- Posts: 8
Battery Automated Dispatch (Version 2020.2.29)
- Alk
21 Jan 2022 03:29 #10536
by Alk
Replied by Alk on topic Battery Automated Dispatch (Version 2020.2.29)
Hello again,
my previous post was a technical failure, so I am re-posting from the beginning for clarity (with minor corrections also).
I have an issue using the automated dispatch option in a standalone battery - PPA single owner model. I am using version 2021.12.2 but thought that this thread would be appropriate for my post, please let me know if I should start a new one.
To simulate a utility BESS (50MW - 2h capacity) earning revenue only from energy arbitrage, I'm setting the PPA price=1 and TOD factors as multipliers instead of using the Merchant Plant ( see "Cleared capacity Merchant Plant" ). Choosing automated dispatch -> perfect look-ahead -> 24 hours period, the "profit" (= revenue - electricity_purchase ) was much lower than I would expect. The "Electricity to/from grid" time-series showed that the system realized one cycle per day, maybe even less on
average. Looking at the average daily per-hour price of electricity, one
would expect a 2h system to make two full cycles per day on average (cycle degradation penalty already set to 0).
So I switched to manual dispatch under the following scheme, which indeed raised the profit significantly.
Not understanding why the crude manual dispatch gave better results, I returned to the automated dispatch running a parametric analysis with look-ahead period as the input and revenue, electricity_purchase as outputs. Results are shown below: a period of 6 hours gives the best outcome and from that point on any increase in the period gives a lower profit than the previous. The manual solution gave a profit of ~1.38E+06 (for reference).
From my understanding, higher look-ahead period should result in better optimized results but clearly this is not the case. Furthermore, changing the frequency to update dispatch variable gives exactly the same results. Am I getting something wrong here or is there indeed an issue in the algorithm?
After Paul's suggestion, I include the SOC time-series for manual (red), auto24 (orange) and auto6 (blue) and attach the .sam file.
my previous post was a technical failure, so I am re-posting from the beginning for clarity (with minor corrections also).
I have an issue using the automated dispatch option in a standalone battery - PPA single owner model. I am using version 2021.12.2 but thought that this thread would be appropriate for my post, please let me know if I should start a new one.
To simulate a utility BESS (50MW - 2h capacity) earning revenue only from energy arbitrage, I'm setting the PPA price=1 and TOD factors as multipliers instead of using the Merchant Plant ( see "Cleared capacity Merchant Plant" ). Choosing automated dispatch -> perfect look-ahead -> 24 hours period, the "profit" (= revenue - electricity_purchase ) was much lower than I would expect. The "Electricity to/from grid" time-series showed that the system realized one cycle per day, maybe even less on
average. Looking at the average daily per-hour price of electricity, one
would expect a 2h system to make two full cycles per day on average (cycle degradation penalty already set to 0).
So I switched to manual dispatch under the following scheme, which indeed raised the profit significantly.
Not understanding why the crude manual dispatch gave better results, I returned to the automated dispatch running a parametric analysis with look-ahead period as the input and revenue, electricity_purchase as outputs. Results are shown below: a period of 6 hours gives the best outcome and from that point on any increase in the period gives a lower profit than the previous. The manual solution gave a profit of ~1.38E+06 (for reference).
From my understanding, higher look-ahead period should result in better optimized results but clearly this is not the case. Furthermore, changing the frequency to update dispatch variable gives exactly the same results. Am I getting something wrong here or is there indeed an issue in the algorithm?
After Paul's suggestion, I include the SOC time-series for manual (red), auto24 (orange) and auto6 (blue) and attach the .sam file.
Attachments:
Please Log in or Create an account to join the conversation.
Moderators: pgilman