[](https://mybinder.org/v2/gh/MauricioSalazare/tensorpowerflow/master?urlpath=lab/tree/examples)
[](https://github.com/MauricioSalazare/tensorpowerflow/blob/master/LICENSE)
[](https://pypi.python.org/pypi/tensorpowerflow/)
[](https://pypi.python.org/pypi/tensorpowerflow/)
[](https://github.com/MauricioSalazare/tensorpowerflow)
[](https://pypi.python.org/pypi/tensorpowerflow/)
[](https://github.com/MauricioSalazare/tensorpowerflow/actions)
# TensorPowerFlow
## What is TensorPowerFlow?
An ultra-fast power flow based on a fixed-point iteration algorithm. The power flow is intended for applications requiring massive
amounts of power flow computations. e.g., electrical load time series, metaheuristics, and electrical grid
environments for reinforcement learning.
## How to install
The package can be installed via pip using:
```shell
pip install tensorpowerflow
```
## Example:
Here we show four examples of what can be done with the packages:
1. *Example 1:* The package comes with a preloaded network case of 34 nodes that can be solved by default.
2. *Example 2:* Solve 10.000 PF for the default network case. The active power for the nodes is generated by sampling a Normal
distribution.
3. *Example 3:* The GridTensor class can generate a random grid using the `GridTensor.generate_from_graph()` method.
The parameters `nodes` and `child`, respectively, can control the number of nodes and branching per node.
4. *Example 4:* We test the grid with a tensor of 3 dimensions. The first dimension is the number of scenarios
(10 in the example). The second dimension is the number of time steps (8.760 to simulate a 30-minute resolution
consumption for one year). The third dimension is the number of PQ nodes in the grid (33 PQ nodes for the base
grid case).
```python
from tensorpowerflow import GridTensor
import numpy as np
from time import perf_counter
#%% Example 1: Solve base case (34 node bus)
network = GridTensor(gpu_mode=False)
solution = network.run_pf()
print(solution["v"])
#%% Example 2: Solve 10_000 power flows on the 34 node bus network case.
network_size = network.nb - 1 # Size of the network without the slack bus.
active_ns = np.random.normal(50, scale=1, size=(10_000, network_size)) # Power in kW
reactive_ns = active_ns * 0.1 # kVAr
solution_tensor = network.run_pf(active_power=active_ns, reactive_power=reactive_ns)
print(solution_tensor["v"])
#%% Example 3: Generate a random radial network of 100 nodes and a maximum of 1 to 3 branches per node.
network_rnd = GridTensor.generate_from_graph(nodes=100, child=3, plot_graph=True)
solution_rnd = network_rnd.run_pf()
print(solution_rnd["v"])
#%% Example 4: Solve a tensor power flow. For 10 scenarios, 8_760 time steps (one year - 1 hr res) for the 33 PQ nodes.
# Meaning that the dimensions of the tensor are (10, 8_760, 33)
network = GridTensor(numba=True) # Loads the basic 34 bus node network.
active_ns = np.random.normal(50, # Power in kW
scale=10,
size=(10, 8_760, 33)).round(3) # Assume 1 slack variable
reactive_ns = (active_ns * .1).round(3) # Constant PF of 0.1
start_tensor_dense = perf_counter()
solution = network.run_pf(active_power=active_ns, reactive_power=reactive_ns, algorithm="tensor")
t_tensor_dense = perf_counter() - start_tensor_dense
assert solution["convergence"], "Algorithm did not converge."
assert solution["v"].shape == active_ns.shape
print(f"Time tensor dense: {t_tensor_dense:.3f} sec.")
```
More examples can be found in the [examples folder](examples) (under development).
Also, you can try the package via Jupyter lab by clicking on the binder icon:
[](https://mybinder.org/v2/gh/MauricioSalazare/tensorpowerflow/master?urlpath=lab/tree/examples)
## Reading and citations:
The mathematical formulation of the power flow can be found at:
> *"Tensor Power Flow Formulations for Multidimensional Analyses in Distribution Systems."* E.M. Salazar Duque,
Juan S. Giraldo, Pedro P. Vergara, Phuong H. Nguyen, and Han (J.G.) Slootweg. [arXiv:2403.04578 (2024)](https://arxiv.org/pdf/2403.04578).
## How to contact us
For any questions, suggestions or collaborations, contact Juan S. Giraldo at <jnse@ieee.org>
Raw data
{
"_id": null,
"home_page": null,
"name": "tensorpowerflow",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.0",
"maintainer_email": null,
"keywords": "power flow, power systems, time series loading",
"author": null,
"author_email": "Mauricio Salazar <e.m.salazar.duque@tue.nl>",
"download_url": "https://files.pythonhosted.org/packages/7c/6d/b275bdbfe2ece3924bc35d863995ee54e550761ace83de0f2173dd21e087/tensorpowerflow-0.0.7.tar.gz",
"platform": null,
"description": "[](https://mybinder.org/v2/gh/MauricioSalazare/tensorpowerflow/master?urlpath=lab/tree/examples)\n[](https://github.com/MauricioSalazare/tensorpowerflow/blob/master/LICENSE)\n[](https://pypi.python.org/pypi/tensorpowerflow/)\n[](https://pypi.python.org/pypi/tensorpowerflow/)\n[](https://github.com/MauricioSalazare/tensorpowerflow)\n[](https://pypi.python.org/pypi/tensorpowerflow/)\n[](https://github.com/MauricioSalazare/tensorpowerflow/actions)\n\n\n# TensorPowerFlow\n\n## What is TensorPowerFlow?\nAn ultra-fast power flow based on a fixed-point iteration algorithm. The power flow is intended for applications requiring massive\namounts of power flow computations. e.g., electrical load time series, metaheuristics, and electrical grid\nenvironments for reinforcement learning.\n\n## How to install\n\nThe package can be installed via pip using:\n\n```shell\npip install tensorpowerflow\n```\n\n## Example:\n\nHere we show four examples of what can be done with the packages:\n1. *Example 1:* The package comes with a preloaded network case of 34 nodes that can be solved by default. \n2. *Example 2:* Solve 10.000 PF for the default network case. The active power for the nodes is generated by sampling a Normal \n distribution.\n3. *Example 3:* The GridTensor class can generate a random grid using the `GridTensor.generate_from_graph()` method.\n The parameters `nodes` and `child`, respectively, can control the number of nodes and branching per node.\n4. *Example 4:* We test the grid with a tensor of 3 dimensions. The first dimension is the number of scenarios \n (10 in the example). The second dimension is the number of time steps (8.760 to simulate a 30-minute resolution \n consumption for one year). The third dimension is the number of PQ nodes in the grid (33 PQ nodes for the base \n grid case).\n\n```python\nfrom tensorpowerflow import GridTensor\nimport numpy as np\nfrom time import perf_counter\n\n#%% Example 1: Solve base case (34 node bus)\nnetwork = GridTensor(gpu_mode=False)\nsolution = network.run_pf()\nprint(solution[\"v\"])\n\n#%% Example 2: Solve 10_000 power flows on the 34 node bus network case.\nnetwork_size = network.nb - 1 # Size of the network without the slack bus.\nactive_ns = np.random.normal(50, scale=1, size=(10_000, network_size)) # Power in kW\nreactive_ns = active_ns * 0.1 # kVAr\nsolution_tensor = network.run_pf(active_power=active_ns, reactive_power=reactive_ns)\nprint(solution_tensor[\"v\"])\n\n#%% Example 3: Generate a random radial network of 100 nodes and a maximum of 1 to 3 branches per node.\nnetwork_rnd = GridTensor.generate_from_graph(nodes=100, child=3, plot_graph=True)\nsolution_rnd = network_rnd.run_pf()\nprint(solution_rnd[\"v\"])\n\n#%% Example 4: Solve a tensor power flow. For 10 scenarios, 8_760 time steps (one year - 1 hr res) for the 33 PQ nodes.\n# Meaning that the dimensions of the tensor are (10, 8_760, 33)\n\nnetwork = GridTensor(numba=True) # Loads the basic 34 bus node network.\nactive_ns = np.random.normal(50, # Power in kW\n scale=10,\n size=(10, 8_760, 33)).round(3) # Assume 1 slack variable\nreactive_ns = (active_ns * .1).round(3) # Constant PF of 0.1\n\nstart_tensor_dense = perf_counter()\nsolution = network.run_pf(active_power=active_ns, reactive_power=reactive_ns, algorithm=\"tensor\")\nt_tensor_dense = perf_counter() - start_tensor_dense\nassert solution[\"convergence\"], \"Algorithm did not converge.\"\nassert solution[\"v\"].shape == active_ns.shape\nprint(f\"Time tensor dense: {t_tensor_dense:.3f} sec.\")\n\n\n```\n\nMore examples can be found in the [examples folder](examples) (under development).\nAlso, you can try the package via Jupyter lab by clicking on the binder icon:\n\n[](https://mybinder.org/v2/gh/MauricioSalazare/tensorpowerflow/master?urlpath=lab/tree/examples)\n\n## Reading and citations:\n\nThe mathematical formulation of the power flow can be found at:\n\n> *\"Tensor Power Flow Formulations for Multidimensional Analyses in Distribution Systems.\"* E.M. Salazar Duque,\nJuan S. Giraldo, Pedro P. Vergara, Phuong H. Nguyen, and Han (J.G.) Slootweg. [arXiv:2403.04578 (2024)](https://arxiv.org/pdf/2403.04578).\n\n## How to contact us\n\nFor any questions, suggestions or collaborations, contact Juan S. Giraldo at <jnse@ieee.org>\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "Ultra fast power flow based in Laurent series expansion.",
"version": "0.0.7",
"project_urls": {
"repository": "https://github.com/MauricioSalazare/tensorpowerflow"
},
"split_keywords": [
"power flow",
" power systems",
" time series loading"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "215957d08e8c3f1d077f164d678f9f785b445a3495d5417ec3e1774a1cdf9892",
"md5": "acc3afb15aaac1c95fe75f8908159c8a",
"sha256": "2ffb3f1871588ca7b9a3125e4aa2f5db6902b0267aa5d75f32c324cc5cceb94d"
},
"downloads": -1,
"filename": "tensorpowerflow-0.0.7-py3-none-any.whl",
"has_sig": false,
"md5_digest": "acc3afb15aaac1c95fe75f8908159c8a",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.0",
"size": 31270,
"upload_time": "2024-04-17T15:28:09",
"upload_time_iso_8601": "2024-04-17T15:28:09.813083Z",
"url": "https://files.pythonhosted.org/packages/21/59/57d08e8c3f1d077f164d678f9f785b445a3495d5417ec3e1774a1cdf9892/tensorpowerflow-0.0.7-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "7c6db275bdbfe2ece3924bc35d863995ee54e550761ace83de0f2173dd21e087",
"md5": "c58697e68dc2c9cf8c89a381ded7da52",
"sha256": "f6ec888a61cd50c025e5f8bbb99460ccd0628886e1242acb1ba9aa0b30804ddf"
},
"downloads": -1,
"filename": "tensorpowerflow-0.0.7.tar.gz",
"has_sig": false,
"md5_digest": "c58697e68dc2c9cf8c89a381ded7da52",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.0",
"size": 27763,
"upload_time": "2024-04-17T15:28:11",
"upload_time_iso_8601": "2024-04-17T15:28:11.648056Z",
"url": "https://files.pythonhosted.org/packages/7c/6d/b275bdbfe2ece3924bc35d863995ee54e550761ace83de0f2173dd21e087/tensorpowerflow-0.0.7.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-04-17 15:28:11",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "MauricioSalazare",
"github_project": "tensorpowerflow",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"requirements": [
{
"name": "numpy",
"specs": [
[
">=",
"1.22.4"
]
]
},
{
"name": "pandas",
"specs": [
[
">=",
"1.4.2"
]
]
},
{
"name": "matplotlib",
"specs": [
[
"<=",
"3.5.0"
]
]
},
{
"name": "scipy",
"specs": [
[
">=",
"1.8.0"
]
]
},
{
"name": "tqdm",
"specs": [
[
">=",
"4.64.0"
]
]
},
{
"name": "numba",
"specs": []
},
{
"name": "networkx",
"specs": [
[
">=",
"2.7.1"
]
]
},
{
"name": "seaborn",
"specs": [
[
">=",
"0.11.2"
]
]
},
{
"name": "pandapower",
"specs": [
[
">=",
"2.8.0"
]
]
},
{
"name": "mkl",
"specs": []
},
{
"name": "psutil",
"specs": []
}
],
"tox": true,
"lcname": "tensorpowerflow"
}
JSON
