1) (the time to be excluded) ).

In addition, optimization adjustments such as comparing the power generation amount estimation accuracy when the condition (2) (irradiance intensity threshold) is changed and adopting the condition with the highest accuracy result can be performed.

DESCRIPTION OF SYMBOLS 1 1st estimated electric power generation amount 2 2nd estimated electric power generation amount 3, 3 'Estimation parameter learning part 4, 4' Electricity generation amount estimation parameter 5, 5 'Electricity generation amount estimation part
11 CPU 12 Main storage device 13 External storage device 14 Input device 15 Display device 16 Bus
S-1 Data measurement S-2 Data exclusion process S-3 Parameter learning S-4 Estimated parameters S-5 Estimated power generation

Claims (3)

  A first estimated power generation amount when the initial power generation performance is maintained, obtained by learning parameters from data obtained by measuring meteorological condition values and power generation amount for a certain period of time after the start of operation; Due to the difference from the second estimated power generation amount obtained by learning the parameters from the same data as the above under the same weather conditions as the data of the first estimated power generation amount during the most recent fixed period after the period has elapsed. A method for evaluating the performance of photovoltaic power generation, comprising evaluating the performance of photovoltaic power generation.   The first estimated power generation amount and the second estimated power generation amount are obtained by measuring weather condition values and power generation amounts related to power generation in the photovoltaic power generation system over a certain period of time and accumulating them as data. From this, the inappropriate power generation amount data determined in advance is excluded, and the remaining power generation amount data is sequentially updated to learn the estimation parameter, thereby determining the estimation parameter, and the generation parameter is determined from the estimation parameter. The performance evaluation method for photovoltaic power generation according to claim 1, wherein the calculation is performed in consideration of a weather condition value related to. The power generation data determined in advance to be inappropriate power generation data includes power generation data at times when the solar altitude is low, power generation data below a threshold of solar radiation intensity, and power generation measurement data for a certain period. As described above, the power generation amount for the period is estimated using the estimation parameters calculated from the remaining power generation data excluding the power generation amount data at a time when the solar altitude is low and the solar radiation intensity threshold value or less. Estimate the amount of power generation using the estimated model parameters calculated from the power generation amount data at the time when the data error deviates by more than a certain value and the actual data of other power plants that are structurally similar. The method for evaluating the performance of photovoltaic power generation according to claim 2, characterized in that it is one or a plurality of power generation amount data at a time when an error from the data deviates by a certain value or more.

Images